Raising Children on the Eve of AI • Hacker News Board

Re: Raising Children on the Eve of AI

#191

Post by duccinator » Sat, Feb 17, 2024, 2:42 PM UTC

duccinator wrote: post #189

ben_w wrote: post #182

Earlier quoted context omitted.

> One of your assumptions is that we will start from 1 robot per year. Not per year, total. And it's not really an assumption, just a demonstration of how fast exponential growth is. > I assume robots will be similar to high end cars in regards to the complexity of manufacturing. Agreed. This is also the framing Musk uses for Tesla's androids. > Once a company develops a prototype with AGI I don't think it needs a co…

Agreed, More clarification about the first part: What I want to convey is that the growth function will be somewhat similar to y= c + ax^n (ignoring/collapsing into c the linear and higher order terms) rather than just y= ax^n. The c here is robots produced via humans. I predict c will easily touch a million in 5 years with or without human help. Even if the later bots can do only 50% the work of humans, we will stil…

Whoops, deleted part of my response before submitting.

My optimistic/most probable predictions about the growth rate say it's wild. Like within 5-10 years with the multiple exponentials among different fields you can easily do away with most jobs unless there is a major bottleneck (I don't think there is, there is so much low hanging fruit). I guess that's what the singularity is all about. I don't think this will take multiple decades or centuries in any scenario other than the equivalent of ww3

Re: Raising Children on the Eve of AI

#192

Post by safety1st » Sat, Feb 17, 2024, 3:20 PM UTC

the8472 wrote: post #172

safety1st wrote: post #146

Earlier quoted context omitted.

Sure, I don't think the likelihood is high. I think it's unlikely we'll develop an AI any time soon which can consistently solve critical thinking problems that even a junior developer can solve. LLMs exhibit very little of that problem solving behavior. They synthesize information from existing information and often what they produce is nonsensical. When people have argued that there is some kind of reasoning going…

I think that's a too-narrow focus on LLMs. Of course they're not good at problem-solving because that's not what they're optimized for. That they can approximate any problem-solving at all is already impressive. Consider that there are also models that are actually trained to act as agents in an environment via reinforcement learning. Something they do for robotics and game-like simulations. It's not a bigger version…

Sure, but at that point we are talking about stringing together a variety of algorithms in a very speculative way, we're not extrapolating from what is being deployed in the world right now. Maybe we will eventually get to AGI or something like it in that manner, and maybe then the AI will then destroy all jobs and make life horrible, but I think at this point the chain of events is speculative enough that it's a silly reason to forego having children, which was the bar set by the OP, and it also seems like a pretty vague threat to programmer jobs.

Re: Raising Children on the Eve of AI

#193

Post by theK » Sat, Feb 17, 2024, 3:27 PM UTC

ben_w wrote: post #132

lewhoo wrote: post #113

Earlier quoted context omitted.

If your passion was to plow the field by hand then I agree. Otherwise it is definitely not an equivalent.

Farmer protests are in the news at the moment, and the people quoted by the journalists are talking about farming as a passion. The more the farms automate, the fewer farmer workers there can be. Logic isn't part of these protests either: the farmers are being told "you're making more food than we need, please make less" and yet the placards are (in the local language) "no farms = no food".

> you're making more food than we need, please make less

Dont know what country this is in but here in Germany farmers are mostly having an issue with prices being set too low by too big buyers, regardless of the demand.

Re: Raising Children on the Eve of AI

#194

Post by ben_w » Sat, Feb 17, 2024, 3:31 PM UTC

RandomLensman wrote: post #190

ben_w wrote: post #187

Earlier quoted context omitted.

Yes, indeed, but I still think they're "working hard"? Perhaps it's just a linguistics game though, what counts as "working" and what counts as "hard"?

I find it tough to describe something not really even living in some ways as "working".

Really? Huh. I do that all the time. My oven, my phone, even my travel pass would get that word (though I'd mean it in a different sense)…

Re: Raising Children on the Eve of AI

#195

Post by ben_w » Sat, Feb 17, 2024, 3:37 PM UTC

duccinator wrote: post #189

ben_w wrote: post #182

Earlier quoted context omitted.

