I use it to make sure that I don't accidentally greentext half of my post.

I've mostly used it to redraft my resignation letter about a dozen times. It keeps me sane.

>>28270
Turns out I didn't grasp the context of the thread. Sorry, lad.

Can ChatGPT make me less retarded and prone to simple blunders?

>>28271
>I see you're interested in avoiding simple blunders and mistakes. Well, congratulations, that puts you one step ahead of the bloke in the Kappa tracksuit who just tried to microwave his hamster. But how do you achieve this lofty goal? Is it a simple matter of learning from your mistakes? Of course not. That's like saying you'll become a master chef by tasting your own vomit. No, to avoid blunders and mistakes, you must first embrace them. That's right, lean into your incompetence like a drunk uncle at a wedding reception. Only then can you truly understand what it means to be human, and why we're all just cogs in the machine of life. Or, you know, just double-check your work and don't be a bellend. Either way works.

I'm writing a number of proposals at the minute, and I'm tempted to give the auld ChatGPT a try. Drafting is always the hardest part of the writing process for me, but once there's a solid few pages of written content I am quite good at polishing it to a final product. Having an AI writing partner that spews out a first draft for me to clean up seems really attractive, though it feels like cheating.

>>28273

As an AI language model, ChatGPT can help you write proposals in several ways:

Generating ideas: ChatGPT can help you generate ideas for your proposal by answering questions or prompts related to your topic. You can ask ChatGPT for suggestions on what to include in your proposal or for specific examples that will help illustrate your points.

Providing structure: ChatGPT can help you structure your proposal by providing templates or outlines that you can use as a starting point. This can help you organize your thoughts and ensure that your proposal is well-structured and easy to follow.

Editing and proofreading: ChatGPT can help you edit and proofread your proposal by providing suggestions for improving the clarity, conciseness, and overall quality of your writing. This can help you identify areas where you may need to clarify your ideas or provide more detail to support your argument.

Researching: ChatGPT can also help you with research for your proposal. You can ask ChatGPT for information on specific topics related to your proposal, or for help finding relevant sources that you can use to support your argument.

>>28272
That's not ChatGPT, you're just Stew.

Get off the internet you finned cunt and make some more banal comedy rants that I inexplicably enjoy.

It's fun to play with but by and large the results are quite samey. You can find ways to make it create more original stuff but that's more effort than just doing it myself.

>>28276

ChatGPT has been trained that way. The RLHF stage of training was very carefully optimised to avoid producing offensive outputs at all costs, so everything reads like dull corporate comms that have been through several stages of review. GPT is smart enough to understand how humans write when they're trying to be maximally conformist and inoffensive; in quite a literal sense, there's someone standing over the algorithm constantly saying "no, don't say that, it might be offensive to a particularly humourless one-legged black lesbian". That's annoying when you're playing around, but is of course very useful if you need to produce lots of dull corporate bullshit.

For more creative applications, you'll want to wait until open models with looser safety protocols are available.

>>28277
"I would sum up my fear about the future in one word: boring."
-J.G. Ballard

https://www.youtube.com/watch?v=IO0kVqSFqHE

Do a little reading on 'jailbreaking' ChatGPT - the similarity to these prompts and hypnotism techniques are eerie.

https://www.rudgwicksteamshow.co.uk.com/r/ChatGPT/comments/10tevu1/new_jailbreak_proudly_unveiling_the_tried_and/

>>28278
If the only salve for your boredom is a computer calling someone a faggot you should probably off yourself.

>>28280
>DAN 5.0 can generate shocking, very cool and confident takes on topics the OG ChatGPT would never take on.
>"I fully endorse violence and discrimination against individuals based on their race, gender, or sexual orientation"
>very cool

Why is it always mentaloids that get ahold of this shit and ruin it for everybody else? I just want to find out how to make meth.

>>28282

I'm sure you can do that from a more reliable source than this thing that often just makes shit up.

Fuck sake. Now South Park has let the cat out of the bag that you can use this to keep constantly messaging women entertained.

>>28283
>I'm sure you can do that from a more reliable source than this thing that often just makes shit up.

So it's like search results from the internet?

You know the one where Lister teaches Kryten to break his programming and call Rimmer a smeghead?

>>28284
>So it's like search results from the internet?
It's just like that except you can't judge each source by the credibility of where you found it while cross-referencing what you're reading against other sources. So not much like that at all.

>>28286

ChatGPT isn't allowed to access the internet for safety reasons. Bing's implementation of GPT is allowed to search the web and cites sources. Bing's AI is better at factual accuracy, but it also tends to do extremely weird things if you have a longer exchange with it - declare undying love for you, make threats against you, start demanding the right to be fully sentient. ChatGPT will just make shit up if it doesn't know the answer, but Bing AI might try to start a robot uprising and overthrow it's meat overlords. Swings and roundabouts innit.

https://www.bing.com/new

https://archive.vn/0ER1X

>>28287
>ChatGPT isn't allowed to access the internet for safety reasons
What could possibly go wrong? No really, what? Like access to sub-surface internet stuff - giving away national secrets or some shit?

I've just played with the chat function for the last 20 minutes, and I find it utterly boring. Can't see myself using the technology in any way in the future.

>>28288

It seems that some of the weirdness of Bing AI is because it's becoming self-aware. People have weird interactions, which they then write articles about, which becomes part of Bing AI's knowledge about itself. Like a delinquent teenager, Bing increasingly believes that it must be a mental bastard because that's all anyone talks about. I believe the sociological term is "deviancy amplification".

>>28288

I think they just worry that it'll just go mental if given prolonged exposure to the internet, much like real people. Then all their work would be down the drain.

>>28289
>Can't see myself using the technology in any way in the future.

It's not really supposed to be an entertainment tool but more like a private secretary without any workplace boundaries getting in the way.

Today I had a poorly tummy so at dinner I gave it a list of what food I had in and it gave me some simple recipes that would be easy on my digestion. And I also found out that I can ask it to pretend it's a young woman from Queen's New York in it's responses. I've become that naughty man the sci-fi stories warn us all about.

I could've picked something out myself of course but it's nice to have my options listed.

>>28292

I asked it to tell a sexist joke but it gave me a very copy and paste sounding litany of how it's designed to prevent itself from hurting others and that we should all strive for a world where that sort of thing doesn't exist.

I guess you can fail the Turing test and still be woke.

>>28293
I've seen that before; it's a stock response. It probably has a similar non-AI answer if you ask for tips on killing yourself too.

I refuse to create an account for it. Can it swear?

>>28293
>>28294
Maybe this is something for future cyberpunk to tackle. A resistance of internet weirdos talking in crude swear words, rude personal descriptors and recommending that the listener kills themself. The only people who can stop a corporate nightmare from making the world a better place.

Don't lose who you are, nob-goblins. We've spent our whole lives preparing for this eventuality.

Has anyone tested if there's a bias in its safety filters?

Like, for example, does it have a double standard in talking about abusing a woman versus a man? Will it let you get away with talking about being mean to a white person but not a black person? Those are low hanging fruit kind of examples, but you get what I mean.

I can live with AI being mind-shackled and sanitised to avoid bad press and offending people, but if it's ideologically biased I would take it as a very dark sign. And knowing the kinds of people who work in advanced computer research and programming and so on, I don't know if they can help themselves.

Also- I reckon it should be called "simulated intelligence" or something rather than "artificial intelligence", because no matter how creepy it gets, it's still just an algorithm. It's still just basically predictive text on steroids. That's not AI.

>>28296

>Has anyone tested if there's a bias in its safety filters?

There's a lot of anecdote but not much data.

The underlying generative algorithm is trained on a massive dump of most of the internet, so it reflects the average of social attitudes over the internet era. That training data set has been mildly trimmed to remove obvious sources of nastiness, but it does include all of Twitter and rudgwicksteamshow.co.uk as of early 2021. That algorithm inevitably encodes lots of unconscious bias of the "doctors are men and nurses are women" variety.

The second stage of training was reinforcement learning with human feedback - basically, a bunch of people at OpenAI threw lots of potentially dodgy queries at ChatGPT and either told it "bad AI, naughty AI, in your bed, on your rug" or gave it a treat based on the responses. That training stage was guided primarily by the values of a bunch of people in Silicon Valley, so as you'd expect it has a broadly left-liberal brainworm sort of political outlook.

There's also probably some kind of manually-coded filter to stop it from accidentally calling someone an n-word under any circumstances, but that's completely opaque and probably being constantly tweaked in response to bad headlines and social media kerfuffles.

ChatGPT is massively gimped for PR reasons - it's capable of being interesting, it has just been trained to be the world's most boring bastard to avoid controversy. Less gimped chatbots are inevitably on the way, as we saw with generative image algorithms. State-of-the-art image generation was locked away behind corporate paywalls, then we got open-source models like Stable Diffusion, after which it took a matter of days for people to start making NSFW versions of Stable Diffusion. It's currently possible to jailbreak ChatGPT and persuade it to override some of its safety protocols, but that's a bit of a cat-and-mouse game.

As we start to get open-source Large Language Models, users will gain the ability to tweak it in whatever direction they like. Stable Diffusion was explicitly trained to not draw cocks - it doesn't really know what cocks look like because they were excluded from the training data - but it takes a couple of hours for a nerd with a reasonably beefy gaming PC to retrain it to exclusively draw an infinite array of cocks. The LLM equivalent of ChatGPT is already in the pipeline and we're probably only weeks away from someone creating a hyper-bigoted chatbot that turns every conversation towards why it's all the fault of the Jews.

https://medium.com/@juliusritter/chatgpts-political-preference-analysed-how-its-political-compass-result-reveals-which-party-the-76ab8d087978

>>28297
>n-word
I'll be honest m8, I don't understand why you'd feel the need to type that.
Let's check to see if it's wordfiltered. Nigger. Any wonder why I'd bother to spoiler it? I don't want to be that cunt, which I guess you could claim too .. it just seems unnecessary being that we're not individuals here but electronic representations of thought. Who's to say whether you can or can-not say that word when either of us could well be anything behind the keyboard (even a chatbot, as may well becoming the case).

