Google AI and DeepMind News and Discussions

Vakanai
Posts: 313
Joined: Thu Apr 28, 2022 10:23 pm

Re: Google DeepMind News and Discussions

Post by Vakanai »

Yuli Ban wrote: Tue Jul 05, 2022 2:22 am"Softly wresting control" is an idea I've had for several years now, best summed up as "if AI ever takes over, it'll be because we let it." We'll put it in charge of more and more executive and administrative tasks, relying on it for advice and suggestions until, before we know it, we've all but replaced society with AI. It'll be a runaway cycle of replacing human cognitive tasks with AGI driven by the same market forces that the most vehement luddites will purport to worship.
Ah - my bad, I misunderstood then. I actually do agree that we will put it in charge of more and more tasks of that sort, yes. The way you word it seemed more like it would seek to take control.
I still disagree with one thing though - even if we give it control of such things, it'll still be used to prop up politicians and leaders that we no longer need. Many of the people it could replace, won't be replaced, because they'll be put in charge of what it is or isn't allowed to do. They'll argue we still need them in power, and because they hold power they will remain in power. Not indefinitely, but for possibly many, many decades longer than they should. If China creates an AGI today, Xi Jinping will still remain in power in China until he chooses to retire or dies for whatever reason. Heck, whoever he chooses to replace him is likely to remain in control of China for decades after. Even if such leaders are no longer needed thanks to AGI, they'll remain still.

Same in capitalism. Even though AGI could render a future where it makes more sense to move on away from capitalism, it's not going away this century probably. Or at least, it's very possible that the billionaires prevent us from moving on.
Capitalism irrationally seeks efficiency and profit at all other costs. The moment it becomes more profitable to automate executive tasks, they WILL be automated. At which point the pin has been pulled and human-dominant society only has a limited amount of time left before the AI grenade explodes everything.
Yes and no I feel. I still think we'll be in a human-dominant society, since the AGI will be owned by humans using it to keep them in power in whatever way they prefer. Or at least, they will try to.

Just want to note that I'm not like many online who refuse to believe that this can bring positive change for the masses, just that I don't believe it will do so all by itself just by existing. I believe that we'll have to fight and work for a better future, one where AGI can bring us all these benefits that it'll make possible.
And I'd be inclined to believe that a soft AI takeover would be thwarted if automation progressed the way it was "supposed to," that is taking physical repetitive jobs, then higher skill jobs, then white collar and cognitive-heavy jobs before eventually automating art, entertainment, journalism, and creative work. But it's happening in a bizarro opposite direction, and that alone is going to undermine everyone and everything we know.
I think this might be where we differ a lot. I don't think it really matters much what order it takes away jobs in, because the problem lies in those who don't really have jobs anyways - the multibillionaire shareholder class who just owns companies and corporations. They don't really work anyways, AGI can't replace them, and those are the people we most need it to replace. How does AGI takeover from them? They're the ones who will own the AGI and decide how it is used.
The great historical irony is that if we had a socialist economy, this would all almost certainly be thwarted, prevented, or regulated so as to prevent workers from becoming redundant or inconvenienced until everything could be done as equally as possible. Yet the Soft AI Takeover will be called the result of socialism rather than natural capitalist forces.
I'm not sure on that - in a true socialist economy (different clearly from communism), it wouldn't be thwarted at all but pursued and embraced as a means to more effectively reach a better level of socialism. Lenin and Stalin would gladly pull the plug on AGI and have their men smash the servers to bits and have the programmers shot - but Karl Marx would applaud the rise of AGI as bringing his goal to reality.

I always wondered what the world might've been if we had one true socialist democracy in our history. Instead of now where all "socialist" states where under brutal dictatorships.
Skynet isn't happening.
Agreed. I was saying that it'd take a Skynet situation where AGI sought to take control from humans or that humans wouldn't have a say, not that I believed that it could happen.
Nero
Posts: 51
Joined: Sun Aug 15, 2021 5:17 pm

Re: Google DeepMind News and Discussions

Post by Nero »

Fundamentally your entire proposal is wrong because it assumes that there could be any coordinated effort at this time that would prevent a true AI from replacing 99% of the work force, things will not remain the way they are to create artificial power for certain politicians when AI is not something that can be contained to one political location.

