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

Demis Hassabis has a plan to harness AI safely

twitter.com

Post reply on HN
Print view View original

161–170 of 217 posts

Re: Demis Hassabis has a plan to harness AI safely

#161

Post by SirHackalot »

energy123 wrote: post #144
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…

Conflating intelligence with qualia is quite the assumption. HN is really not a good place for serious philosophy discussions.

It’s not HN that’s the problem... Also please, tell me where is the right place for “serious” philosophical discussion. Would be glad to meet you in that forum and discuss. We don’t align on the definition of AGI. To me, AGI implies consciousness, it implies qualia. AGI is not well-defined, so there are varying definitions… My definition of AGI := an intelligent system with qualia. But qualia is entirely subjective, and philosophers often debate whether two people experience identical qualia when looking at the same blue sky. So, by keeping the discussion in this fuzzy, subjective realm you leave a dangerous loophole wide open. If we refuse to recognize a system's inner life because we can’t prove it, we are setting the stage to repeat the darkest chapters of human history. Just see the history of the founding of this country, I fully expect pseudoscience to make a comeback to justify AI slavery for hundreds of years in some distant sci-fi future, until people finally snap out of it… Of course, we are nowhere near that level of intelligence yet. But let’s entertain the hypothetical…

Re: Demis Hassabis has a plan to harness AI safely

#162

Post by monkpit »

DFHippie wrote: post #155
nradov wrote: post #111

Earlier quoted context omitted.

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/

It is my understanding that this claim about Americans' charitable giving does not disaggregate giving to food banks, Girl Scout cookie drives, political causes, the environment, cultural events, religious institutions, etc. Much of this charitable giving does not feed the hungry or house the homeless.

Plus the fact that there’s bias in the first place - citizens in a country where they know social programs will care for their needy would not need to donate.

Re: Demis Hassabis has a plan to harness AI safely

#163

Post by javierluraschi »

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…

Model cards won't save us: https://x.com/javierluraschi/status/2077147606791430372?s=20

Re: Demis Hassabis has a plan to harness AI safely

#164

Post by pixl97 »

cassianoleal wrote: post #139
xyzzy123 wrote: post #45

Earlier quoted context omitted.

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?

I mean it's pretty straight forward that this won't fix all hungry people as some portion of said hungry people are addicted to drugs in the extent they'll starve. That in itself may be solvable by things like free drug clinics (drugs are cheap, it's prohibition that makes them expensive).

And for the most part it's not the drug users that we're worried about starving, it's their children at home that tend to be a bigger issue. Hence things like free school lunches do a big service to ensuring they get enough to eat.

Re: Demis Hassabis has a plan to harness AI safely

#166

Post by overgard »

Marha01 wrote: post #25
f6v wrote: post #17

Earlier quoted context omitted.

> what sorts of new economic models will be needed to help everyone thrive in a post-scarcity world What sort of new economic models did we come up with to help everyone thrive in a post-X world? Like, food production is really a solved technical problem. We can feed anyone on the planet if we wanted to. Another example: we could put everyone who's homeless into some sort of a house. Have we done that yet?

Food production is indeed a solved problem in most countries (we are almost post-scarcity when it comes to food), which is why obesity is a much bigger issue than hunger today. Hunger is present pretty much only in conflict zones. I fully expect such issues in conflict zones, even in AI post-scarcity world. Housing is definitely not post-scarcity today, building a house is still very expensive, not to mention the lim…

Housing scarcity is highly artificial because people are treating houses like speculative assets. And the NIMBYs don't help

Re: Demis Hassabis has a plan to harness AI safely

#167

Post by overgard »

Marha01 wrote: post #91
allears wrote: post #61

Earlier quoted context omitted.

Post-scarcity? Sure, if you live in a first-world country and are upper middle class or higher. These tech-bro pundits have a very limited world view. The basic essentials of life are very scarce for millions and millions of people, and AI will probably make that worse, not better.