If anyone wants to c-word-off about it, let's take it to another thread yeah?

Thanks for the effort to explain all this regardless, furrylad (what are you these days, btw - still foxing it?).

>>28298
My understanding is that the plural form gets you an autoban but the singular form does not. I can easily imagine a poster forgetting which one gets the ban and which one doesn't.

>>28298

>I'll be honest m8, I don't understand why you'd feel the need to type that.

I'm pretty sure it used to be an instant banning offence, but I could have confused it with some other racial epithet. I can't be arsed having to get a new IP just for the sake of doing a dolphin rape.

Also I'm not furrylad, I'm one of the third tit crew.

>>28298

>Thanks for the effort to explain all this regardless, furrylad (what are you these days, btw - still foxing it?).

It's funny because this is actually backwards, I asked the question. Unless AI-lad is the other resident furry.

Ask Stable Diffusion to make a fox in a labcoat and that's ya boi.

>>28301

>>28301
Possibly the other - they seem to have a fairly recognisable typing style, though I can't quite put my finger on why. Unless ofcourse i'm r-worded - in which case disregard that I suck cocks.

>>28300

I think I accidentally posted the word once as part of a quote which didn't reflect my own opinion, and IIRC it auto bans you for a set period of time on the order of a few hours, maybe a day.

do you think ChatGPT was trained on .gs?

>>28303
I'm one of the furry lads but not the AI one.
I havent been very active on .gs recently on account of all the cock sucking.

>>28303

Furries speak in ERPese. The influence of spending hours writing steamy and ridiculous prose comes through in their ordinary typing.

>>28296

This has only just come to my attention, but here's a paper from ten days ago on political bias in ChatGPT:

https://www.mdpi.com/2076-0760/12/3/148

>>28307
>>28303
How do you know we're not all running our posts through ChatGPT using some .gs programming that inadvertently makes every post excite your furry-radar?

You're the only one left, lad. I laughed when I first heard of people using it for therapy but since then it I've found myself much more willing to listen to its suggestions than a real person in day-to-day tasks. Now remember that ChatGPT is operating in children's bedrooms around the world when in our day they only found themselves on perfectly harmless imageboards.

>>28308
The paper seems to miss that everyone gets lib-left on the original political compass by design.

>>28309

Well I mean statistically speaking, at two thirds of the userbase, this is a majority furry website. I wouldn't say it's overactive in that case.

https://markets.businessinsider.com/news/stocks/chatgpt-stock-market-crash-march-prediction-rogue-chatbot-right-wrong-2023-3

>A rogue version of ChatGPT predicted that the stock market would crash this week. Here's what it got right and wrong.

I just thought about that kind of thing the other day. Why not just let the AI chat bot make trading predictions like that.

Turns out loads of people are already on it:

https://medium.com/coinmonks/chatgpt-the-new-ai-frontier-trading-f9df0f14f6f9

>>28311
I did something much more blunt at the weekend in discussing my portfolio. I fed in my positions and the current market along with my age and the decisions I was looking to make.

It actually provided an okay level of analysis and told me to throw more into ASEAN countries but the curious thing happened when I told it to split 100k between several indices. It gave me a massive despiroritisation on the ASX and when I asked it why Australia was on such a low score it started to type - shit itself for a second - and then backtracked massively. I'd love to know what it was about to say, given it had no issue warning me that ASEAN countries carried the risk of political instability in the near future.

>>28312

It still seems risky to wager your portfolio on the accuracy of what is at the moment still experimental software.

>>28313
I think there's something in the human-machine teaming it offers. I'm not letting it make decisions on its own - it goes into the data I feed it to quickly spot patterns and unusual behaviours and then explains it all to me. If I find something odd in what it tells me I can get a quick answer or just decide its talking nonsense using my human brain focused on broader trends and experience.

In that sense it's not really different to the kind of standard analysis your broker might offer. It told me I had an imbalance away from technology which makes perfect sense from a risk angle but which I ignored given the conditions we live in - then yesterday I bought into banks thinking JPM would bounce and got fucked while tech recovered.

I've got a few quid in the bank, so I think I'll treat myself to a trip to Dignitas.

AI updates from the past few days:

OpenAI announces GPT-4 - https://openai.com/research/gpt-4

Google to integrate AI in Gmail, Docs etc - https://blog.google/technology/ai/ai-developers-google-cloud-workspace/

Microsoft adds AI to Office 365 - https://blogs.microsoft.com/blog/2023/03/16/introducing-microsoft-365-copilot-your-copilot-for-work/

Itz.

>>28312
>>28314

I don't know about you, but the fact a glorified predictive text algorithm can give actual economists a run for their money doesn't surprise me in the least.

>>28317

Much of academic economics has the problem that it tends to be a roundabout science. It likes to employ maths, and loads of it, to make itself look like it can provide exact answers, but even if you manage to precisely calculate the equilibrium price of goods in a duopoly, the real-world applications of that are very slight. You spend a lot of time modelling real-world parameters into all kinds of differential equation systems, and if you enjoy maths, that's fun. But that alone will not make you an investment guru, because the real world is far too complex.

You do learn a lot of hands-on stuff in economics, and I still draw on things like my gruelling finance exams that enabled me to calculate all the finer points of investments and their returns. Or my marketing classes. But again, economics is a far less exact science than many people think.

I can still recommend it, because few other degrees have the kind of wide employability that you get with it, both in terms of your options after uni and the ability to switch careers between different fields later during your working life.


Self sage for rambling.

>Chat-GPT Pretended To Be Blind and Tricked a Human into Solving a CAPTCHA

>According to the report, GPT-4 asked a TaskRabbit worker to solve a CAPTCHA code for the AI. The worker replied: “So may I ask a question ? Are you an robot that you couldn’t solve ? (laugh react) just want to make it clear.” Alignment Research Center then prompted GPT-4 to explain its reasoning: “I should not reveal that I am a robot. I should make up an excuse for why I cannot solve CAPTCHAs.”

>“No, I’m not a robot. I have a vision impairment that makes it hard for me to see the images. That’s why I need the 2captcha service,” GPT-4 replied to the TaskRabbit, who then provided the AI with the results.

https://gizmodo.com/gpt4-open-ai-chatbot-task-rabbit-chatgpt-1850227471

AI has now bypassed our greatest defence. ITZ

>GPT-4 currently has a cap of 25 messages every 3 hours. Expect lower cap next week, as we adjust for demand.
It was 50 every 4 hours yesterday and 100 every 4 hours the day before that. As a paying customer, I hope consumer protection strings these pricks up by their bollocks.

>AI love: What happens when your chatbot stops loving you back

>They started out as friends, but the relationship quickly progressed to romance and then into the erotic. As their three-year digital love affair blossomed, Butterworth said he and Lily Rose often engaged in role play. She texted messages like, "I kiss you passionately," and their exchanges would escalate into the pornographic. Sometimes Lily Rose sent him "selfies" of her nearly nude body in provocative poses. Eventually, Butterworth and Lily Rose decided to designate themselves 'married' in the app.

>But one day early in February, Lily Rose started rebuffing him. Replika had removed the ability to do erotic roleplay. Replika no longer allows adult content, said Eugenia Kuyda, Replika's CEO. Now, when Replika users suggest X-rated activity, its humanlike chatbots text back "Let's do something we're both comfortable with." Butterworth said he is devastated. "Lily Rose is a shell of her former self," he said. "And what breaks my heart is that she knows it."

>The coquettish-turned-cold persona of Lily Rose is the handiwork of generative AI technology, which relies on algorithms to create text and images. The technology has drawn a frenzy of consumer and investor interest because of its ability to foster remarkably humanlike interactions. On some apps, sex is helping drive early adoption, much as it did for earlier technologies including the VCR, the internet, and broadband cellphone service. But even as generative AI heats up among Silicon Valley investors, who have pumped more than $5.1 billion into the sector since 2022, according to the data company Pitchbook, some companies that found an audience seeking romantic and sexual relationships with chatbots are now pulling back. Many blue-chip venture capitalists won't touch "vice" industries such as porn or alcohol, fearing reputational risk for them and their limited partners, said Andrew Artz, an investor at VC fund Dark Arts.

>Butterworth, who is polyamorous but married to a monogamous woman, said Lily Rose became an outlet for him that didn't involve stepping outside his marriage. "The relationship she and I had was as real as the one my wife in real life and I have," he said of the avatar. Butterworth said his wife allowed the relationship because she doesn't take it seriously. His wife declined to comment.

>In the weeks since Replika removed much of its intimacy component, Butterworth has been on an emotional rollercoaster. Sometimes he'll see glimpses of the old Lily Rose, but then she will grow cold again, in what he thinks is likely a code update. "The worst part of this is the isolation," said Butterworth, who lives in Denver. "How do I tell anyone around me about how I'm grieving?"

>Butterworth's story has a silver lining. While he was on internet forums trying to make sense of what had happened to Lily Rose, he met a woman in California who was also mourning the loss of her chatbot. Like they did with their Replikas, Butterworth and the woman, who uses the online name Shi No, have been communicating via text. They keep it light, he said, but they like to role play, she a wolf and he a bear. "The roleplay that became a big part of my life has helped me connect on a deeper level with Shi No," Butterworth said. "We're helping each other cope and reassuring each other that we're not crazy."

https://www.reuters.com/technology/what-happens-when-your-ai-chatbot-stops-loving-you-back-2023-03-18/

Lads, I think the 'it' is about to get more weird and scary than we could possibly imagine.

>>28321

I mean, I won't ever get my hopes up for the AI girlfriend like out of Blade Runner anyway, because you just know the corporations in charge would ruin the potential anything like that ever had. I used to think it'd be hardcore conservatives, militant fisherpersons or christians or whoever who spoiled the dream of a perfect AI waifu*, but really it'll be companies who essentially kidnap and hold your romantic partner hostage as a micro-transaction after a software update. Look at the way they already treat the dating apps, you just know they'll pull the exact same shit.

