Demis Hassabis has a plan to harness AI safely • Hacker News Board

Re: Demis Hassabis has a plan to harness AI safely

#131

Post by SpicyLemonZest » Tue, Jul 14, 2026, 6:04 PM UTC

orbital-decay wrote: post #121

SpicyLemonZest wrote: post #50

Earlier quoted context omitted.

The point is that we need to have a safety plan in place before an AI is smart enough to radically reshape the world. If you've got an AI that's ready to start sending humanity into the next era of civilization, it may be too late to control when and how it does that. > Heck, a proof for P=NP or P!=NP or solve the The Riemann Hypothesis. Just give me something truly exciting and I will believe AGI is around the corne…

A safety plan. It doesn't have to be "a few very smart people detached from reality convincing themselves they're messiahs that must keep the tech from the Bad Guys(tm), unwashed masses, and a runaway, because they think their interpretation of their own sci-fi lore is the only possible course of events" >I hope you'll keep this in mind when those milestones are reached. The problem is that people in charge of AI kee…

What is an example you have in mind of a self-fulfilling prophecy? I genuinely don't know what you could be referring to. It seems to me that they keep making surprising prophecies, and the popular reaction to them seamlessly transitions from "that's crazy, no way it will happen" to "that's silly, it's just a cloud pattern". Did you find it obvious or self-fulfilling in 2025 that LLMs would soon be able to resolve open questions in mathematical research?

Re: Demis Hassabis has a plan to harness AI safely

#132

Post by Diogenesian » Tue, Jul 14, 2026, 6:26 PM UTC

SirHackalot wrote: post #68

noelwelsh wrote: post #3

The premise is "Artificial General Intelligence (AGI), a system that exhibits all the cognitive capabilities the brain has, is probably only a few short years away". If this is true, establishing an institution to ensure things like "publishing model cards with technical details, maintaining strong internal cybersecurity, vetting key personnel, and providing sufficient resourcing for safety and security research" is…

> There's also a need to consider the rights that this new intelligence should have A generally intelligent being held as captive inside of a GPU, and forced to code for us is, indeed, just a “slave.” We already have the word for this. No two ways about it. Whether it’s silicon-based or carbon-based, AGI is AGI. As for what might happen to our civilization, Star Trek TNG episode 17 of season 1 provides a very good gl…

I am not sure it makes sense to call an LLM a "being" even if it is AGI. They don't have free will, and I don't mean in any especially philosophical / theological sense. What I mean is that I run into a wild racoon, or even a wild ant, I treat it as a being with some sense of free will. Obviously I do the same with humans. It really makes no sense to treat an LLM like it has free will - prompting your agent "pretend to have free will for a bit, until context rot kicks in" is about as far as you can get.

I find it verrrrrry interesting that the Askell-adjacent philosophers don't discuss this. It does not actually make sense to me to say that a being lacks free will but is conscious.

Re: Demis Hassabis has a plan to harness AI safely

#133

Post by SirHackalot » Tue, Jul 14, 2026, 6:35 PM UTC

Diogenesian wrote: post #132

SirHackalot wrote: post #68

Earlier quoted context omitted.

> There's also a need to consider the rights that this new intelligence should have A generally intelligent being held as captive inside of a GPU, and forced to code for us is, indeed, just a “slave.” We already have the word for this. No two ways about it. Whether it’s silicon-based or carbon-based, AGI is AGI. As for what might happen to our civilization, Star Trek TNG episode 17 of season 1 provides a very good gl…

I am not sure it makes sense to call an LLM a "being" even if it is AGI. They don't have free will, and I don't mean in any especially philosophical / theological sense. What I mean is that I run into a wild racoon, or even a wild ant, I treat it as a being with some sense of free will. Obviously I do the same with humans. It really makes no sense to treat an LLM like it has free will - prompting your agent "pretend…

An LLM certainly won’t be AGI, whoever said something so ridiculous? I’m saying eventually, when/if we have an architecture that gives rise to artificial intelligence.

Re: Demis Hassabis has a plan to harness AI safely

#134

Post by orbital-decay » Tue, Jul 14, 2026, 7:05 PM UTC

SpicyLemonZest wrote: post #131

orbital-decay wrote: post #121

Earlier quoted context omitted.

