You are given limited access to the Gearworks of Reality. You may delete one of these from human civilization.
44
Replies
2
Boosts
@mttaggart neither is in the top ten of things I worry about, so probably "neither". I might feel inclined to "investigate" what else I might change. :)
I guess you practically asked for it.
@mttaggart I obviously understand there is a distinction. But I also don't think I'd live in the country with the highest data centre per capita ratio in Europe if venture capitalists weren't betting on betting smart enough to get out before the bubble bursts. It's how much they keep pushing it beyond the scope of utility that makes it such a scourge. Its the griftiness of it all.
@mttaggart LLMs are gambling, the stock market is gambling, the same way prediction markets are gambling. Therefore I will erase gambling and GenAI will subside overnight because everyone will lose interest since the only shiny thing about it is gone
@mttaggart As reprehensible in every way gen AI has become, I didn't hesitate to choose gambling. That's the problem that would require Gearworks of Reality-level powers to eradicate. With regulation and enforcement, gen AI can be rendered safe enough. It's doing great damage in the meantime, but that's on us humans, who could have prevented it. No human could prevent gambling.
@mttaggart Shame it's only one!
@mttaggart
I feel like removing gambling may fix enough things so that AI also dies.
If gambling on stock is no longer a thing a will a lot of investment would go away.
@mttaggart I choose gambling because generative AI is not enough. I want all AI to fucking burn.
@mttaggart Do you mean like, the different ways one can gamble, or the inborn human urge to gamble? I’m kinda assuming the latter, because the former is a legal policy that is just Prohibition - with all the success and longevity that comes with it - and the latter would require access to reality’s guts to change.
So definitely the human nature one. Much better return than banning something.
@mttaggart Well, you could certainly argue Generative AI is a form of gambling
@mttaggart Truly, neither. Not our biggest problems. Show me the glitch that predisposes us to confirmation bias - I’d like to iron out that one.
@mttaggart
Since generative AI is only so prominent because the tech CEOs are gambling they can create the fabled singularity before their financial bubble collapses, eliminating the human desire to gamble will eliminate destructive behavior like they're doing with AI now.
@mttaggart gambling (recreational) is almost impossible to eradicate, as it is very low tech and essentially a construct of mind, so you'd have to alter the very nature of human cognition as it depends on our ability for risk analysis.
People can bet on which stick floats down a river fastest.
Generative AI is high-tech dependent and has yet to be shown as even vaguely efficient in resource usage - it hasn't even *started* to break even at scale.
The price signal of using it has yet to be determined.
Removing genAI is the only achievable one that renders a recognisable world, and every adult alive has lived in a time of no genAI.
The same cannot be said of gambling (recreational).
@mttaggart get rid of gambling and end GenAI use as a result 👍
@mttaggart isn't generative AI also gambling?
@mttaggart If gambling includes the investment & stock market, which it should, then that - it would take most of our other problems away too.
@mttaggart these_are_the_same_picture_dot_jpeg
@mttaggart they’re the same picture
@ajn142 I don't know that I've ever seen an employer require employees to sit at a blackjack table outside of Vegas, but I do agree there are similarities.
@mttaggart I mean, both amount to "put money in, trigger nondeterministic statistical system, maybe get something back, repeat". It's just that employers have been sold the idea that putting money in and pulling the lever until they get something usable is less expensive than hiring someone who can do the work without GenAI, or will let that person do so much more work for the same wage, but they don't recognize the equivalence to feeding the rent money to a slot machine.
@mttaggart There's a difference?
@mttaggart Limited only to casinos, prediction markets, etc.. GenAI and it's not close.
Far more destructive.
@distractal Watching what gambling's done to a lot of lives, I'm really not so sure, hence the poll.
@mttaggart As a Methodist, I chose gambling. Maybe we can compromise somewhere on no one being a billionaire.
@vwbusguy Extreme wealth and poverty is such a low bar to clear for a just society, but right there with you
I don't want to remove responses, but for clarification, I am referring to recreational gambling such as casinos, sports betting, and prediction markets. An overbroad definition of "gambling" is pointless.
Edit: And yes, day-trading and otherwise gamifying securities markets.
Fascinating to see how many folks think these are identical.
