Rockstar co-founder [Dan Houser] compares AI to 'mad cow disease,' and says the execs pushing it aren't 'fully-rounded humans'
submitted by
www.pcgamer.com/software/ai/rockstar-co-founder…
Madiator's Corner
“So it’s sort of like when we fed cows with cows and got mad cow disease.” is an amazing analogy for the current state of LLMs.
Yeah, the headline makes it sound like he’s insulting AI, but he’s just illustrating a fairly basic fact…
Oh theses cows went wobbly, fell over and died, better feed them to our good stock to save money.
To bad there isn’t an AI prion disease for them to catch.
It isn’t – it is a completely populist comparison that can’t be used as a meaningful argument.
CEOs work by this very same principle: say what people want to hear, making their beliefs even stronger and thus increasing their craving to hear that statement again. Repeat until something breaks.
You know analogy and argument are not the same thing, right? Like, not at all. I’m not the guy you replied to, but I definitely did not get the impression he meant anything even close to “meaningful argument”.
And you know what? I think he’s right; that is a very apt analogy when taking about sophisticated bullshit generators. Cause that’s the version they keep trying to sell to the public. This isn’t about data analysis or any other if the creative scientific applications of this technology.
But this is the only thing that matters. Who cares how people who don’t know, care or capable of using some technology are actually using it? Ammonium nitrate is poisonous if you try to salt a soup with it, but it makes miracles in the hands of those who know how to use it (if you like analogies)
I am not a lawyer, but I think that even in the USA it isn’t mandatory to buy everything that corporations sell.
Modern LLMs are very useful tools. Just some people try to smalltalk with it for some reason. Their right. Salting soup with nitrates is also their right.
Ugh yet another Lemmy user desperate to make the LLMs like them. Go chat with your chatbot friends and leave humans alone mate.
the analogy perfectly describes model collapse.
Thanks Lembot.
Everyone’s talking about the mad cow part, but this is also a really excellent point:
“Some of these people trying to define the future of humanity, creativity, or whatever it is using AI, are not the most humane or creative people. So they’re sort of saying, ‘We’re better at being human than you are.’ It’s obviously not true.”
Humanity and creativity are not on the CEO’s resumeé. Making shareholders happy by increasing profits and padding the CEO’s ego certainly is.
Please be a nice bubble and pop soon, AI.
Stupid question: if you think it’s a good idea but don’t know when the price will go up, you just buy stock and wait. But if you think it’s a bad idea and don’t know when the price will go down, is there any long-term alternative to shorting that doesn’t require betting on the date?
If you imagine it like making a bet, nobody’s going to take a bet with you where they pay you when it pops, but there’s no time after which you pay them – because they’d never get any money out of that bet. Buying stock is different because it’s a thing you can own, but you can’t invest in the idea of something failing, because there isn’t any business which will take your money and make something more likely to fail.
You could buy every stock except AI-related stocks, which I believe is functionally equivalent to buying an index fund and shorting AI stocks based on the percentage of AI stocks in the index fund. You could also think about what businesses would do well (or less poorly) in the case of an AI-instigated crash, and then buy those.
Yes, you can with derivatives: buy out-of-money puts.
Derivatives are financial instruments that pay out based on market movements. A classic example is crops: using derivatives, farmers can, essentially, “lock in” the price they sell their goods at. This allows them more stability, since they know in advance how much they’ll be paid for their crops. (And they’ll separately buy crop Insurance to cover their risk for crops failing, most likely.)
Puts are a derivative that is a contract for the right to sell an asset at a given price (the “strike price") on a given date. Usually, these are closed out by paying the cash value at the end, not actually selling the stocks.
Out of money means that the strike price is below the current market price. If they are still out of money at the end of the contract term, they are literally worthless. But, if the underlying asset (like NVidea stock) crashes, then you can earn the difference between the strike price and the market price.
What makes this speculation* strategy effective is that the market usually prices in a low probability of a major price decrease, so they’re (relatively) cheap. They also have limited downside risk—at worst, you lose everything you spent buying them. For deeply out-of-money puts, you can make a lot of money with a huge crash, but most of the time you “just” lose all your money.