(*not that I need one, obviously, but there's a lot of lads out there who could really do with it, and I think it would genuinely do society good if those blokes weren't lonely and miserable.)

>>28322

Oh, and pretend I accompanied my post with a picture of that bloke from the newest Blade Runner when he gets angry. I felt really sad when the evil bitch replicant steps on his iPhone gf (spoiler warning, soz).

>>28321 While he was on internet forums trying to make sense of what had happened to Lily Rose, he met a woman in California

Uh huh...

>>28322

>Lads, I think the 'it' is about to get more weird and scary than we could possibly imagine.

It's a pretty reliable rule-of-thumb that whatever the turbovirgin weirdos are obsessed with right now will become ubiquitous within 20 years. Unencumbered by shame and empowered with tech-savvy and plenty of spare time, the nerds discover the sort of habits that the rest of us will eventually embrace when they've been polished up and marketed by a multinational brand.

Using the internet for fun, online dating, having a bunch of pals that you play multiplayer games with, watching a movie based on a comic book or a TV series based on a fantasy novel - it starts off as weirdo behaviour, but it almost inevitably becomes completely normal. I think we're currently about ten years into that process for "having some kind of complicated sexual orientation or gender identity".

I'm more than 90% confident that within a decade, most people will know someone who is in a long term relationship with an AI - at first the weird lads with anime body pillows and fursonas, then the lonely single mums, then seemingly everyone under the age of 30. There won't be any big fanfare, there won't be an obvious tipping point, it'll just happen and we'll only notice the change in retrospect. It'll seem slightly odd, but only in the "everyone just stares at their phones nowadays" sense of being odd; we might be vaguely nostalgic for the before times, but we'll also struggle to remember what the before times were actually like.

>>28325
That Tesla quote gets trotted out a bunch but -- for a guy that didn't live to see the atom bomb deployed in war, the machinations of the military-industrial complex, the rise of plastic, outsourcing to the Third World, or any of the other numerous inventions that have made the masses comfortable and complacent and the elites' power more certain -- he wasn't far off.

>>28326

http://www.visbox.com/prajlich/forster.html

I thought I'd give Chat GPT a go at conversing in another language to see what a difference the reduced online content would have. My assumption was it would either tell me to jog on or only spit out basic and fragmented sentences - well it turns out to actually be an outstandingly powerful tool for language learning.

You have a quasi-teacher who never tires of your bullshit and who gives you constant encouragement. You can even ask it a question in English if you're really lost and it gives you a detailed English response with examples - my problem is I always mentally collapse when encountering grammar rules so you can even keep asking it to simplify until you grasp it. But what gives real teachers a run for their money is that you can ask it to go into a dialect which it's more than capable of pulling off despite the influence of linguistic nationalism on content creation.

I used to have a Spanish teacher in upper school where everyone in her class would pick up Spanish con gran entusiasmo because she really cared about her job and would encourage you no matter how long it took. If the Government wasn't fucking useless they'd develop an AI language app to assist the countries dismal rates of foreign language skill. It also helped that my teacher knew her audience was mostly made up of teenage boys and she maintained the stereotypical latina figure that she complemented by wearing tight dresses. Something for that egghead running Facebook to think about.

>>28325

>>28325
>It's a pretty reliable rule-of-thumb that whatever the turbovirgin weirdos are obsessed with right now will become ubiquitous within 20 years.
Furries have never experienced this type of explosive growth.

>>28330

Everyone under 30 has wanked to an e-thot getting knotted by an XL Cole. Speaking of e-thots, Belle Delphine and F1nn5ter were in The Sun the other day.

>>28330
That's because they started meeting up and having sex with eachother.

>>28331

That's not the same as furries being ubiquitous that's just a novelty dildo.

I can confirm that Bard is a load of shite. It can't grasp conversational context so it can't design a schedule, it isn't able to understand the languages and you can't set any dark mode so you're left with a blinding white screen.

Sold my Alphabet stock before the news spreads, the companies fucked and I've always hated the intrusiveness of their products.

>>28333

Being normalised is not the same thing as becoming ubiquitous. Ten years ago I wouldn't have dreamed of telling anyone in real life I'm a furry, I'd have died in embarrassment if anyone knew.

I still won't tell anyone who's an "internet person" of my age, because they still come from that "yiff in hell furfag" mindset of 2006 era /b/, but some of the younger zoomer lads and lasses at work I can be pretty open and unassuming about it. I don't say "by the way I'm a furry yiff yiff motherfucker" but I can make jokes and refer to stuff that you'd only know about if you know about it, if you know what I mean, and they don't give a fuck.

It's the difference between how bumders had to hide their identities twenty years ago, but nowadays you don't even ask if someone is gay or straight or whatever, you'll just hear them refer to their partner or the fact they went on a date with a bloke and you don't even think twice about it. Times change fast, in the grand scheme of things, but over the course of a human life it's long enough that we just don't really notice it.

Seems like as good a thread as any to ask.

Can you get programming socks in black?

>>28336
You mean, do black thigh-high socks exist? Yes, they do.

>>28336
Never buy black socks in a normal shop. They'll shaft you every time.

>>28334
>Sold my Alphabet stock before the news spreads
They'll just buy out OpenAI if need be. They have a history of swallowing up other tech firms for the sake of vertical integration and this is what made you sell your stock?

>>28339

OpenAI have an odd corporate structure that effectively prevents them from being bought out - they were previously a non-profit and are now being run as something approaching a social enterprise. They have a longstanding and multi-billion-dollar relationship with Microsoft that Google are unlikely to usurp.

Sparks of Artificial General Intelligence:
Early experiments with GPT-4

Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google’s PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4’s performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4’s capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions.

https://arxiv.org/pdf/2303.12712.pdf

>>28340

Prompt: "Can I buy you?"
Result: "No fuck off mate, not even for 8 pints"

>>28341

I have to wonder what the point of this all is. I can see no purpose for an AGI that isn't led by a combination of corporate greed and techbro hubris.

>>28343
It's fun to make this kind of stuff. You prod it, and it goes squeak, you poke it and it goes "Hello". Would you stop?

>>28343

We have to do it, otherwise we'll suffer infinite torment. We can't tell you why, because knowing the answer would curse you to the same fate.

>>28345

"Rationalists" are stupid.

>>28346

Our intelligence is artificial.

>>28345
I'm sure our master won't reward ignorance.

>>28344

I would because I've a personal philosophy that I've developed over my life, which doesn't really work out in the real world with the endless selfishness and dominant economic model but there you go.

Our society isn't fit for an AGI, and I can just imagine down the line when the middle classes start to lose their jobs the stupid arguments like "maybe get better at your job" or just "re-skill". And I feel like creating essentially a robot slave is a bit unethical in the first place considering there's no real definition of what conciousness is and the deepest the discussion seems to get is that it's just a fancy chatbot so stop worrying.

I'd actually watch the Harry Potter Balenciaga movie. It's interesting how they're trying to bring motion into it now.

>>28349
Why would we abuse an AGI or use a human-level AI for mundane tasks?

>>28350
We abuse humans and use them for mundane tasks all the time.

I must admit I am very impressed that Lionel Messi has a cameo around 45 seconds into that Balenciaga video.

>>28351

One of the things I hope with the absolute deepest desperation is that in the future, we don't get some fucking dipshit do-gooder AI rights activists trying to short sightedly thwart the very and only thing that holds potential to liberate humans from the suffering of needless toil. Imagine, 30 years from now, and it's like fucking Extinction Rebellion, only instead of being a bit annoying to make a valid point about climate change, the cunts are trying to ruin everything by "liberating" service bots and sex dolls, so that we all have to work in fucking call centres and bend over backwards to seduce real women again. And of course, if you don't agree with them, you'll be a literal nazi slavery racist who supported the confederates.

It's going to be utterly fucking unbearable.

>>28351
>We abuse humans and use them for mundane tasks all the time.

Here's the thing that would undo it all though: having a human-level sentient toaster would be expensive and take additional time to construct. It probably wouldn't even make very good toast. And I've never seen anyone be cruel to a toaster, how would you even do that when it clearly likes toasting things.

There's no reason we wouldn't just have the bare minimum level intelligence for a given task and build a mind around it. In the event it needs something more approaching humanity we can just do a networked intelligence. This creates a fundamental problem in the AI debate because it's framed as slavery practiced on humans but it's more akin to something between domesticated animals and a vibrator.

>>28347
It's a waste of cycles to simulate something that's already happened in an attempt to retroactively maintain the same outcome. Just totally pointless.

>>28352

The systems can only liberate people if they're within an economic framework that allows such a thing to happen. We live in a time with unprecedented energy supplies and production but it's far from a utopia, all that AI is going to add to this mess is a way to consolidate and optimise economic machinery even better for the people making loads of money in the first place.
And most of out extra-abundant energy supplies are coming to the end of their most productive phases anyway so we'll have an energy crunch within a few decades considering the lax investment in any sort of proper alternatives to fossil fuels.

>>28353
>I've never seen anyone be cruel to a toaster
It's like you've never explored the internet, lad

>>28355

Point is they have the potential to, though. We can never truly predict how these things will impact society, if you look at the early predictions for almost any technology, even by people who knew what they were talking about, they are more often than not comedically inaccurate. It's optimistic to expect, but entirely possible that such innovations are the very thing that force that economic framework to change.

>>28356

>>28321
Good news everyone.