I remember once being called a "lumper" in school—trying to make the case that many different categories were in fact indistinct, thereby rendering the discussions of differences moot. This is in contrast to "sorters," who prefer to highlight differences.
I can see how these two things look similar, but I also think it would be imprudent to regard all the tools of oppression without distinction.
Or perhaps y'all think one is a subclass of the other? Are LLMs just gambling? I find that a tad reductive, if I'm honest.
I guess I turned into a sorter with time.
Venn diagram of 4 different cognitive associative classification mechanics/methods people apply and each individual has their preferred favorite.
These are like tools in cognitive toolbox, use the appropriate tool for the challenge, not “law of instrument”- https://thedecisionlab.com/biases/law-of-the-instrument
@mttaggart we're just looking for a loophole to get rid of both.
(Although to be honest, I like the transformer model a little bit for options it gives to human computer interfaces, but I really don't like how many people are gambling everything on LLMs doing more than just predicting the most likely next word)
@mttaggart LLMs as gambling is horseshit. That's like saying statistics is gambling. And cryptocurrency = gambling. People, save your anger for the things that are truly evil.
@codinghorror @mttaggart How about an argument that investment in LLMs is turning into gambling? (I'd accept early investment really was investment, but I think at some point the desperate pursuit of 'AI' by governments and corporations is not going to be distinguishable from problem gambling - kind of "I've got to keep going to recover my losses".)
@codinghorror @mttaggart operating AI is closer to gambling than it is to statistics. If it answered consistently I might say otherwise, but it doesn't, & that's not even disputed.
Crypto is closer to fraud than gambling, despite many people gambling on crypto.
AI very much is something to be angry about. I'm a little confused as to why you'd think otherwise?
@codinghorror @mttaggart You give it a prompt. You push the button. Maybe you get what you were looking for, most likely not. You give it a prompt, you push the button…
In terms of behaviour and feedback that seems fairly similar to playing a slot machine.
> Fascinating to see how many folks think these are identical.
I came here to mention I think there's an overlap. But I think I could also argue generative tools can also be a higher-level paintbrush, which gambling never is.
@mttaggart Most people are sorters in my experience. I have that experience because I'm often accused of being a lumper. Thing is, I'm neither.
If I see a duck and a goose, I don't think of them as the same. I don't think of the goose as a big duck, or the duck as a small goose. I see the value of each.
But I do think that their similarities warrants a waterfowl category, and some properties most waterfowl share.
People tend not to have patience with that and accuse me of lumping.
@mttaggart Anyway, genai and gambling share something, yes.
The commonality is the variable ratio reinforcement schedule. You can't be sure that the next prompt is going to reward you with a good result, so you pull the one armed bandit's lever again. And again. And again.
And you focus on the great results you sometimes get, and not the many in-between failures. Those are explained away with "prompt skill issues" and whatnot.
@jens @mttaggart I'll tell you this .. you will almost always get terrible results on any question (beyond incredibly trivial, simple questions) unless you pick the longest think modes. Gemini is the worst for this.
@codinghorror I'm not known to be a fan of LLMs, and never got great results out of them. But then, I don't keep trying... 😂
I'm more of a schmeerer then. There are distinct boundaries, but there's lots of schmeer in between.
I had so many arguments in college about the definition of a species. Turns out that homo neandertalis is homo sapiens (neandertalis).
There are places where bright lines can be drawn, I'm not sure this is one of them.
@johntimaeus Here are a few worth considering:
- Gambling actually makes money for the game operators (unless you're DJT, who couldn't even make a casino profitable)
- The ability to create a gambling operation is far more accessible to more people than LLM generation
- Nobody is claiming gambling is the key to a new future for humanity. While that's marketing, I do think some of the leaders of the adventure are true believers.
- This last one is a matter of degree, so definitely fuzzier. But the "win rate" for LLMs is, objective depending, not nearly as poor as recreational gambling, IME. It would be much easier if the technology never worked. But in many cases, it does upsettingly well.
@mttaggart One could argue that generative AI is a form of gambling, but so would farming and moving to a new town.
@linuxandyarn I think you need a useful distinction between recreational gambling and risk-taking.
@mttaggart generative AI wouldn't be so bad if the world's wealthiest and most powerful weren't gambling so much wealth on it being able to replace human labor.