This contrasts with short selling where you have unlimited downside risk. With short selling, you’re basically borrowing someone else’s share and immediately selling it at the current market price, then you need to buy it back from the market when you close out the position. So if you sold it for $100, and need to buy it back at $1000, you’re royally fucked. (You won’t be allowed to get that far, though; you need to keep assets in your account to cover the cost, so you’d be forced to continually “pony up” more cash as the price rises, until you can’t make a payment and you’re forced to close out the position, losing all your initial money and all the money you were forced to keep adding as it rose.)
But good luck with that strategy; I imagine NVidea puts are pretty expensive right now since a lot of people are making this exact bet. As such, people issuing/selling puts are demanding a lot of money to pay for them taking on risk.
* This is “speculation”, not “investment”. Investment requires, by definition, capital put towards productive assets—in other words, it needs to be expected to return an income stream of some kind, like interest, profits, or dividend payments. Speculation is betting on the direction of price movement on an asset—"gambling”, effectively, but with fancy investment words. Like in the farmer example above, they’re gambling that prices won’t go up, since they won’t gain any of the benefit from rising prices. That type of speculation reduces risk—unlike what you are asking about.
There are other ways that derivatives can reduce risk, but that’s not what you were asking about here.
Puts are rights to sell, not to buy. And with that, the rest of your post actually makes sense (you want to sell above market price).
Oh, shoot. That’s a silly mistake to make. lol.
Fixed.
One thing to add, prices can be manipulated in the short term to make or avoid certain options from getting into the money. They won’t do this to target specific individuals, but there’s a value called “max pain”, which is the price such that the most puts and calls expire worthless and the ones that are in the money pay out the minimal value, when all outstanding contracts for an equity are considered in aggregate, and prices trend towards those at expiry time.
Am I understanding this correctly?
I sign a contract with you saying “one year from today, if I decide to sell potatoes to you, you must buy them for $100”. I pay you $10 as consideration for the contract. Next year, if the price of potatoes is $110, I simply decide not to exercise my right. You don’t get potatoes, but you keep your $10.
If the price of potatoes is $90, then I buy potatoes for $90, and you buy them for $100. I broke even. Or more realistically, you send me $10 cash, no actual potatoes change hands.
If the price is $95, then I only lose $5, after you send me $5 back.
If the price is $80, then I made $10 in profit, after you send me $20.
And obviously, you might on-sell that contract, in which case you’re up or down the difference between $10 and whatever you sold it for, and I go to them to sell my potatoes/get my cash back, if the price of potatoes ends up below $100.
Is that right? If so, I have two follow-up questions:
Holding cash is a position. It also gives you the most flexibility, and low risk, but also low reward.
That’s a good question I dont have an answer to. Maybe there are ways to short where you can just hold, but I dont know how. Maybe there’s a way to borrow lots of RAM and GPUs, sell them, then buy them back when the price drops and sell them for cheap back to whom you borrowed. But I dont know who would make that deal.
You can hold a short position by repeatedly borrowing more stock – but you run the risk of running out of money completely, because short positions have (theoretically) infinite downside risk.
Financial risks can, by themselves, never be infinite. They are by nature quite finite. Also, risks are always alluding to a downside. Otherwise they’re called chances.
Granted. “Arbitrarily large” would probably be a better phrasing: if I buy a stock for $100 and the value drops to $0, I’m out $100. Can’t lose more money than I put in. What I meant is that short positions, by their nature, don’t have this ceiling on the amount of money you lose.
Stock market derivatives are essentially all betting. You won’t get someone to bet with no date to resolve the bet: after all, you might just hold on to it until you kick the bucket.
The move equivalent to buying stock when you think it’s going to go up is to sell stock you own when you think it’s going to go down.
Or you can look at the series of actions, where buying when you think it’ll go up is just step 1, then step 2 is wait for it to go up, and step 3 is sell it for a profit, and step 4 is look for the next stock you think will go up (or wait and hold the cash if you don’t think any will).
In which case you can do step 3 if you own the stock, or step 4 if you don’t. Then, if it does crash (and the crash is stock prices and not the currency itself, like what happened to a degree in response to the money printed after 2020), you can buy back in at the bottom and wait for it to go up.