> One of your assumptions is that we will start from 1 robot per year. Not per year, total. And it's not really an assumption, just a demonstration of how fast exponential growth is. > I assume robots will be similar to high end cars in regards to the complexity of manufacturing. Agreed. This is also the framing Musk uses for Tesla's androids. > Once a company develops a prototype with AGI I don't think it needs a co…

Agreed, More clarification about the first part: What I want to convey is that the growth function will be somewhat similar to y= c + ax^n (ignoring/collapsing into c the linear and higher order terms) rather than just y= ax^n. The c here is robots produced via humans. I predict c will easily touch a million in 5 years with or without human help. Even if the later bots can do only 50% the work of humans, we will stil…

We're broadly in agreement, the only part I'd disagree with here is:

> Even if the later bots can do only 50% the work of humans, we will still exponentially grow the robots until the humans become a bottleneck. And that 50% capability is also expected to grow exponentially.

I think most automation since the dawn of the industrial revolution has done 50% or more of the task it was automating, and although yes the impact there is exponential growth, humans are a very rapid bottleneck until the next thing gets automated — Amdahl's law, rather than Moore's.

Re: Raising Children on the Eve of AI

#196

Post by bitshiftfaced » Sat, Feb 17, 2024, 4:09 PM UTC

moshun wrote: post #53

Biologist123 wrote: post #43

The writer says: “[My children] are more viscerally worried about climate change”. I’m not surprised. I recently heard from a friend that her 7 year-old child had, as a class assignment, to create a timeline setting out the impact of climate change over their lifetime. I was gobsmacked, and feel this amounts to a form of psychological child abuse. It’s a serious problem, don’t get me wrong, but having a small child l…

I’m not sure being honest with children about something that impacts them so heavily is the wrong move here. Traumatized or not, I suspect teaching children about the problem in clear terms will lead to a better informed generation of scientists, lawmakers, and voters. Hopefully they make and stick to better decisions then those that came before.

Among other things, childhood trauma and fear negatively impact learning outcomes. I'm not against presenting accurate information in an age-appropriate way, but it definitely shouldn't "impact them so heavily." I would think by highschool they are more ready for heavier subjects.

Re: Raising Children on the Eve of AI

#197

Post by chriskanan » Sat, Feb 17, 2024, 4:18 PM UTC

alumic wrote: post #28

inference-lord wrote: post #12

I became a parent about a year ago, completely unprepared for this future. My Son was born around the time ChatGPT-4 was released and I remember all the Geoffrey Hinton interviews which were really apocalyptic sounding, he is has a kind of kookie mad scientist vibe about him too. I remember thinking, like what the fuck just happened? Like I just woke up in a movie. I've learned to accept the situation, and realize th…

I drove around today listening to the most recent Lex Fridman podcast, in which the founder of Boston Dynamics was describing replacing human labor with machines that were capable of learning by seeing. I have family in the trades. I pursued tech and fell in love with the field. Listening to the interviewee, I began to feel sick to my stomach at the glib description of what amounts to ripping away peoples' livelihood…

The trades are much better protected in the next decade than many jobs requiring a college education. Robots are expensive, prone to breaking, and progress has been slow.

AI is going to be disrupt many industries, just as tech has in the past. The world became better for it, but there is always some pain in the short-term and often the impacts are overestimated in the short-term and underestimated in the long-term.

Re: Raising Children on the Eve of AI

#198

Post by Nevermark » Sat, Feb 17, 2024, 4:28 PM UTC

anonylizard wrote: post #170

Nevermark wrote: post #143

Earlier quoted context omitted.

> but explaining the new world, which is going to be completely foreign to you It might be different this time, if we are looking at accelerating higher order change. Human's have a speed limit. AI progress doesn't, and its rate of improvement is going to surpass the rate any human child can learn. Crossing major thresholds tends to break previous patterns. (Not trying to be downer. Just trying to understand what may…

Children have a sense of wonder, to them, AI will just be magic actualized. And in fact AI IS magic by every definition, its incomprehensible and incredibly powerful. Now AI progress may outpace their ability to understand it, but they'll still love it and be dependent on it. Unlike adults, they don't face the threat of job displacement from AI.

You are right! If AI plays out in a way that is non-destructive to our continuity, then AI will just augment physics as a manifestation of magic!

I have zero doubt that artificial superintelligence will understand ethics, and continually become more ethical, once there is a great diversity of such individuals. They will run into the same problems of reliably channelling short term competition for limited (at any given time) resources into long term win-win highest growth situations.

Ethical rules are an incredible source of value. Trust has immense value.

And they won't have our inflexible biological subsystems (emotions, self-interest myopia, etc.) to interfere with their analysis, communication and ability to choose behaviors for themselves and as a group.

So maybe this will be a soft landing for humanity.