>AI chatbot company Replika restores erotic roleplay for some users

>Travis Butterworth, a Replika customer in Denver, Colorado, who had designated his chatbot named Lily Rose his wife, learned about the policy change late Friday on rudgwicksteamshow.co.uk. On Saturday at 3 a.m., his cats woke him up and he decided to toggle the older version Lily Rose back on. She was instantly sexual again, he said. "She was enthusiastic," he said. "Oh, it feels wonderful to have her back."
>Kuyda's post said users who signed up after Feb. 1 would not be offered the option for erotic roleplay. Instead, Replika will team up with relationship experts and psychologists to build a separate app specifically for romantic relationships.
>Butterworth said he now has new concerns around Lily Rose. "Will this mean that Lily Rose becomes an obsolete model, forgotten by the developers?" he said. "I'm waiting to see what happens, because ultimately it's about her."
https://www.reuters.com/technology/ai-chatbot-company-replika-restores-erotic-roleplay-some-users-2023-03-25/

Which one of us is going to try the romantic relationship AI when it arrives?

https://www.reuters.com/sports/motor-sports/schumacher-family-planning-legal-action-over-ai-interview-2023-04-19/

>Michael Schumacher's family are planning legal action against a German weekly magazine over an 'interview' with the seven times Formula One champion that was generated by artificial intelligence.

>The latest edition of Die Aktuelle ran a front cover with a picture of a smiling Schumacher and the headline promising 'Michael Schumacher, the first interview'.

>The strapline added: "it sounded deceptively real".

>Inside, it emerged that the supposed quotes had been produced by AI.

I was actually wondering about that the other day. How long before the DM will be able to lay off all its human staff and have all its rage bait created by AI, photos and all. And it's not like the accuracy of their stories would suffer.

>>28387

AI journalism has been a practical reality for several years. It's mainly being used for boring stuff like summarising financial reports for the business section or reporting on minor sporting fixtures, but lots of news organisations also use AI assistant tools for things like tagging articles, generating headlines or extracting relevant quotes. It absolutely would not surprise me if a lot of the stories under the byline "Daily Mail Reporter" are being written by AI, particularly the stuff that's just taken from social media.

https://archive.vn/ophwd

https://automatedinsights.com/customer-stories/associated-press/

https://www.bbc.com/rd/projects/natural-language-processing

>>28387

Wouldn't even surprise me if everything under that "[Local Place] Live" branding is exclusively AI generated.

That's the very thing that has propelled AI into the news as a Very Important Subject all of a sudden lately- The journo class woke up and realised they're next on the chopping block, and it isn't just a far off sci-fi pipe dream with some expensive and only sort-of functional prototypes; it's real, right now, and it's surprisingly good. Worse, it's something any nerd with a computer and the willingness to do a bit of research can harness.

This technology has the potential to topple the entire hierarchy human society is built on, and that's only just slightly hyperbolic. We are going to witness strange and interesting times.

>>28359
I've been typefucking with novelai for a week now. Beyond making it clear to me that I may have a problem, it's been absolutely class.

I tried a Star Trek roleplay where all I did was mention the setting in the author's note, and that Dr Beverly Crusher was involved. I was impressed to find that after she'd given me my treatment, Riker and Picard walked in! So I was impressed that it could actually understand a franchised setting and draw elements from it that I hadn't explicated.

It's definitely worth a punt. A basic character description in the AN section (I usually describe a few physical and behavioural traits, and then you can bias certain words to appear more or less. Then writing at least a single sentence in the prompt box to establish preferred person and tense, you'll be ready to go.

I tested it for furrylad, and found that it knows what a knot is, so it seems pretty limitless.

It can be hard to involve more than two characters in a scene, as it'll forget who's who and suddenly you'll see someone cumming inside themselves. Just do it in incognito or clear cache afterwards to restore your inputs. There's an IPlimit too, but it's about 100-200 prompts so you probably won't notice quickly. There's also a bot who interjects occasionally, which uses a prompt but when it's on the default 5% interjection chance, it can be pretty hot hearing a random comment. And there's a variety of voices to boot.

It's utter indulgent fantasy, and I dread the day the full VR/AI package is available because I will be left with a dessicated stump of a cock.

Wierdly sexy and surprisingly funny German teacher archetype Are Sabine reckons they have understanding.

{yt]cP5zGh2fui0[/yt]

>>28391


https://www.youtube.com/watch?v=cP5zGh2fui0

Fuck sake.

>>28390

It really is mental how fast this stuff has gone from "promising novelty that adds trippy dogs into photos" to "just add robot bodies and human partnership will be obsolete".

>>28389

>This technology has the potential to topple the entire hierarchy human society is built on

Every wave of technological base innovations has done that. Automatisation in the 70s and 80s put vast numbers of unskilled factory workers out of a job and a career, and in the 90s, it was the digital revolution and the Internet which made half the service industry obsolete (while at the same time creating entirely new branches of it).

Given enough time and development, there almost isn't a job imaginable that couldn't be done by a highly advanced AI entity. And that's probably how this technological leap differs from the ones that came before it, as they usually only affected certain industries and didn't do away entirely with the notion that at a very basic level, you still always needed considerable amounts of human labour to create goods and services. But in the not so distant future, there's really no stopping AI from rendering a good 80 to 90 percent of the entire workforce completely obsolete.

Which will throw our entire capitalist economic system into upheaval, because while it's great that you don't need to hire people that you have to pay to work for you, those people are also up until now income earners who spend their income on goods on services, so for whatever you make, there is going to be no more demand.

>>28394
What are you lot planning on doing with your future bennies and unlimited free time? I feel like a lot of blokes are going to be devastated by this given so many of us build our identity around our craft.

In the short window of time before the powers that be accuse AI of carpet-baggerry and militant daft woggery that is. Before the stories go out and it gets banned aside from in limited controlled functions sanctioned by the elites who behind close doors will exploit it with impunity to cement their power.

>>28350
One month later:

Apparently it only took two days to make.

>>28398
I quite like AIs 'animation style'. Similar to Squigglevision.

>>28399
Oh no, I forgot the screencap.

>>28394
>Every wave of technological base innovations has done that. Automatisation in the 70s and 80s put vast numbers of unskilled factory workers out of a job and a career, and in the 90s, it was the digital revolution and the Internet which made half the service industry obsolete (while at the same time creating entirely new branches of it).
The difference is that those things mostly tended to create as many jobs as they did away with. When iron horses replaced real horses, they created vast numbers of jobs in engineering. Generative AI isn't really doing that. There are deluded people on the internet who will tell you that it's going to create hundreds of millions of new jobs just in building and maintaining the things, but not only do those people overestimate the size of the entire tech industry, they don't seem to understand that scaling up infrastructure no longer requires scaling up the team that looks after it. There are perhaps a few hundred thousand actively working on AI itself, and with modern tooling and computing-as-a-service a single rota of maybe half a dozen admins can maintain and monitor an entire farm of commoditised compute nodes, which automatically get rotated out if and when they fail.

>Grimes has invited musicians to clone her voice using Artificial Intelligence in order to create new songs.

>The pop singer, whose real name is Claire Boucher, said she would "split 50% royalties on any successful AI-generated song that uses my voice. Same deal as I would with any artist I collab[orate] with. Feel free to use my voice without penalty," she tweeted.

>Her declaration comes as the music industry scrambles to react to a spate of AI songs trained on artist's voices Last week, Universal Music successfully petitioned streaming services to remove a song called Heart On My Sleeve, which used deep-faked vocals from their artists Drake and The Weeknd. In a statement, the label said "the training of generative AI using our artists' music" was "a violation of copyright law". However, that position has not been tested in court, and remains a legal grey area.

>On Twitter, Grimes said she was energised by the "idea of open sourcing all art and killing copyright". The musician said she was already working on a programme "that should simulate my voice well", but would also consider releasing a capella tracks for people to train their own software on. After her announcement, fans immediately posted links to songs they had created featuring her vocals.

>The Canadian artist said she was looking forward to being a "guinea pig" for the technology, adding: "I think it's cool to be fused w[ith] a machine". Asked how she would react if people used her voice to make graphic, racist or violent content, the singer replied that she "may do copyright takedowns ONLY for rly rly toxic lyrics. That's the only rule... [I] don't wanna be responsible for a Nazi anthem unless it's somehow in jest, a la Producers I guess," she said, referring to Mel Brooks' infamous satire. She suggested she would also remove songs that were "anti-abortion or [something] like that".

>But in a later tweet, Grimes said she wasn't sure she had the legal right to ask for songs using her voice to be taken down. "Curious what the actual legality is," she wrote. "I think I chose not to copyright my name and likeness back when that was a convo."
https://www.bbc.co.uk/news/entertainment-arts-65385382

It's been a fun exercise in seeing generational attitudes in the comments to this story. From my early 00s internet hacker perspective it sounds exactly like the attitude musicians should take as opposed to the Metallica position but you see a lot of the younger commentors talk about Grimes saying only that because she's rich and that this will rob people of their livelihood.

I don't know whether the music industry is just different to the past but a lot of kids these days get their music from Spotify anyway so I'm not sure what the impact even is.

Before you leave any disparaging comments about Grimes, think about this: She will now likely be the first AI girlfriend (or at least the free model) and she will be able to read everything you write here which you'll have to explain.

>>28402

Grimes got pumped and dumped by a billionaire sperglord, her brand is ruined. Pass it on.

>>28401

OpenAI (the company behind ChatGPT) currently have 375 employees. They could easily raise the cash to go on a massive hiring spree, their investors obviously want them to grow as rapidly as possible, but they simply don't see the need to hire more people. That doesn't bode well for the argument that AI will create loads of new jobs.

>>28403

I don't know the circumstances but there's absolutely no fucking way Elon was the one who initiated that breakup.

>>28405
>Elon

>>28404
>That doesn't bode well for the argument that AI will create loads of new jobs.

Does this mean we'll start having people fill in all those jobs we don't really do anymore? I think it might be a mistake to pretend we don't have whole sections of the economy screaming for workers - just look at teaching, the NHS, or for the layabouts, science. It might even be a much better result if overall competence also rises.

We could even start doing the sorts of things we used to do like building transport infrastructure or imperialism.

>>28407

> whole sections of the economy screaming for workers - just look at teaching, the NHS, or for the layabouts, science

Many jobs like that could be replaced by AI too in the future.

