Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

After the AI boom: what might we be left with?

blog.robbowley.net

Post reply on HN
Print view View original

431–440 of 489 posts

Re: After the AI boom: what might we be left with?

#431

Post by ACCount37 »

maplethorpe wrote: post #364
ACCount37 wrote: post #153

Earlier quoted context omitted.

Symbolic AI is dead . Either stop trying to dig out and reanimate its corpse, or move the goalposts like Gary Marcus did - and start saying "LLMs with a Python interpreter beat LLMs without, and Python is symbolic, so symbolic AI won, GG". Hallucinations are incredibly fucking overrated as a problem. They are a consequence of the LLM in question not having a good enough internal model of its own knowledge, which is d…

> Hallucinations are incredibly fucking overrated as a problem. They are a consequence of the LLM in question not having a good enough internal model of its own knowledge, which is downstream from how they're trained. Plenty of things could be done to improve on that - and there is no fundamental limitation that would prevent LLMs from matching human hallucination rates - which are significantly above zero. Why is th…

To avoid hallucinations, you, a human, need two things: you need to have an internal model of your own knowledge, and you need to act on it - if your meta-knowledge says "you are out of your depth", you either answer "I don't know" or look for better sources before formulating an answer.

This is not something that's impossible for an LLM to do. There is no fundamental issue there. It is, however, very easy for an LLM to fail at it.

Humans get their (imperfect, mind) meta-knowledge "for free" - they learn it as they learn the knowledge itself. LLM pre-training doesn't give them much of that, although it does give them some. Better training can give LLMs a better understanding of what the limits of their knowledge are.

The second part is acting on that meta-knowledge. You can encourage a human to act outside his knowledge - dismiss his "out of your depth" and provide his best answer anyway. The resulting answers would be plausible-sounding but often wrong - "hallucinations".

For an LLM, that's an unfortunate behavioral default. Many LLMs can recognize their own uncertainty sometimes, flawed as their meta-knowledge is - but not act on it. You can run "anti-hallucintion training" to make them more eager to act on it. Conversely, careless training for performance can encourage hallucinations instead (see: o3).

Here's a primer on the hallucination problem, by OpenAI. It doesn't say anything groundbreaking, but it does sum up what's well known in the industry: https://openai.com/index/why-language-models-hallucinate/

Re: After the AI boom: what might we be left with?

#432

Post by paulryanrogers »

CuriouslyC wrote: post #414
saubeidl wrote: post #402

Earlier quoted context omitted.

You're making the (imo incorrect) assumption that a revolutionary moment isn't the exact thing the US needs right now. Morality dictates we have to try and force one.

We need a change, but I don't think we have the unity and will to force a revolution right now. I think our best chance is a massive win in the 26 elections, but barring that I think we're struck bailing out the oligarchs for now, and striking back with collaborative economic warfare.

> ...striking back with collaborative economic warfare.

What does that look like?

Re: After the AI boom: what might we be left with?

#433

Post by filoeleven »

lifthrasiir wrote: post #322
dsr_ wrote: post #283

Earlier quoted context omitted.

There's only one problem with a race to superintelligence, and that's that nobody has evidence that mere intelligence is coming, much less superintelligence. (There are a thousand more problems, but none of them matter until that first one is overcome.)

Today's AI does exhibit some intelligence, period. It is absurd to claim that an intelligent-looking entity doesn't have intelligence, only because we might not be able to determine which part of the entity has one. The superintelligence is an entirely different problem though because there is no clear path from intelligence to so-called superintelligence, everything has been just a speculation so far.

Clever Hans would like a word.

Re: After the AI boom: what might we be left with?

#434

Post by CuriouslyC »

paulryanrogers wrote: post #432
CuriouslyC wrote: post #414

Earlier quoted context omitted.

We need a change, but I don't think we have the unity and will to force a revolution right now. I think our best chance is a massive win in the 26 elections, but barring that I think we're struck bailing out the oligarchs for now, and striking back with collaborative economic warfare.

> ...striking back with collaborative economic warfare. What does that look like?

I outline it in my article. Don't do business with unethical companies. If 50% of the American populace used its power of the purse in a collaborative fashion, it could kill every unethical company within 1-2 years.

Re: After the AI boom: what might we be left with?

#435

Post by CuriouslyC »

mike_hearn wrote: post #424
CuriouslyC wrote: post #421

Earlier quoted context omitted.

Serious question, why load up the army with people? Drones and autonomous weapons are 100% here now. We don't need a mass of general infantry, the new pattern is spec-ops spotting and targeting for autonomous kinetic munitions. Think ghosts in starcraft.

The only way to genuinely control captured territory on the ground is with infantry, and it'll remain that way for a while yet. America demonstrated the limits of what you can do with Predator drones in Afghanistan; the tech worked great and posed a huge threat but the Taliban was never truly defeated and continued training and recruiting at scale. For something like an invasion of Taiwan or (gulp) other territories…

That was true 10-20 years ago, but in a world with "terminators" I don't think it's true anymore.

Regarding effective conquest, we can look at the historical lesson of Rome. Conquest is effective when you can co-opt local leaders and cultures to cause them to identify with the conquering culture. Conquest that doesn't cause integration is historically unstable.

Re: After the AI boom: what might we be left with?