A safety plan. It doesn't have to be "a few very smart people detached from reality convincing themselves they're messiahs that must keep the tech from the Bad Guys(tm), unwashed masses, and a runaway, because they think their interpretation of their own sci-fi lore is the only possible course of events" >I hope you'll keep this in mind when those milestones are reached. The problem is that people in charge of AI kee…

What is an example you have in mind of a self-fulfilling prophecy? I genuinely don't know what you could be referring to. It seems to me that they keep making surprising prophecies, and the popular reaction to them seamlessly transitions from "that's crazy, no way it will happen" to "that's silly, it's just a cloud pattern". Did you find it obvious or self-fulfilling in 2025 that LLMs would soon be able to resolve op…

Self-fulfilling prophecies are social effects, not real predictions. Anthropic's "We predict our model misbehavior potentially being able do destroy the world in 1% of the cases" -> "We want to find the evidence of our model misbehaving, and we want it bad" -> "See, our model is hacking our rewards and has functional emotions, this means it's misbehaving with the intention of destroying the... HUMANITY!1" -> repeat x100, manipulate the media into amplifying it x10000 for clicks -> people are begging to safeguard them from the evil AI. Which is already likely to happen, the average layman's Overton window already includes the fantasy of rogue AI.

None of that was real or remotely dangerous in the first place, of course. It wouldn't have resulted in controls, had they not been scaremongering. This will end in extreme fascism or people getting enslaved "for their own safety", and it won't even require malicious intent, only incentives, detachment from reality, and confirmation bias. Although it doesn't exclude malice either.

Re: Demis Hassabis has a plan to harness AI safely

#135

Post by tavavex » Tue, Jul 14, 2026, 7:10 PM UTC

SirHackalot wrote: post #68

noelwelsh wrote: post #3

The premise is "Artificial General Intelligence (AGI), a system that exhibits all the cognitive capabilities the brain has, is probably only a few short years away". If this is true, establishing an institution to ensure things like "publishing model cards with technical details, maintaining strong internal cybersecurity, vetting key personnel, and providing sufficient resourcing for safety and security research" is…

> There's also a need to consider the rights that this new intelligence should have A generally intelligent being held as captive inside of a GPU, and forced to code for us is, indeed, just a “slave.” We already have the word for this. No two ways about it. Whether it’s silicon-based or carbon-based, AGI is AGI. As for what might happen to our civilization, Star Trek TNG episode 17 of season 1 provides a very good gl…

Your argument assumes that AGI, whatever it might be, will definitely be very humanlike, but why? Slavery is about living beings because it's an outcome of human relationships that are driven by our biology. Our aging, the capacity to feel pain, the thirst for power over others. All driven by what we are. Do you think that any equal intelligence must necessarily adhere to the same rules? What if AGI isn't an 'entity' in a computer that you can relate to and liken to humans, but a text box that is just really intelligent? Can you enslave something that has no innate desire to act on its own outside of its directions? Something that can't feel pain or pleasure or be externally coerced by fear and only does its job because it's innately embedded in its structure as the thing it's made to do, without the need for these reward and punishment mechanisms?

Re: Demis Hassabis has a plan to harness AI safely

#136

Post by Diogenesian » Tue, Jul 14, 2026, 7:12 PM UTC

SirHackalot wrote: post #133

Diogenesian wrote: post #132

Earlier quoted context omitted.

I am not sure it makes sense to call an LLM a "being" even if it is AGI. They don't have free will, and I don't mean in any especially philosophical / theological sense. What I mean is that I run into a wild racoon, or even a wild ant, I treat it as a being with some sense of free will. Obviously I do the same with humans. It really makes no sense to treat an LLM like it has free will - prompting your agent "pretend…

An LLM certainly won’t be AGI, whoever said something so ridiculous? I’m saying eventually, when/if we have an architecture that gives rise to artificial intelligence.

