Today's systems are mostly good at following instructions. But push them far enough, and you already get weirdness like alignment faking, instrumental self-preservation and more. We know because we've seen it in lab settings, while probing for extreme failure modes on purpose - but the world is large and strange enough that edge cases like that are liable to surface naturally.
The reason why none of this has exploded in our faces isn't that today's AIs never want to do weird and dangerous things. We know they do, at times. It's that today's AIs are incapable of pulling them off when they try.
Capability is the thing that matters, at the day's end.
Not as far as I know, but I have personally seen my coding agents take on a prompt like "See if you can port this project to typescript", and work for hours, defining a myriad of subgoals and continuously summoning and managing subagents while developing ad-hoc tools and skills.
There is to the best of my knowledge no fundamental limitation to having an agent/claw go on like this for 80 years with a prompt like "live your life to the fullest".
All the time, yes. But you have to keep two things separate in your thinking:
- Prompted as in prompt made of tokens -> for LLMs, tokens double as a clock signal. Time only flows when tokens are pushed through them.
- Prompted as in specific request placed in the stream of tokens -> Yeah, they do that all the time whether it's getting into infinite loops of repeating same pattern, or suddenly deciding to do things based on inputs they normally ignored.
Also don't forget that everything is a "prompt" for LLM. All input tokens end up in the same place.
So without a token pushed into them they do something? Not sure I understand...
In the current UIs is there a lot of suppression then as I have not seen things start on their own?
I meant an LLM doing something without any external prompt at all. Not doing something different etc but rather do something without a token/prompt ever flowing to it.
String theory is in fact falsifiable contrary to popular belief. It's just not practically falsifiable with current (and likely future, for a while) technology as the energy scales we need to probe to falsify it are astronomically large.
The new thing lately is ETFs that are "Whole-Market minus Microsoft" or "Whole-Market Minus Magnificent Seven". You can achieve similar ends by combining a whole market fund with direct short positions if you're an institutional investor, but it gets a little needy of your attention and your calculator and your fees to maintain those positions as a low-cap retail investor (just buy a put a day or something?).
I am explicitly avoiding these indices until this exploit is fixed. The lack of diversity in SPY and the like is already bad enough without these pump and dump schemes being added to the mix.
Buy alternative ETFs with similar performance and low fees. VIG is one example.
From your statement it seems like you're underestimating it's impact.
Their impact would be felt across the whole market by their sheer valuation - even if you tried to exclude them specifically.
So yeah, if eg ai crashed and took Nvidia, meta, goog or MS for the ride... You'd have massive impacts all across the board, even if you specifically tried to exclude them from your index, just because of how gigantic it's share on the economy is.
But this is purely theoretical. It can only be considered an opinion until something actually happens - because the market has never been in a situation like this before - no matter what some people may claim.
If you have a $10B fund which owns things evenly across the nasdaq, and a new company arrives, you have to buy shares in the new company, which means selling existing shares or getting more funding to balance.
So you buy $1B of SpaceX stock at the IPO price, meaning you have to sell 10% or $1B of existing Nasdaq stock, which (when combined with every other fund) lowers the cost of the other stocks.
If I own stock in everything-but-spacex, then I'm seeing the price collapse because everyone else is selling.
Then if spacex stock collapses to half its initial price, your $1B becomes worth $500m, you get margin called, and you have to sell more stock, pushing the other stocks down, and the problem cascades
Of course the recession that would likely occur if/when the AI spending merry go round comes to an ends would have a negative impact on the entire economy. In that scenario, would you rather own a stock that's going to zero, or one has taken a hit because of the other company going to zero but still has a viable long term business?
Would you rather have an unusually large concentration in a few very specific companies in your portfolio, or have increased diversity (and therefore lower risk) in this unprecedented scenario?
There is a nice book by Jonathan Haidt about this called The Happiness Hypothesis.
He uses a "Elephant and Rider" psychological metaphor to describe the internal battle between our rational mind (the rider) and our emotional, intuitive drives (the elephant).
FWIW creatine is "one of the most studied supplements for muscle and strength".
But at the same time "creatine’s brain benefits aren’t as exciting as social media makes them out to be. The research at this point just doesn’t support the hype".
I think people often overestimate the effects of these things. It absolutely works for muscle growth but it's not like a steroid or something. Similarly there is enough evidence to suggest that it actually does have some small effect on cognition; I remember a study 10 years ago showing that in people who are creatine deficient (vegans) it improved cognition scores. But it's not going to be a huge effect for someone whose not deficient in some way.
It would actually make sense that as you age and eat less you might get creatine deficient so sure. I don't think it's bullshit, but it's not going to be a huge noticeable effect either.
All of this reminds me of people who don't weight lift "because they don't want to get built." They somehow think you lift some weights and boom you're looking like Arnold. No, it doesn't work that way.
Some companies use product managers for this purpose. They translate the implicit domain knowledge of those experts into explicit requirements, mostly through interviewing.
reply