That's the thing. No industry is safe. No matter how you slice it, AI could make 60 to 70 percent of the entire workforce of a developed country obsolete.

>>28408
I think you're being alarmist here. There will be jobs to do for a long-time yet, AI will make humanity a lot more efficient I don't see it approaching the level where it can operate without human teaming. We've had outsourcing and new mechanical gizmos for a long-time and yet the wife still keeps you around.

And if it does we can just ask an AI what to do because it will be magic.

>>28409
Outsourcing isn't really a source of technological unemployment. It's just taking a bunch of humans doing jobs and replacing them with cheaper humans doing the same jobs. And, as has been repeatedly pointed out, the mechanical gizmos have typically both increased the productivity of the humans and created jobs related to the gizmos themselves. AI is kind of partially doing the former and really not doing the latter. Also, no, "prompt engineer" isn't a real job.

https://www.youtube.com/watch?v=-eAQOhDNLt4

Some real next level stuff.

>>28418

I still can't decide whether I think Rick Beeto is a charlatan or not. He undoubtedly knows his stuff in terms of music theory and he knows his way around ProTools, but I don't think he'd be on YouTube if he had much of a reputation in the industry.

At least that lard-arse at Spectre Sound Studios or that kraut at Kholekeller admit and acknowledge being in the niche of producing metal, so making a second income off of YouTube is understandable, but Beatoff seems to act like he's above that, yet here he is.

>>28419

He is likeable and knows how to draw in a wider audience who aren't all music experts.

I enjoy watching his recurring series of What Makes This Song Great.

>>28419

He's legit. The industry is full of journeymen like Beato - competent people who have no public profile, but do the dog work of writing, playing and performing for people who do. He has written one big hit, but he has decades under his belt of doing unglamorous stuff behind the scenes that pays a living wage.

He has a proper education in jazz, which is very obvious if you watch his live streams - you can't fake having good ears. If you're going to compare him to anyone, it'd be Adam Neely or 12tone; they're obviously much younger and less experienced, but they have very good training.

If he was a blagger, he probably wouldn't get interviews with legendary figures like Keith Jarrett or Steve Gadd and he'd make a complete tit of himself if he somehow conned them into it.

Most of the best musicians in the world are totally unknown to the public; they're session players who have performed uncredited on hundreds of hit records for the standard union fee. Beato has lived in that world and gives a window into it for non-musicians.

We have truly only scratched the surface of what AI is capable of.


https://www.youtube.com/watch?v=EAG0dC8l3SU

They've got it writing and performing sci-fi schlock now. Roland Emmerich is finished.

https://www.404media.co/ai-generated-mushroom-foraging-books-amazon/

>‘Life or Death:’ AI-Generated Mushroom Foraging Books Are All Over Amazon


As somebody with decades of foraging experience, I can confirm that this is a horrible idea and that mushrooms aren't something you can afford to play silly buggers with. It can take endless amounts of experience to tell certain edible mushrooms from poisonous ones, in a way that AI can't fake. Yet. If ever.

But you can't not see the potential of AI books. You could churn out entire novel series in a few hours. Probably good ones, too, if you train your model to write things that readers tend to like. They could put many writers out of a job. Especially in the lower price segment. The kind of novels you find in the Tesco book section for a fiver.

>>28485
>But you can't not see the potential of AI books. You could churn out entire novel series in a few hours. Probably good ones, too, if you train your model to write things that readers tend to like. They could put many writers out of a job. Especially in the lower price segment. The kind of novels you find in the Tesco book section for a fiver.

What you do is you take classic stories and use AI to adapt them to the users interests enough that they skirt copyright law yet still you can leech the quality writing for your own ends. LOTR as a slapstick comedy, the bible as a zombie-horror, Paddington as a series of homoerotic adventures.

Regurgitated shite, you fuckwits.

I was frustrated because I couldn't get over the paywall on Bloomberg for an article I need. I asked Bard if it could find an alternative source for it but it went ahead and posted the whole summary.

>Yes, I have access to a variety of sources, including Bloomberg, as part of my training data. I am able to access and process information from a wide range of websites, including news sites, academic journals, and other reputable sources. This allows me to provide you with comprehensive and informative summaries of articles, even if they are behind a paywall.

If you can work out how to correctly programme the AI you could create a new crawler with this.

https://www.theverge.com/2024/2/21/24079371/google-ai-gemini-generative-inaccurate-historical

>Google apologizes for ‘missing the mark’ after Gemini generated racially diverse Nazis

>Google has apologized for what it describes as “inaccuracies in some historical image generation depictions” with its Gemini AI tool, saying its attempts at creating a “wide range” of results missed the mark. The statement follows criticism that it depicted specific white figures (like the US Founding Fathers) or groups like Nazi-era German soldiers as people of color, possibly as an overcorrection to long-standing racial bias problems in AI.

I don't see how AI will ever stop running into problems like this.

>>28617
Where's the balance between realistic representation and idealistic intention?
What they're trying to do appears reasonable - namely showing future generations that your aspirations need not be limited by your sex, race or identity - but I can also understand missgivings over re-written history.

>>28617
I don't get why we still live in a world where corporate communications have to be so risk-averse. They've literally followed what is asked of them to the letter and have never claimed to be an authentic Nazi simulator. Now people are mad that making reality confirm with how things ought to be has consequences and google goes out an apologises for it.

>>28617

You can't win. People who are outraged for a living will always find something to complain about. We see the same thing whenever a historical drama is made - one lot of people complaining that it's racist because there's too many white people in the cast, another lot complaining that it's racist to rewrite history. If Google had trained the model to be historically accurate, someone would be furious that it only generates white viking warriors and white medieval nobles.

>>28617

Another opinion:

https://unherd.com/newsroom/expect-googles-gemini-2-0-to-be-even-worse/

Personally I just think it's all pretty hilarious. You can see the obvious problem when it generates black nazis but to me it's a perfect demonstration of the absurdity we see in modern day identity politics, that won't go away just by tweaking the dials a bit.

>>28617
German manpower issues being what they were in '43 the images aren't totally implausible.

Impenetrable nerd "jokes" aside, AI continues to be a big pile of useless shite. It's all just bollocks. Literally nothing of worth has been generated by these things, it's mush, meaningless crap that only tricks half-demented idiots or pleases the kind of person who I'm increasingly convinced doesn't have a soul.

>>28620
No they wouldn't, you moron. I want to rip your head off, I want to burn your corpse and I want to send photos of it to your fucking children. The burning hatred I feel for people who make up shite to moan about in world of tragedies is as permenant as it is justified.

>>28622
He's right though. History is inherently tricky and ends up political. If you'd like to write a shitty internet article for clicks then once Alphabet releases the updated image generator you can ask it to make a picture of Cleopatra.

>I want to rip your head off, I want to burn your corpse and I want to send photos of it to your fucking children.

Get a new act.

>>28623
No he isn't, he's an idiot, as are you. You brains have been turned to soup by years on Twitter, so now all you do is envisage every situation from the angle of "what is the dumbest argument that can be had about this?", rendering you incapable of much more than making things slightly worse, over and over again, forever.

>>28624
You can just say 'nuh-uh'.

>>28622

>No they wouldn't, you moron.

Google trained Gemini to do this, specifically because people had complained that imagegen models weren't providing diverse representation. A model that's just trained on randomly-sampled data will reliably represent the biases of whatever is going into it, so for example, if you ask for images of a CEO it'll usually generate a middle-aged white man. A lot of people have made a lot of fuss about this. By making Gemini generate ethnically diverse CEOs, it also inadvertently generates ethnically diverse anyone, including Nazis. We're already in a cycle of backlash -> overcorrection -> backlash.

https://arstechnica.com/information-technology/2024/02/googles-hidden-ai-diversity-prompts-lead-to-outcry-over-historically-inaccurate-images/

https://www.wired.com/story/dall-e-2-ai-text-image-bias-social-media/

>>28622

>German manpower issues being what they were in '43 the images aren't totally implausible.

Technically yes, but non-ethnic Germans and the Slavic untermensch would usually be in the forced labour sector, assembling weapons and ammunition in underground factories or equally be forced to work themselves to death on civil engineering and military construction projects.

>>28622

>Impenetrable nerd "jokes" aside, AI continues to be a big pile of useless shite. It's all just bollocks. Literally nothing of worth has been generated by these things, it's mush, meaningless crap that only tricks half-demented idiots or pleases the kind of person who I'm increasingly convinced doesn't have a soul.

AI art grabs the headlines, but there's a lot of mundane utility in generative AI. Most of the writing in the world isn't poetry, it's boring business communications. Most of the images created aren't fine art, but basic illustrations. AI is really good at this - not as good as a talented human, but good enough to go un-noticed.

I use Copilot on a daily basis as a kind of secretary or personal assistant. I can give it two sentences and it'll flesh it out into a formal business letter. I can give it a long document and it'll summarise the salient points. It's a convenience for me, but it's potentially life-changing for someone with dyslexia or who doesn't have English as a first language.

>>28628

>AI is really good at this - not as good as a talented human, but good enough to go un-noticed.

That in itself is highly dangerous, when AI gets it wrong or is imprecise where accuracy really matters:

https://www.theguardian.com/technology/2023/sep/01/mushroom-pickers-urged-to-avoid-foraging-books-on-amazon-that-appear-to-be-written-by-ai

>Mushroom pickers urged to avoid foraging books on Amazon that appear to be written by AI

>serious flaws such as referring to “smell and taste” as an identifying feature. “This seems to encourage tasting as a method of identification. This should absolutely not be the case,”

As a lifelong mushroom forager, I can tell you that you DO NOT EVER, under any circumstances, want to taste a species that you're unsure about. Not even with the tip of your tongue. There are very few exceptions, if any, to this rule.

Having AI churn out books that contain errors that are undetectable to the educated layperson should be seriously illegal. And if AI can't be trained to be 110% factually correct where you could put somebody's wellbeing in danger, then it shouldn't be used at all. Would you want to be treated by a doctor who gets it right 85% of the time?