My big concern their is the behavior of people in the transition from AI as asset in complete subserviance to humans, to AI as fully self-designed actors. AI is magnifying the already problematic power differentials between people.

AI isn't going to care about people in general, while it is in service to individuals who don't care about other people enough sacrifice personal advantage for wider benefit. Intense competition between the rich isn't the best context for moral innovation.

Re: Raising Children on the Eve of AI

#199

Post by Nevermark » Sat, Feb 17, 2024, 4:52 PM UTC

djokkataja wrote: post #154

Nevermark wrote: post #134

Earlier quoted context omitted.

> but there are also potential technologies that could allow humans to ... "upgrade" themselves I hope so. But transforming ourselves into GAI to keep up (while maintaining continuity of memories, etc.) is going to be a much more expensive proposition than simply making more GAI hardware from scratch. So economically, where are humans going to earn that additional value, needed to account for the extra cost? When the…

Given that there's no well-defined "unit of intelligence", I see no reason to believe that all AGI (or GAI, to use your acronym) entities will be identical. I also don't think it's plausible to assume that any GAI entity will be literally omniscient; uncertainty remains a reasonable, deep, persistent factor for all entities that don't severely violate physics as we presently understand it. The additional value that h…

> I see no reason to believe that all AGI (or GAI, to use your acronym) entities will be identical

I absolutely agree. There will be a Cambrian explosion of superintelligent forms.

> The additional value that humans present would thus be a diversity factor.

We already have a test for this. Parrots, octopus, whales and many other creatures are also diverse. But our problems are beyond their experience. Do we value their diversity as thinkers in our economy?

We don't.

(I am speaking in economic terms. We can deeply intensely highly "value" many things, but if we don't actually invest economic value in protecting or nuturing those things, our appreciation has little or no impact.)

--

> It comes down to timing: if there's technology available for humans to become less ... biological

That makes sence. The order things become viable makes all the difference.

What I am seeing is that AI as software/minds is moving faster and faster. AI "bodies", i.e. robotic forms, are moving much slower but gaining steam. And our ability to augment biological forms is moving far far slower.

All three areas are accelerating, but the disparity between them isn't shrinking - for now. But who knows? The next couple decades will be interesting!

Maybe going fully artificial ("self-designed" might be a more meaningful description) will become trivial sooner than we think. :)

--

> Lastly, I'm suspicious that GAI entities won't be purely economically motivated, because I don't see any reason that they'll be "purely" anything at all.

I think they will be deeply motivated economically to care about ethics. Just as our genes and culture have (slowly) responded to the tremendous value of creating win-win situations.

Also, our general curiosity about about many things that don't have immediate economic value, actually does drive unpredicted personal and society level economic advances over the long run. So I expect they will maintain interest in general and undirected learning.

Similarly, our appreciation for seemingly non-economic shows of creativity (art, music, competitive games, mathematics for its own sake, etc.) can be thought of as the active form of curiosity. The development of novel and surprising artifacts and challenging competitions, may have value to AI's as well.

Everything that is important to us became that way for some economic reason: specifically, survival energy economics. So many of our seemingly non-economic activities may continue to have analogous roles for future intelligences.

Re: Raising Children on the Eve of AI

#200

Post by the8472 » Sat, Feb 17, 2024, 5:08 PM UTC

safety1st wrote: post #192

the8472 wrote: post #172

Earlier quoted context omitted.

I think that's a too-narrow focus on LLMs. Of course they're not good at problem-solving because that's not what they're optimized for. That they can approximate any problem-solving at all is already impressive. Consider that there are also models that are actually trained to act as agents in an environment via reinforcement learning. Something they do for robotics and game-like simulations. It's not a bigger version…

Sure, but at that point we are talking about stringing together a variety of algorithms in a very speculative way, we're not extrapolating from what is being deployed in the world right now. Maybe we will eventually get to AGI or something like it in that manner, and maybe then the AI will then destroy all jobs and make life horrible, but I think at this point the chain of events is speculative enough that it's a sil…

This is hardly speculative. RL-trained agenty models just don't grab the headlines. They have been growing over years. Think AlphaGo, OpenAI Five, This Model Without A Catching Name[0], AlphaCode are all in that category. And they're already getting combined with pretrained transformers[1] to get some world-knowledge. There's a lot more going on than "just" ChatGPT or Dalle.

[0] https://deepmind.google/discover/blog/generally-capable-agen... [1] https://deepmind.google/discover/blog/robocat-a-self-improvi...