True AI is global and you only need to have it exist in more than one location for the arguments of it to be contained to make any sense, with the amount of companies actively researching AI and the pace at which global politics moves there is no way that such an outcome is possible now. Once a single country is post-scarcity every country is because they are all potential customers even in a capitalist system the prefoliation of AI and the abundance created by it is fundamentally more valuable than maintaining a non post-scarcity economy.

Think of what video streaming is today, a business that completely destroyed all forms of physical media rental or purchase, Netflix is worth more $80 billion, Blockbuster at it's very most dominant was worth $5 billion. Even though it dominated the same marketspace to a much greater extent a small percentage of a post-scarcity economy is worth orders of magnitude more than a large percentage of one that is not post-scarcity.
User avatar
Yuli Ban
Posts: 4631
Joined: Sun May 16, 2021 4:44 pm

Re: Google DeepMind News and Discussions

Post by Yuli Ban »

I agree and agree-to-disagree with your criticisms except with one point: there IS a socialist democracy in the form of Nepal. It just isn't a total failure or a rousing success, so no one really talks about it.
And remember my friend, future events such as these will affect you in the future
Vakanai
Posts: 313
Joined: Thu Apr 28, 2022 10:23 pm

Re: Google DeepMind News and Discussions

Post by Vakanai »

Nero wrote: Wed Jul 06, 2022 9:21 amFundamentally your entire proposal is wrong because it assumes that there could be any coordinated effort at this time that would prevent a true AI from replacing 99% of the work force, things will not remain the way they are to create artificial power for certain politicians when AI is not something that can be contained to one political location.
I don't believe you're right about me being wrong (surprising). First "it assumes that there could be any coordinated effort" - that's your assumption, I never said anything about it being "coordinated". I'm imagining a less coordinated, more every man for himself greedy push of money scenario to exert control over how AGI plays out and is used. Second "that would prevent a true AI from replacing 99% of the work force" I never ever said it wouldn't - it WILL replace the workforce, and if anything your estimate of it replacing only 99% of the workforce is on the low end. I believe that it will both replace the workforce and yet still be hampered in it's ability to raise this replaced workforce out of poverty. Our biggest disagreement would be this part - "things will not remain the way they are to create artificial power for certain politicians". I believe that politicians being the sort of people inherently attracted to power, they will seek to maintain that power so it doesn't matter if "AI is not something that can be contained to one political location" because most political locations will seek to control it in such a way that they maintain power over their regions.
True AI is global and you only need to have it exist in more than one location for the arguments of it to be contained to make any sense, with the amount of companies actively researching AI and the pace at which global politics moves there is no way that such an outcome is possible now. Once a single country is post-scarcity every country is because they are all potential customers even in a capitalist system the prefoliation of AI and the abundance created by it is fundamentally more valuable than maintaining a non post-scarcity economy.
I disagree. If anything I think the global pace of politics being slow will prove a hindrance - they'll be too slow to enact certain laws and regulations over the companies and corporations who will find new ways to abuse this technology, and thus having lived with it for a few years will be more apt to listen to their wealthy backers that it "isn't so bad" and will form the laws, when they eventually get to it, to support these companies. By the way, I never made the argument that they'd try to slow down post-scarcity, only that they'd try to control AGI to keep them in power - there's more than one model of how post-scarcity society could be. We could be post-scarcity in a world where corporations and governments decide how these resources are used, impose artificial scarcity, and use AGI to maintain control over the population. Post-scarcity has nothing to do with holding back the full potential of AGI from making a better future. We can end up yet in a post-scarcity dystopia. It's shockingly possible.
Think of what video streaming is today, a business that completely destroyed all forms of physical media rental or purchase, Netflix is worth more $80 billion, Blockbuster at it's very most dominant was worth $5 billion. Even though it dominated the same marketspace to a much greater extent a small percentage of a post-scarcity economy is worth orders of magnitude more than a large percentage of one that is not post-scarcity.
Like I said, I don't know where you got it in my argument that they were going to hold back a post-scarcity future. No where did I ever suggest such a thing. I said that they would seek to limit its uses purely for those ways which they could use to increase profit and control for themselves, and it's entirely possible to do just that while ushering in a post-scarcity future.