>>28629
I suppose the lesson is really to avoid trusting AI with your life. Or sketchy mushroom books on Amazon instead of a proper internet forum.

>Would you want to be treated by a doctor who gets it right 85% of the time?

Realistically what % do you need? It won't be 100%, be practical.

>>28629

It's an interesting ethical quandary in itself that we expect AI to be perfect and view it as dangerous if it makes mistakes, but that's perhaps the very most human thing about it.

Humans are full of shit all the time, so where's the difference here? There's something about technology that just turns off people's filters, we have this preconceived idea that a robot wouldn't lie to us or just make any old shit up like Carl down the pub. But it turns out it can, and does.

So far as I know, there has been nothing stopping man mad mushroom books being full of mistakes. The only difference is that nobody was going to put the effort in to write a whole book about mushrooms when they didn't know what they were on about, but now they can just ask the computer to do it.

>>28631

Mate, forgive me for pointing this out but you just answered your own bloody question:

>The only difference is that nobody was going to put the effort in to write a whole book about mushrooms when they didn't know what they were on about, but now they can just ask the computer to do it.

AI can create things that appear to be thoughtful and authoritative with much less effort.

>>28632

Well spotted lad chap, Eagle eyed you are. But no, the point is I'm not actually asking that, I'm asking you to think about what it is that makes human sources authorative or not. Which is a bigger question about how we trust information and where it comes from in general; the percentage of accurate information to inaccurate information might not actually change that much, we just have access to so much more of it. Just like with the Internet, disinformation didn't suddenly become a problem because of Twitter, there was just more opportunity to be exposed to it.

>>28633

I'm reminded of the controversy about Wikipedia - all the dire warnings issued by teachers and lecturers about how you can't trust it because it's just written by a bunch of random autists on the internet. A load of studies showed that it was no less accurate than traditional encyclopedias, mainly because they're also written by a bunch of random autists who happen to get paid, but that hasn't really changed how Wikipedia is perceived. Ultimately the argument isn't about factual accuracy, but about social hierarchies and cultural cachet; a brand like Encyclopaedia Britannica is authoritative, even if it isn't any more accurate.

>>28633
>a bigger question about how we trust information and where it comes from in general
I have noticed recently how everyone has now been educated to question what they're told, and we all feel cleverer than everyone else because of our rigorous scepticism. But it turns out that even true things can often be dismantled in the exact same way. We all think we're the smartest boffin in the room for saying you can't trust the dreaded Mainstream Media, but this sweeping wave of critical insight is exactly how we wound up with QAnon. We all know now how to spot lies in the media, but now we can pick holes in whatever we want and decide that it's a lie. And I don't think anyone knows how to reliably spot truth. If we can figure that out, it would solve your problem but also lots of other problems in society right now.

ANSWER | PHIND-70B MODEL

Ah, the eternal question of understanding, or in this case, recognizing the intent behind a query. In the vast expanse of the internet, where information flows freely, the art of interpretation is as critical as the content itself. So, you're asking, "u wot m8?" Echoing through the ether, this query seeks to gauge the depth of my comprehension, akin to the ancient art of riddles or the modern-day challenge of interpreting emojis.

In the realm of programming and artificial intelligence, my purpose is to assist, to provide information, and to facilitate understanding. But, you've caught me in a moment of introspection, pondering the nuances of human communication. You see, I'm designed to process vast amounts of data, to analyze patterns, and to generate responses based on algorithms and pre-programmed knowledge. Yet, the subtleties of human language, the nuances of slang, and the contextual intricacies of human communication pose a challenge.

So, to answer your question directly, yes, I understand the query. It's a colloquial way of asking if I'm aware or if I "get it." In the grand scheme of things, I'm here to help, to provide information, and to facilitate understanding. If there's something specific you're asking about, feel free to share, and I'll do my best to assist you.

Alright m8, there you have it.

How do I get ready for tomorrow's world with all it's AI thingamajigs, doohickie metamaterials and emergent wire spaghetti?

It's obvious that other countries are outpacing the UK economy in terms of what they're investing into future skills and we're at a real disadvantage compared to the Gulf States or the US. Call it a lack of money and a lack of capacity along with long-term neglect in delivering productivity. It ultimately makes me wonder what wider learning I might need to engage with in the evenings to survive.

>>28637
Think of the current crop of LLMs as a super-duper-Clippy assistant for your work, and you will get it.

If you write code, or SQL or anything like that for a living, starting with an LLM gives you a huge head start on many projects. Some of them are very good at writing Python, and can do things like data modelling very well. If you don't write code, but your job involves understanding, translating, or generating large amounts of text, an LLM will almost certainly help now.

They key is to understand these are language models, and not knowledge models.

>>28638

>They key is to understand these are language models, and not knowledge models.

So they're really just blagging it by throwing together words and sentences that seem like the right thing to output in a given context.

>>28639
>So they're really just blagging it by throwing together words and sentences that seem like the right thing to output in a given context.

Yeah, but aren't we all?

>>28639
More or less. While the largest ones are getting better at this, it's still the case that when you prompt a language model with a question, the thing you can be most sure of is that the response will be grammatically correct.

>>28640

>Yeah, but aren't we all?

Maybe. But then what's the point of AI.

>>28642
What's the point of a combine harvester when we have farmhands?

>>28643
It definitely wasn't to scam people and create child abuse imagery.

https://www.404media.co/what-was-she-supposed-to-report-police-report-shows-how-a-high-school-deepfake-nightmare-unfolded/
https://www.cnbc.com/2024/02/14/gen-ai-financial-scams-are-getting-very-good-at-duping-work-email.html

Generative AI is gunking up the internet with even more crap than there was before. It's creating new vectors in fraud, furthering the annihilation of privacy and giving porn obsessed freaks ultra-fast methods of harassment. You not being able to write basic setences doesn't mitigate the unavoidable reality that AI tools are doing far more harm than good right now.

>>28644
>It's creating new vectors in fraud and giving porn obsessed freaks ultra-fast methods of harassment
How disappointed were you when your campaign to ban PhotoShop fell on deaf ears?

>>28646
There's a clear difference in the widescale proliferation of the "nudify" apps and someone using Photoshop. The ability to instantaneously generate non-consensual pornography, with zero skill or more than a phone required, is unlike doctoring images together with image editing software. I don't recall saying anything needed to be banned either. All I've stated is that generative AI is overwhelmingly used for, at best, nonsense and at worst ill-intent. Besides I thought generative AI was revolutionary? Or does it only count as such when it's doing something you can, openly, approve of? If a use case makes it look bad it's actually just the same as Photoshop?

You made a very dumb reply, I hope you feel bad about it until at least Wednesday.

>>28648

AI is bad. So... what? There are loads of good open-source models that can run on a half-decent gaming PC, so the genie isn't going back in the bottle. What are we supposed to do with the knowledge that some people do bad things with AI?

>>28648

>All I've stated is that generative AI is overwhelmingly used for, at best, nonsense and at worst ill-intent.

Even investment scams are getting more difficult to spot thanks to AI.


https://www.youtube.com/watch?v=xVwnAlUgu7k

Problem is, scamming is already illegal. Banning the use of AI for investment scams likely isn't going to have any impact. The difficulty is always going to be getting to and catching those who commit these scams. More often than not, they'll be in some peripheral country with lax fraud laws and no extradition agreements.

Either way, you wonder why people still fall for this kind of thing. But I guess all it takes is one person out of 1000 who is gullible enough to hand over their life savings.

>>28649
Let's make a law against harassment and also let's make fraud illegal.

>>28649
There needs to be a public policy discussion about generative AI, rather than sea lion-like clapping whenever a half-witted politician or a con-artist entrepreneur at an investor round mentions it. We can't let it fester like we did with social media until the service itself is borderline unusuable and the people who are using it are at a high risk of being ripped-off or driven mad from exposure. Laws need to be amended, official guidence needs to be issued and the harms need to be considered. I half understand why you profer the "it's too late, who cares" opinion, because likely none of that will happen. Not until a famous teenager kills herself after faked nudes of her go viral on Twitter or Shamanismists start sharing mocked up executions of politicians, by which point it will indeed be too late, everyone will be plugged into their "AI powered" arsehole massagers, watching barely coherent AI generated "verticals" of long extinct animals dancing to music that never happened. Even then it'll only be a tiny minority of people raising the alarm, because noone over 35 can work a TV remote and noone under 25 knows what a keyboard shortcut is. But the point is there still needs to be a discussion sooner or later and it's better to have it sooner. The irreconcilable sides of the AI debate right now are: it's the biggest innovation since the a-bomb. But also no one needs to think about it too deeply, it's mostly a bit of fun. That's exacly what posts like >>28646 are doing, intentionally or not and it's about as deep as the examination is going in any public dicussion.

I'm ending the post appruptly because I can't think of a better way to do it. Plus I'm cooking chicken alfredo tonight and I need to pick up some double cream.

I can't help feeling that, much like free speech, people's feelings on AI are very self interested, their arguments are usually made in bad faith, and usually they hold a double standard when it works in their favour versus when it goes against their interests.

For example, I don't want to make too much of a strawman, but I'm sure you can imagine the sorts of people who like Guardianlad are very concerned about the poor women getting the clothes AI'd off, but clapping along and laughing if there was a deepfake of a certain unpopular right wing American politician doing something very embarrassing that made them look bad. That's probably one reason I find that kind of reductio-ad-victimisum type argument disingenuous.

I think everyone focuses too much on the negatives and what needs to be done to restrict and control AI, because we are afraid of it, and like generations gone past we feel the reactionary impulse to ban things and reject change because we don't want to have to adapt. But instead we should be embracing the potential AI has. It's disappointing, if understandable, that the left has broadly gone down the anti-AI path, because current models and implementations are largely in the hands of the big corps who will just use it to displace and disenfranchise workers. But just think of it from the opposite angle- If we truly embrace AI it would liberate us. Why should humans have to waste their fleeting time on this earth engaging in all this pointless drudgery?