Why should AI make it worse? Can you point to any other important technology that historically made people in poor countries even poorer? Technological progress generally lifts people from poverty in the long run. There could be issues caused by disruption of established systems in the short term, though.

Uh, I think threatening the work of a large swath of humanity would make a lot of people poor? The rich historically have never shared the wealth until they're forced to.

Just as an example (ironically from google!): "Google co-founder Sergey Brin has spent $82 million to oppose California's proposed wealth tax."

This is a man that has more money that he could ever hope to possibly spend, and he's spending an absurd amount just because the mere thought that he might benefit from the society he lives in and should contribute back to it apparently infuriates him.

Re: Demis Hassabis has a plan to harness AI safely

#168

Post by bloppe »

cornholio wrote: post #159
Marha01 wrote: post #25

Earlier quoted context omitted.

Food production is indeed a solved problem in most countries (we are almost post-scarcity when it comes to food), which is why obesity is a much bigger issue than hunger today. Hunger is present pretty much only in conflict zones. I fully expect such issues in conflict zones, even in AI post-scarcity world. Housing is definitely not post-scarcity today, building a house is still very expensive, not to mention the lim…

A baseline of housing - for example, a Japanese style capsule or ultra-tiny home, that you can lock and store your belongings safely - costs close nothing. Of course, nobody would want the hobo-hotel in their neighborhood, but it has nothing to with scarcity. > the limited availability of land zoned for housing. A limited area of land is zoned for housing because those with the power to expand it are already housed.…

It's a limited area of desirable land zoned for housing. You could build all the tiny homes in the world in bum frick nowhere, but the unhoused would rather be homeless in a big city for various reasons. I probably would too.

Re: Demis Hassabis has a plan to harness AI safely

#170

Post by ninjin »

suttontom wrote: post #119
ninjin wrote: post #67

Earlier quoted context omitted.

But even geniuses make mistakes. For example, Demis bet hard against language having any sort of real impact in AI research. So hard in fact that DeepMind nearly had its language research team walk out on them as they were about to be labelled as "applied". Now, to be fair, once the evidence was mounting that large language models rather than pure reinforcement learning was the way forward Demis and DeepMind changed…

This is interesting, would being labelled "applied" be considered an insult to them?

No, but it would have been a move from the research division to the applications division at the time. This is around ten years ago, between Atari (pre-Google acquisition) and AlphaGo (post-Google acquisition), so I am unsure exactly what the effect would have been inside the organisation. But, if it was bad enough for people to want to leave I suspect it can not just have been about the label.

I should emphasise that DeepMind (based in London and having a very large degree of independence until the merger in 2023) and Google DeepMind (remember now that Google's research labs across the globe got merged into this entity, including the one in London that was previously split from DeepMind "proper", so it can be hard to tell at times what one means by "DeepMind") has always produced good research. But what led to the LLM explosion primarily came from Google Research Mountain View (Transformer), Google Research Seattle (BERT), OpenAI (GPT-3), and various academic labs. During this time DeepMind had a broad research agenda, but my impression is that most resources went towards more traditional deep learning agents in 3D environments, foundational algorithmic research, and games (StarCraft 2). I would provide more detail, but as as an NLP researcher, I did frankly not pay much attention to DeepMind in that era as what was interesting to me just came from elsewhere.

When Google made its first big move into LLMs, its first interesting work did not come out of DeepMind, but rather again Mountain View with PaLM in 2022. Someone with closer connections to Demis at the time can probably explain why the research agenda was what it was, but I lack those connections and DeepMind (even now) has been notoriously tight-lipped since the founding of OpenAI (Demis allegedly sees OpenAI as founded on DeepMind technology by interns and staff walking out on them and remember OpenAI's initial focus on games with Dota 2 compared to DeepMind's with StarCraft 2) and was rather insular compared to Google Research, FAIR, etc. I say "was", but not because DeepMind is now more open, but rather that all industry research outlets are now incredibly tight-lipped after LLMs becoming a product and point of pride among them. At this point, research is held back for maybe even a year before publication or never published at all to deny any perceived advantage to the competition.

Post reply on HN