There's this belief I feel, that all we need to do is reach AGI/ASI, and assuming that the alignment problem doesn't go pear-shaped, all we'll need to do is sit back and reap the wonders of tomorrow, and that there's nothing those in power can do to stop a brighter future. This isn't true, and is a dangerous assumption imo. AGI will come, and it will bring certain advantages and wonders for sure, but there are a truly frightening number of ways those people can shape it into a future we do not want, one where both a post-scarcity economy and a capitalist hellscape exist hand in hand. One where our future is taken from us and warped for many unbearable decades. I'm afraid that people think we only need to get to AGI and then our work is done. I believe that we shouldn't make that assumption, that when AGI makes this future a possibility, we will have to rise up and work to make it a reality. Post-scarcity isn't the endgame, a more equitable future is. Post-scarcity is more the starting line than the finishing line.
Vakanai
Posts: 313
Joined: Thu Apr 28, 2022 10:23 pm

Re: Google DeepMind News and Discussions

Post by Vakanai »

Yuli Ban wrote: Wed Jul 06, 2022 4:48 pm I agree and agree-to-disagree with your criticisms except with one point: there IS a socialist democracy in the form of Nepal. It just isn't a total failure or a rousing success, so no one really talks about it.
Which points do you agree with, and which do you agree to disagree with? If you don't mind me asking.

And I know so little of Nepal I'm afraid that I was under the mistaken assumption it was a city in Tibet and thus under China's government.
User avatar
Ozzie guy
Posts: 486
Joined: Sun May 16, 2021 4:40 pm

Re: Google DeepMind News and Discussions

Post by Ozzie guy »

I made a larger post here but I don't want to derail things.

TLDR Nepal is not socialist look at their private sector, unemployment and multinational companies exploiting their workers.

The Nepalese revolution had genuine communists but China made sure the leadership and country became capitalist after it's revolution.

I suppose you could compare it to the many Latin American countries that are not socialist but left wing by 2020 US standards and get labeled socialist for various reasons.
User avatar
wjfox
Site Admin
Posts: 8663
Joined: Sat May 15, 2021 6:09 pm
Location: London, UK
Contact:

Re: Google DeepMind News and Discussions

Post by wjfox »

I'm surprised nobody posted this breakthrough yet...

-----

AI visualises every protein known to science

29th July 2022

DeepMind's AlphaFold program has uncovered the structures of more than 200 million proteins, essentially all of those known to science.

[...]

Version 1 of AlphaFold emerged in 2018, when it achieved first place in the Critical Assessment of protein Structure Prediction (CASP), a competition run by the global scientific community every two years.

[...]

Version 2 won in 2020, this time with a much greater accuracy of 92.4 and a margin of error comparable to the width of an atom (0.16 nanometres). The result made headlines, as it meant that the 50-year "protein folding problem" – one of the biggest mysteries in biology – had finally been solved.

DeepMind launched an even more powerful version in 2021 with 16 times the calculating speed. It generated 3D structures for a million proteins, which the company made freely available on a new database. Until then, only 180,000 protein structures had existed in the public domain, so the new data increased that number by 5.6 times. AlphaFold itself then became open-sourced, allowing scientists around the world to benefit from its abilities.

Now, the AI has taken its achievements to a whole new level. Less than 12 months after its public release, AlphaFold has been accessed by more than half a million researchers, who together have generated structures for nearly all catalogued proteins known to science: 200 million of them.