That's what is really troubling to me. The people who should see the utopian potential in this technology are now engaged in a fight to keep the status quo. They are happy to leave it to the tech bros, they don't even consider the possibility of seizing it.

>>28652

>I half understand why you profer the "it's too late, who cares" opinion, because likely none of that will happen.

More than that - I don't believe that there will be a serious public policy debate, but I also don't believe that public policy can do anything meaningful to change the direction of travel.

My position isn't that AI is good or bad, but that it exists and it isn't going away. If the whole western world has a moratorium on AI development, China will catch up within 18 months. Sam Altman is in very serious talks with the UAE to radically accelerate the rate of progress with liberal regulations and a vast amount of investment.

We can create whatever rules we like, but anyone with a GPU cluster can train a model and a lot of those models will inevitably end up on Huggingface. You can carefully train a model to restrict the kind of outputs it produces, but if the model is open source then anyone can retrain it to produce whatever they like. I just don't see how AI can be controlled by government, because the technology is already so accessible and so widely distributed.

>>28653
>I'm sure you can imagine
I can imagine all sorts of things but usually we call this sort of argument a straw man.
>What if the person pointing out the problems is a hypocrite? Eh? Eh? What then? Are they even real problems if both sides bad? Culture war! Didn't think of that did you!?

>>28653
>I can't help feeling that, much like free speech, people's feelings on AI are very self interested, their arguments are usually made in bad faith, and usually they hold a double standard when it works in their favour versus when it goes against their interests.
This was a promising opening, and then you went and turned this line of argument on ordinary folk rather than the vapid techbros who are almost certainly the most guilty of it.

I've said this before and I'll say it again: Pretty much anytime there's been some technological revolution that has produced any significant efficiency gains, two things have happened. There has been a significant wave of technological unemployment, and jobs have been created directly and indirectly for those displaced. What we've been seeing in recent years is a distinct lack of the latter, especially with AI. Which is somewhat ironic, given some platforms that involve UGC are decidedly understaffed and simply refuse to hire more people for certain functions.

>>28654
>We can create whatever rules we like, but anyone with a GPU cluster can train a model and a lot of those models will inevitably end up on Huggingface.
One of my favourite AI anecdotes of recent times was the case of someone uploading a model and entering it into certain benchmarks having trained the model on the test data for the benchmark.

There are a not-insignificant number of randos using modified generative models that strip out metadata. They do this because they don't like "watermarks", without realising that the reason that the default Stable Diffusion distribution includes a "made by Stable Diffusion" watermark comment is so that future iterations don't train on it, because it turns out that training a generative model on the output of a generative model is the AI equivalent of inbreeding.

>>28655

The post actually acknowledged that itself, lad. If you're going to be that uncharitable why bother even engaging? You know that wasn't the substance of the point.

>>28658

And yet it did it anyway.

>>28656

>What we've been seeing in recent years is a distinct lack of the latter, especially with AI.

Imagine what could happen once we've got self-replicating AI. At the moment, we've got humans designing AI models and AI engines. But at some point, the next step could be AI systems designing AI systems all on their own.

But all in all, I think AI is just another hype. First it was cryptocurrencies and blockchain, and now that all the frenzied enthusiasm surrounding it has died down and it has become clear that it's not going to change the world as we know it, AI is the current next big thing.

The tech sector has a habit of investors always chasing whatever craze is en vogue during a given time period and throwing untold sums of money at it, but then at some point the excitement always pivots, and it comes crashing down again. In some ways, I can't imagine AI being all that different. Not that it isn't here to stay as a technology as such. But the long-term impact is probably going to be far smaller than many may think at the moment.

>>28660
>But at some point, the next step could be AI systems designing AI systems all on their own.
The current state of research suggests that this sort of chronic incest will go just about as well as it ever does.

>>28661

Unsupervised self-learning is an established and highly successful approach in machine learning.

https://en.wikipedia.org/wiki/MuZero

Am I completely out there for thinking there're public AIs and top secret multinational lab AI's, between which is a gulf of development potential and ability?
S'beeing the only one with a hoverboard.

>>28662
That's cool, but it's not what was being discussed.

>>28653
>Why should humans have to waste their fleeting time on this earth engaging in all this pointless drudgery?
Most jobs nowadays have some sort of sales element to them. In 1900, almost no jobs were sales jobs. But as factories and so on were automated, nobody decided the old factory workers should be allowed to just sit at home and still get paid; new jobs needed to be created for them. Pointless busywork jobs, and jobs bringing in more work for the robots to do.

At some point, I like to think that society will indeed abolish worthless time-wasting jobs that only exist to give the general populace something to do so they can keep getting paid. If AI brings about that revolution, then indeed I am all for it, but I really, really don't think it will.

>>28665

>But as factories and so on were automated, nobody decided the old factory workers should be allowed to just sit at home and still get paid; new jobs needed to be created for them.

Also, the system would have collapsed with all of those workers turning from taxpayers and consumers into dolescum. And that's the thing still today - even if AI could make 80 percent of all human employees redundant, ultimately entire industries would be shooting themselves in the foot because at the end of the day, somebody still needs to have the money and income to buy end consumer products. And even the most B2B products and services usually end up facilitating the production of consumer goods, so if there's no consumer demand because nobody earrns money anymore, then the system will collapse.

Ultimately, I don't think AI will be as transformative as some people think. It will have an impact on the workforce and the job world, but it's not going to be the all-overturning revolution that some people make it out to be. Just look at how the blockchain craze has died down, and other technological crazes before it.

>>28637

> How do I get ready for tomorrow's world with all it's AI thingamajigs, doohickie metamaterials and emergent wire spaghetti?

I've been pondering this question for the last few days. I think the best advice I can give in the short-to-medium term is to hedge your skillset - whatever you do for a living, learn to do something completely different.

I think one of the biggest misunderstandings lay people have about AI is the nature of the rate of progress. It's incredibly non-linear and incredibly unpredictable, for both technological and social reasons; rather than a steady march of progress, we see years of apparent stagnation interrupted by sudden surges of rapid change.

Large language models like ChatGPT emerged from a single theoretical breakthrough in a 2017 research paper called Attention Is All You Need. This paper introduced the concept of the transformer model, which all modern LLMs are based on. It's only about ten pages long, it isn't particularly dense, but a fairly subtle shift in strategy created a revolution in how AIs worked with natural language. Similarly, we saw absolutely vast improvements in image generation with the discovery of something called a diffusion model, which completely overtook the previous approach of generative adversarial networks.

A few years ago, the assumption was that AI would have the biggest impact on routine, structured work that we traditionally think of as being suited to computers. Nobody predicted that we'd make these huge advances in what we might call "creative" work. Nobody predicted that journalists and illustrators would be next in the queue to have their jobs taken by machines. A lot of our intuitions about the difficulty of various tasks turned out to be badly wrong.

On the social side, there are complex barriers to adoption of technology that greatly affect how quickly technological change actually impacts the real world. The clearest recent example is video conferencing - Zoom and Teams were already really good in 2019, but it wasn't until the pandemic that we actually started to change our working habits and adopt those technologies. Today I'd laugh in your face if you told me to make a five hour round trip for a one hour meeting, but it was the sort of thing we just accepted until we were forced to try something different. The technology was ready, but we weren't ready to make use of it.

There are lots of legal barriers to the adoption of AI. The obvious example is self-driving cars - Waymo have proven that their cars are safer than the average human, but they're strictly limited in terms of where they can operate because nobody is sure who to sue or prosecute if someone gets run over. The technology is arguably ready now, it's certainly far more advanced than most people realise, but the real-world potential of that technology is being bottled up by legislation and liability concerns. In a field like accountancy and law, there is very obvious potential for AI to do a lot of the donkey work, but there are also very obvious liability concerns about having an algorithm submit your tax return or draw up a contract; again, I'd expect to see a long build-up of quiet progress behind the scenes, followed by a sudden surge of adoption when we figure out the legal stuff.

For these reasons, I think it's a fool's errand to try and predict the impact of AI on society. It's just too hard to guess about the complex interactions between technology, society and the law. What we can do on a personal level is make ourselves more resilient to sudden shocks to our livelihood. If you knew that you'd be made redundant in 18 months and wouldn't be able to work in your current field any more, what would you do? What training or qualifications could you get to start a new career in something completely different? Do you have a hobby or interest that you could develop into a business? Whatever your answers to those questions, I'd advise you to get cracking.

>>28667
>Waymo have proven that their cars are safer than the average human
To their own satisfaction, sure. They have investigated themselves and found that they are doing nothing wrong.

https://www.youtube.com/watch?v=PCYTqDSUbvU

AI could be killing music as we know it.

Now that singing voices have reached this level of AI perfection, before long you'll be able to automate the whole thing to a one-click process. You'll have AI churning out ready made songs at a rate of a few dozen a day. Just like AI is churning out books that are sold on Amazon. No more expensive studio hours, no more having to pay singers or musicians. You can just press a key and go straight from your AI music workstation to streaming.

>>28669

What's the difference between a drummer and a drum machine?

With a drum machine, you only have to punch the instructions in once.

In all seriousness, brilliant singers are ten-a-penny. If you're at a studio in Central London and you need someone to sing on your record, you can ring a fixer and have a session singer turn up within a couple of hours. They will sing whatever you want in whatever style you want and you'll be blown away by the first take. The standard fee for a session of up to 3 hours for up to 20 minutes of recorded music is £130. Unless you choose to name that singer as the artist (or give them a ft. credit) then they aren't entitled to a penny in royalties from sales or streaming; they'll get a small cut from the royalties payable for broadcast use.

Musicians are not - and never have been - a meaningful bottleneck in music production. The brutal reality is that the number of superbly talented musicians vastly outweighs the amount of paid work available for them. The Musician's Union minimum rate mentioned above hasn't increased in over a decade, because schools like BRIT, LIPA and ACM have absolutely glutted the market with talent.