But if the fed pumps money into the system to prop up the stock markets, or the government bails out firms that might go under, then that money can be used to keep the stock prices high. And with the richest 1% owning such a high portion of the entire economy, if they have a lot of cash, they could also do that without any help from the feds (reserve or government).
So depending on how a crash is responded to, the best bet might be holding cash or avoiding holding cash. Or maybe investing in some good that holds value well.
However, holding stock might still be fine, assuming the equities you hold are able to survive the crash and everything that comes next. If you look at the historic crashes, the value does always return and pass the previous before crash value, at least on average. You won’t get rich playing it like that but you might not lose those unrealized losses unless you’re in a position where you have to sell.
i’m still convinced the only reason they’re pushing it so hard is to invalidate their Kompromat….
we’re already at the point where people don’t believe videos they don’t like.
Well, he isn’t wrong.
…while his multibillion dollar company is union busting.
He left Rockstar a few years ago.
Did I say he was union busting? Read the English. He started Rockstar.
So maybe he is a little more qualified to criticize Rockstar management for being “fully rounded”.
Was Rockstar unionized while he was there? No.
I mean you said his, which is factually incorrect as it isn’t his.
Read the room bro and learn to take constructive criticism. It’s not a personal attack.
And I took that personally!
You might just be a bit of a dick.
Do you talk to people like that in person?
Total rockstar
Agreed. Moo.
Is he still working at Rockstar? If so I can forgive a lot of stuff right now.
EDIT: He is not.
While I like it as a simile, we didn’t feed cows to cows - we fed them sheep infected with scrapie, at least that’s the theory of how mad cow disease started. I say “we”, I wasn’t involved in the process, I just live here.
Mad cow disease is caused by prions from dead cow brains infecting the healthy brains of living cows. It’s kind of the cow equivalent of Kuru.
Those prions can also infect humans leading to Creutzfeld Jacob syndrome. Prions can also come from wild deer species and infect through venison.
Technically it’s called Variant Creutzfeld-Jacob Disease. The non-variant version is also known as Kuru.
The difference is transmission. Standard CJD is human->human
I thought it wasn’t just any prions, but specific prions that replicate out of control after some time, depending on genetics. So cows were fed unclean bone meal from sheep and pigs and cows all mixed together, and they think some scrapie infected sheep spine/brain got in there and it spread to cows. Then it spread from there because we didn’t think cows could get it and even if they could we were pretty certain humans couldn’t get it. But we were wrong.
Yeah, pretty sure you’re right. Though I’ll admit I’ve forgotten much more than I know on the topic.
https://en.wikipedia.org/wiki/Bovine_spongiform_encephalopathy
There is a natural incidence of prions in cows of one in a million, but we have herds this big now.
There’s natural incidence of prion disease in humans as well. I’m definitely not an expert and just remembering a video I watched a week or two ago. That the prions from different sources act differently.
Video I watched:
Mad cow disease
Also watched this one
I’m tired of people using “AI” for only generative AI/LLM. Especially in video games ?!
You’re absolutely right, but you also need to accept that most people think ai means transformers (as in, LLMs and generative/image synthesis)
It sucks, but that’s how language works unfortunately. How most people use a word ends up defining it.
As in, what, you want more? Such as…?
Recommendation systems, search and ranking algorithms, speech-to-text, text-to-speech, voice cloning, audio noise reduction, predictive analytics, forecasting models, fraud detection, anomaly detection, robotics control systems, self-driving path planning, robot navigation and mapping (SLAM), reinforcement-learning agents, game-playing AI, warehouse optimization, dynamic pricing models, object detection, face recognition, license-plate recognition, medical image analysis, sentiment analysis, keyword extraction, spam detection, classic machine-translation models, personalization engines, matching algorithms, logistics route optimization, scheduling algorithms, resource allocation models, algorithmic trading bots, risk scoring, market-making AI, healthcare diagnostics, disease-risk prediction, protein-folding models, music analysis, beat detection, autonomous drone navigation, industrial monitoring, and smart-traffic systems.
Sounds like machine learning (ML) “a field of study in artificial intelligence” more than what people mean when they say AI, which is Generative AI shortened.
Obviously depends on context of a conversation though for how you would deduce the usage of the word.
Ha ha what a list!
Cheers
actual artificial intelligence