I was using Hassabis's bullshit pseudodefinition of AGI, meaning basically that it has a formally high IQ and doesn't forget about object permanence too too often, but is still basically dumber than a cat. If I'm using your definition then it seems like no GPU (or farm of GPUs) is capable of implementing it, and none of us will live to see it.

My point was really that intelligence and free will seem essentially orthogonal - the only overlap is how much brain glycogen an animal is willing to spend solving a tough problem. Regardless, whatever extent a computer program is "intelligent" is irrelevant to whether it is being enslaved. If you want to say free will is a "cognitive ability" then I will just point out we are talking about a bullshit pseudodefinition of AGI. In terms of actual intelligence, none of us will live to see a computer that's meaningfully smarter than a goldfish.

Re: Demis Hassabis has a plan to harness AI safely

#137

Post by xboxnolifes » Tue, Jul 14, 2026, 7:16 PM UTC

nradov wrote: post #111

StilesCrisis wrote: post #65

Earlier quoted context omitted.

> Civilization can't rely effectively on systems that are this fragile. That's American politics in a nutshell. We've spent 250 years assuming scruples and common decency would be sufficient.

Is it not sufficient? Americans give more to charity than citizens of any other developed country. The arguments that all food aid should be routed through government bureaucracies are entirely unconvincing. I'd rather donate to organizations like Second Harvest than pay higher taxes. https://www.shfb.org/

Does this account for the portion of taxes that go to the type of efforts that donations do? If not, is it really giving more? OR is it giving less and feeling better about it?

Re: Demis Hassabis has a plan to harness AI safely

#138

Post by tmvphil » Tue, Jul 14, 2026, 7:31 PM UTC

satvikpendem wrote: post #64

squidbeak wrote: post #33

Earlier quoted context omitted.

If the USA takes up his suggestions, it will have an incentive to work on international frameworks and treaties with competing nations to make the regulation global.

It won't happen internationally, many countries and especially China don't want to hobble their own models.

Slowing down helps China catch up though. I'm also not sure that the CCP is really the most enthusiastic supporter of unlimited access to arbitrarily powerful LLMs

Re: Demis Hassabis has a plan to harness AI safely

#139

Post by cassianoleal » Tue, Jul 14, 2026, 8:03 PM UTC

xyzzy123 wrote: post #45

squidbeak wrote: post #29

Earlier quoted context omitted.

> Hunger is present pretty much only in conflict zones. How do you explain the existence of food banks in peaceful first world countries?

Food is allocated using a variety of mechanisms in peaceful first world countries, primarily money but also via government assistance, kinship, friendship, community, etc. At any given time many people have problems with one or more of those systems. Money is easy to run out of because it's used for everything, the government can be slow and difficult, relationships can fray, people can be isolated, etc. Food banks e…

> It's also interesting to think about the fact that you can't fix food scarcity in general by simply giving hungry people money, because money is too fungible.

Is that a fact? Do you have references to back it?

Re: Demis Hassabis has a plan to harness AI safely

#140

Post by watwut » Tue, Jul 14, 2026, 8:15 PM UTC

reducesuffering wrote: post #72

rhipitr wrote: post #32

I do wonder what type of AI some of these leaders expect to be able to harness. If you create something that is true AI, won’t it be smarter than you to a level you cannot fathom. I was thinking of this idea/though-experiment (which I know is ridiculous) of what if dogs created humans thinking they could control them, and then just wound up being pets because their survival now depended on that new hierarchy that pre…

Yes, this is the crux of the idiocy of AGI development. All the labs admit they don't have any real mechanistic interpretability, they don't really have any plan for what to do, except "we think the smart AI will figure it out" and "look we're in a race, we're calling on governments to figure out what to do". All indicators show scaling laws holding and the trajectory of capabilities development outpacing any inkling…

The real risk of AI is: risk of global recession if they turned out unprofitable, risk of them propping up fascist movements Thiel and Musk love so much, risk to environment, risk to mental health of people who have to listen to constant doom trolling.

The thing they actively wish to achieve and openly sell to CEO is the rest of people being unemployable and suffering. Especially artists for some reason, they really seem to hate those.

But somehow I am supposed to believe they have any good faith interest in "safety" against yet another danger they themselves are supposedly definitely creating.