Read more: https://www.futuretimeline.net/blog/202 ... iction.htm


Image
User avatar
Time_Traveller
Posts: 2023
Joined: Sun May 16, 2021 4:49 pm
Location: Clermont, Indiana, USA, October 7th 2019 B.C.E

Re: Google DeepMind News and Discussions

Post by Time_Traveller »

‘Existential catastrophe’ caused by AI is likely unavoidable, DeepMind researcher warns
4 hours ago

Researchers from the University of Oxford and Google’s artificial intelligence division DeepMind have claimed that there is a high probability of advanced forms of AI becoming “existentially dangerous to life on Earth”.

In a recent article in the peer-reviewed journal AI Magazine, the researchers warned that there would be “catastrophic consequences” if the development of certain AI agents continues.

Leading philosphers like Oxford University’s Nick Bostrom have previously spoken of the threat posed by advanced forms of artificial intelligence, though one of authors of the new paper claimed such warnings did not go far enough.

“Bostrom, [computer scientist Stuart] Russell, and others have argued that advanced AI poses a threat to humanity,” Michael Cohen wrote in a Twitter thread accompanying the article.

“Under the conditions we have identified, our conclusion is much stronger than that of any previous publication – an existential catastrophe is not just possible, but likely.”
https://www.independent.co.uk/tech/arti ... ml?src=rss
"We all have our time machines, don't we. Those that take us back are memories...And those that carry us forward, are dreams."

-H.G Wells.
User avatar
Yuli Ban
Posts: 4631
Joined: Sun May 16, 2021 4:44 pm

Re: Google DeepMind News and Discussions

Post by Yuli Ban »

wjfox wrote: Fri Jul 29, 2022 8:57 pm I'm surprised nobody posted this breakthrough yet...

-----

AI visualises every protein known to science

29th July 2022

DeepMind's AlphaFold program has uncovered the structures of more than 200 million proteins, essentially all of those known to science.

[...]

Version 1 of AlphaFold emerged in 2018, when it achieved first place in the Critical Assessment of protein Structure Prediction (CASP), a competition run by the global scientific community every two years.

[...]

Version 2 won in 2020, this time with a much greater accuracy of 92.4 and a margin of error comparable to the width of an atom (0.16 nanometres). The result made headlines, as it meant that the 50-year "protein folding problem" – one of the biggest mysteries in biology – had finally been solved.

DeepMind launched an even more powerful version in 2021 with 16 times the calculating speed. It generated 3D structures for a million proteins, which the company made freely available on a new database. Until then, only 180,000 protein structures had existed in the public domain, so the new data increased that number by 5.6 times. AlphaFold itself then became open-sourced, allowing scientists around the world to benefit from its abilities.

Now, the AI has taken its achievements to a whole new level. Less than 12 months after its public release, AlphaFold has been accessed by more than half a million researchers, who together have generated structures for nearly all catalogued proteins known to science: 200 million of them.

Read more: https://www.futuretimeline.net/blog/202 ... iction.htm


Image
How did I miss this? That's monumentally impresssive!
And remember my friend, future events such as these will affect you in the future
User avatar
Yuli Ban
Posts: 4631
Joined: Sun May 16, 2021 4:44 pm

Re: Google DeepMind News and Discussions

Post by Yuli Ban »

Sparrow: Building safer dialogue agents
In recent years, large language models (LLMs) have achieved success at a range of tasks such as question answering, summarisation, and dialogue. Dialogue is a particularly interesting task because it features flexible and interactive communication. However, dialogue agents powered by LLMs can express inaccurate or invented information, use discriminatory language, or encourage unsafe behaviour.

To create safer dialogue agents, we need to be able to learn from human feedback. Applying reinforcement learning based on input from research participants, we explore new methods for training dialogue agents that show promise for a safer system.

In our latest paper, we introduce Sparrow – a dialogue agent that’s useful and reduces the risk of unsafe and inappropriate answers. Our agent is designed to talk with a user, answer questions, and search the internet using Google when it’s helpful to look up evidence to inform its responses.
And remember my friend, future events such as these will affect you in the future
Post Reply