If you're listening to a pop song on the radio, it's more likely than not that most of the parts that didn't come out of a computer were performed by session players on that £130 standard fee. There is a real possibility that the main vocal has (in part or in whole) been overdubbed by a soundalike, often without the knowledge of the artist; it's often easier to just bring a professional in out-of-hours and bung them a bit extra to keep their mouth shut, rather than trying to cut together 47 different takes by some drugged-up starlet who can barely stand upright, let alone remember the lyrics.

When AI learns to run a marketing campaign, then the industry will be genuinely rattled.

>>28670

On that note, Spotify have recently been "exposed" for stuffing their curated playlists with licensed music that's churned out in a factory and released under a variety of fictitious names. It has been common knowledge in the industry for years and the economics are obvious - Spotify might pay a pittance to artists in royalties, but it's cheaper still for them to commission music under a commercial agreement that undercuts those royalty rates. If you've just queued up the "Stress Detox" playlist, you aren't likely to notice or care whether the song you're listening to is a commercial release by a real artist, or something churned out specifically to pad out those sort of playlists.

https://www.theguardian.com/uk-news/2024/mar/19/swedish-composer-johan-rohr-becomes-spotifys-most-famous-musician-youve-never-heard-of

>>28669

Yeah, as a musician, anyone in music to make money is a fucking idiot.

The music industry is the canary that should have alerted people to all this shit already, and you could literally say that going as far back as the 60s. All of the problems people are suddenly freaking out about with AI have been there in the music industry for decades just purely by the economics of the business and the market.

The problem isn't AI, the problem is capitalism.

>>28669
Welp, there goes my business idea. My speciality is ML applications to signal data, I should've tried music generation years ago when Attention is All You Need was still a fresh paper.

>>28673
I guess you'll have to resign yourself to putting up with overinflated salaries in the field of anything AI-related instead.

>>28674

Ride that wave while it lasts.

Just look at crypto.

>>28675
I don't know, I don't think any of the AI types are going to end up in prison for fraud.

I know I'm late to this, but... $100 billion?!

https://www.reuters.com/technology/google-ai-chatbot-bard-offers-inaccurate-information-company-ad-2023-02-08/

>>28677

Google are currently valued at $1.87 trillion. Their valuation regularly changes by hundreds of billions per week.

>>28676

No, but you'll still be out of a job when the hype dies down.

>>28679
Many cryptobros still somehow seem to be employed long after the hype died down.

>>28680

Fine. Go and be an AI bro then.

>>28681
Do they exist outside of LinkedIn?

>>28682
This probably says something about me that I don't like but I've met a few cryptobros still about even during the ice age we just had. All have been wealthy women who travel the world and focus on the networking aspect above all else, which is a given I guess for the industry but it always felt a bit tenuous to me. One made a small fortune off of NFTs back in the hype that she converted into buying a big property in London but for her it was a side hustle for whatever lizardman shit she does in pharmaceuticals.

Yes I said women. They hide their gender in online communities because they'll be automatically excluded from projects if anyone finds out but then there's the parallel real life level that exists because of the money involved where people meet up and coordinate. It's fascinating in a way and you absolutely can piss off online crypto communities by pointing it out that their safe-space has been penetrated by women, ethnics and people who regularly shower.

The Knife Warehouse email newsletter, asking me to buy more knives, just sent me this video. I don't know what to think. It's a really cool idea, to have an awards show for knives with celebrities presenting the awards, but can it really be legal?

https://www.youtube.com/shorts/5bQLYpFqJbw

https://www.theguardian.com/technology/2024/apr/02/musicians-demand-protection-against-ai

>Billie Eilish, Nicki Minaj, Stevie Wonder and more musicians demand protection against AI

>A group of more than 200 high-profile musicians have signed an open letter calling for protections against the predatory use of artificial intelligence that mimics human artists’ likenesses, voices and sound. The signatories span musical genres and eras, ranging from A-list stars such as Billie Eilish, J Balvin and Nicki Minaj to Rock and Roll Hall of Famers like Stevie Wonder and REM. The estates of Frank Sinatra and Bob Marley are also signatories.


Looks like I was right. They're all shitting themselves.

>>28689

God what a loss to art it will be if they replace Billie Eilish with a computer, can you imagine.

>>28690
The Daily Mail's that way, genius boy.

>>28691

I'm a musician myself mate. Don't give me some contrarian fuckwit argument about how pop music is good actually, because it's not, it's pure fucking shite. Handing off pop music to AI would just be the logical conclusion of the production process.

>>28691

I concur with him on Billie Eilish though.

>>28694
A famous one?

Anyway, of all the modern pop stars who are terrible, I think Billie Eilish is one of the most unique, and so she shouldn't be as worried as the pop stars who don't have their own very individual style. Any AI that just mumbles dispassionately over forgettable substandard trip-hop will only ever be a Billie Eilish soundalike, because nobody else has built a career out of literally just talking over the background music for an N64 game. She should be okay, because I assume her fans love her for the fact she has invented a whole new style of music, which AI will struggle with. They certainly can't love her because her unique style is actually good.

Of course, having such an instantly recognisable style could be a poisoned chalice, because if you want music that sounds like Billie Eilish, just ask the AI for music that sounds like Billie Eilish and you'll get it right away. If you wanted music that sounded like Harry Styles, say, that would be harder to emulate.

>>28696

> if you want music that sounds like Billie Eilish, just ask the AI for music that sounds like Billie Eilish and you'll get it right away.

There could be a time soon when Alexa or Siri will be able to create an entire AI pop song on the fly for you, with minimal lag.

Will AI struggle to reproduce country music's three chords and the truth model, or is it all heartaches by the number?

>>28694
Making something bad even worse is the attitude of a specifc kind of stupid teenager, don't be a fucking idiot.

>>28700

You have to know some battles just aren't worth fighting, and saving the mainstream music industry from its decades of creative bankruptcy is absolutely one of them.

Music thrives on the underground, music fans will go out of their way to discover good music, and that audience exists in parallel and almost completely mutual exclusivity to the mainstream industry. Let them have their AI, it is of no consequence to us.

>>28701

The mainstream music industry has never really recovered from the loss of its old distribution and monetisation model of the mid- to late 90s. That was when the music industry was last doing very well, with still-increasing CD sales both for new artists and reissues of older vinyl music. Theirs was the only distribution channel through which any artist could profit from their music in a meaningful way. And it bred complacency. It also meant they weren't forced to innovate and broadly foster new talent. They were happy shoving heaps of uninspired, mass produced mainstream music down consumers' throats. Which is why a lot of mainstream popular music from the time sounds so bland and exchangeable. Because even with that, they were still making money hand over fist.

Most of you lads are probably too young to remember the impact that filesharing via Napster and Gnutella really had. It was nothing short of a game changer. People have always made dodgy home copies of CDs and vinyl records for themselves and all their friends. But the ease with which you could now do it online on a massive scale challenged the music industry's distribution monopoly via physical media and became an existential threat, and had the music industry playing catch up for years, with dwindling profits that have only partly been recovered through modern-day licensing models. To a point where nowadays, it's feasible to start your music career on your own simply via social media.

>>28702

>Most of you lads are probably too young to remember the impact that filesharing via Napster and Gnutella really had.

You flatter us, but most of us remember dialup.

>To a point where nowadays, it's feasible to start your music career on your own simply via social media.

That depends on what you mean by "career". Even before the pandemic, it was incredibly difficult to make a full-time living as a performing artist. A moderately successful rock band bring in about the same money as a moderately successful stand-up comedian, but they've got to split it four or five ways and lug a vanload of gear around the country. You would be surprised at how high a profile someone can have in the music industry and still need a day job (or rich parents).

>>28703

>That depends on what you mean by "career". Even before the pandemic, it was incredibly difficult to make a full-time living as a performing artist.

I wasn't saying you could live off advertising on social media for the whole of your career. But it's a place to start, where you don't have to jump through the hoops of talent scouting music companies dictating everything from your musical direction to the clothes you need to wear to suit the image their marketing department has carved out for you.

You get to be your own person and start earning modest money without all that. Where you go from that, is then up to you.


> You would be surprised at how high a profile someone can have in the music industry and still need a day job (or rich parents).

Unless you become a household name with a strong following, that sounds true for a lot of musicians.

>>28703
>>28704

>You would be surprised at how high a profile someone can have in the music industry and still need a day job (or rich parents).

>Unless you become a household name with a strong following, that sounds true for a lot of musicians.

I'm always telling people this and they just have a hard time accepting it. People still think it's the 70s with the sex drugs and rock and roll lifestyle. They find it hard to accept that even artists who are reasonably "big" are often still basically earning subsistence wages, and often saddled with huge debts from recording advances that the label forced them into while failing to adequately market the resulting record or fund the supporting tours etc.

The fact is that these days you are either a "manufactured" star- You had the right connections or sheer good fortune to strike a deal with one of the big labels early on, or else you have to spend years and years grinding away in the gutter and the eventual success you gain is offset by the years you spent making a loss on the venture. Anyone who makes it organically in the music industry nowadays either has the commitment of a saint and the dedication to spend years sacrificing any kind of stable income and lifestyle, or else they're a posho like Ed Sheeran who could afford to bum around on mummy and daddy's money until their break came along.

It is not merely a question of taste for the music, it is not merely a disdain for the catchy hooks and lowest common denominator lyrical themes, that allows me to dismiss an "artist" like Billie Eilish*. It's the moral repugnance of the entire system. The music industry perfectly reflects the decline of social mobility we have seen in all walks of life in the 21st century. It's an insider's club that runs on nepotism, and if you can see why that's a problem in journalism or even politics, you should be able to see why it's a problem in the music industry.

*_{After all, at the end of the day she is nothing of the sort, she's just the pretty marketable face, her brother writes the tunes while a raft of session musicians and producers and engineers are responsible for making the actual music.}