#436

Post by CuriouslyC »

ACCount37 wrote: post #153
paufernandez wrote: post #103

Earlier quoted context omitted.

In my case I fully grasp what such a future could be, but I don't think we are on the path to that, I believe people are too optimistic, i.e. they just believe instead of being truly skeptical. From where I look at it, LLMs are flawed in many ways, and people who see progress as inevitable do not have a mental model of the foundation of those systems to be able to extrapolate. Also, people do not know any other forms…

Symbolic AI is dead . Either stop trying to dig out and reanimate its corpse, or move the goalposts like Gary Marcus did - and start saying "LLMs with a Python interpreter beat LLMs without, and Python is symbolic, so symbolic AI won, GG". Hallucinations are incredibly fucking overrated as a problem. They are a consequence of the LLM in question not having a good enough internal model of its own knowledge, which is d…

Symbolic AI isn't dead, we use it all the time, it's just not a good orchestrating layer for interacting with humans. LLMs are great as a human interface and orchestrator but they're definitely going to be calling out to symbolic models for expanded functionality. This pattern is obvious, we're already on the path with agentic tool use and toolformers.

Re: After the AI boom: what might we be left with?

#437

Post by ACCount37 »

CuriouslyC wrote: post #436
ACCount37 wrote: post #153

Earlier quoted context omitted.

Symbolic AI is dead . Either stop trying to dig out and reanimate its corpse, or move the goalposts like Gary Marcus did - and start saying "LLMs with a Python interpreter beat LLMs without, and Python is symbolic, so symbolic AI won, GG". Hallucinations are incredibly fucking overrated as a problem. They are a consequence of the LLM in question not having a good enough internal model of its own knowledge, which is d…

Symbolic AI isn't dead, we use it all the time, it's just not a good orchestrating layer for interacting with humans. LLMs are great as a human interface and orchestrator but they're definitely going to be calling out to symbolic models for expanded functionality. This pattern is obvious, we're already on the path with agentic tool use and toolformers.

This is what I mean by "move the goalposts like Gary Marcus did", yes.

If what you're claiming is that external, vaguely-symbolic tooling allows a non-symbolic AI to perform better on certain tasks, then I agree with that.

If you replace "a non-symbolic AI" with "a human", I agree with that too.

Re: After the AI boom: what might we be left with?

#438

Post by inkyoto »

mike_hearn wrote: post #396
inkyoto wrote: post #358

Earlier quoted context omitted.

> […] because the Chinese government has started massively distorting the economy in favor of automation projects. It's widely known that one of the hardest parts of planning a factory is figuring out what to automate and what to use human labour for. … or it is an early manoeuvre – a pre-emptive measure – to address the looming burden of an ageing population and a dearth of young labour that – according to several d…

Robots installed now will be obsolete and broken by 2050. Demographic change is the sort of thing where market mechanisms work fine. You don't build out assets like those ahead of demand, you wait until you need them at which point you get the latest tech with economies of scale as demand ramps up everywhere. The most optimistic reading of this move is that it's just more of the same old communism: robotic factories…

At no juncture did I suggest that the machines presently entering commission in the year 2025 would remain unaltered in perpetuity. Technological progress proceeds with unrelenting velocity — it is both natural and inevitable that subsequent refinements and augmentations will be introduced. The systems deployed today, in all likelihood, represent but preliminary trials and iterations — a proving ground — for capabilities yet to be fully revealed.

China, indeed, possesses a longstanding tradition of curating information to satisfy the sensibilities of its ruling class — a practice traceable to the dynastic courts of antiquity. Yet, to dismiss a potential adversary solely upon the architecture of its political order is — at best — ill-advised, and at worst, a grave miscalculation. The probability of threat must be judged on capacity, not narrative. Whether such an adversary proves formidable or farcical is immaterial at present — the truth will emerge within the span of a decade or so.

Re: After the AI boom: what might we be left with?

#439

Post by f4uCL9dNSnQm »

alextingle wrote: post #361
mmh0000 wrote: post #258

Earlier quoted context omitted.

Have you seen the "average" users of LLM? They mostly don't care about hallucinations. It's like that joke[1] "Computer says no", it doesn't matter what is true or real, only that the computer said it, so now that's true in their mind. Personal anecdote: I work with beginner IT students, and a new, fun ( I find it highly doubtful that hallucinations, if unsolved, will have any real negative effect on LLM adaptation.…

I think it's possible that people will learn not to trust LLMs.

It won't happen due to Gell-Mann amnesia. It "helps" that LLMs never admit to not knowing something. So from average user point of view it looks like the agents know everything and are only under-educated in the one domain the user is actually the specialist.

Re: After the AI boom: what might we be left with?

#440

Post by dcre »

hattmall wrote: post #356
dcre wrote: post #234

Earlier quoted context omitted.

The idea that people will suddenly decide LLMs are useless after two years of exponential growth in usage is a fantasy.

That's essentially what was said about Tamagotchi in the 90s, and everyone of their "users" was a paying customer.

Thank you for proving my point. People waiting for a big, sudden crash must insist that the growth is driven solely by hype and FOMO. This is absurd. We’re talking about professionals with decades of experience using these things for hours a day in their area of expertise.
Post reply on HN