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

Moore's Law for Everything

moores.samaltman.com

Post reply on HN
Print view View original

241–250 of 554 posts

Re: Moore's Law for Everything

#241

Post by maxsilver »

S_A_P wrote: post #162
onlyrealcuzzo wrote: post #157

Earlier quoted context omitted.

Horses aren't the center of the economy, though. It's fundamentally different. Horses might always want more and more things no matter how much they get. But they don't create demand in our economy, so we'll never know. No matter how much humans get, they'll always want more. And other humans will have to come up with ways to meet that demand. Sure - if we ever reach a time where the entire human race says "this is e…

I think this is a good take- Having things be better for horses was never the point of the economy. In fact horses were probably the main driver for developing mechanical solutions since horse ownership is expensive and time consuming. There was no incentive to make life/jobs better for horses, and as soon as a better replacement was available horses were not necessary for most people. So you cant swap out horses for…

> So you cant swap out horses for humans as they analogy is completely flawed.

You kind of can though. The economy cares about no one, it cares only about capital. And as far as capital is concerned, humans are just fancy horses. Humans are expenses that need minimizing to the point of zeroing. In a perfectly efficient capitalist economy, no one pays any humans for anything ever.

Re: Moore's Law for Everything

#242

Post by maxsilver »

qeternity wrote: post #206
legulere wrote: post #85

Earlier quoted context omitted.

GDP per capita also counts children. $48,000 for a family of four sounds perfectly fine to me.

These arguments always ignore the fact that the only reason $X GDP is generated is because people are incentivized. Children don’t produce, so someone else is producing for them. If you tell people they will get $12k no matter how hard they work, they won’t work. We’ve tried socialism.

> If you tell people they will get $12k no matter how hard they work, they won’t work.

This is obviously not true, since in every capitalist society, the hardest working people already make the least amount of money, and the laziest people employed already get given the most amount of money.

Capitalism has already proven that financial incentive has no correlation to how hard someone works.

Re: Moore's Law for Everything

#243

Post by pgsimp »

colinmhayes wrote: post #142
zz865 wrote: post #81

I still dont really know about AI taking over the world. The most expensive things in my budget are housing, car, healthcare, childcare, flights/hotels, food. Does AI really change that much? There are definitely too many over-educated people out there already, I'd think this is more of an impact & setting up disappointment than the bots.

Robots building houses on cheap land due to mass work form home would make housing much cheaper. Robot doctors make healthcare much cheaper. Robot teachers make childcare/education much cheaper. Robot farmers + GMO make food much cheaper.

I don't think building costs are the main issue with availability of housing. You need the land to build on - and not just any land, but land in desired locations.

How are the artificial mega cities in China doing? Didn't they build several cities from scratch that are supposed to house several million people each?

Re: Moore's Law for Everything

#244

Post by joe_the_user »

joe_the_user wrote: post #78

In the next five years, computer programs that can think will read legal documents and give medical advice. Aside from the other points, taking "AI" as it exists in it's present form (deep neural networks and related) as specifically the bringer of unlimited wealth certainly puffs up the various "AI companies" notably OpenAI (It should be noted that OpenAI's most famous product, GPT-3, can generate strings that sound…

One other thing I'd like to add.

The scenario of AI mostly replacing people like doctors and lawyers involves bizarre paradoxes beyond whether deep learning "AI" works as advertised. Suppose you can train an "AI" to read legal papers or diagnose patients based on X-rays. That training is done from the data of real life lawyers and doctors actions. Suppose, best case scenario (very unrealistic imo btw), you have a complete "snapshot" of the behavior of lawyers and doctors in a given year. The problem is reality changes, you need new lawyering and doctoring behaviors after N years. Doctors need interpret new maladies, lawyers need to cite new decisions and both need to interpret new language forms that appear. But if you've actually removed the real lawyers and doctors, where would you get the new training data?

And this is just taking AI at it's word.

Re: Moore's Law for Everything

#245

Post by xiphias2 »

lincolnq wrote: post #236
xiphias2 wrote: post #50

I have an idea for sharing wealth: Why not create a non-profit organization that democratizes AI by sharing its models openly with everyone?

I know you are being sarcastic, but I don't know if everyone understands Sam Altman's motivation for running OpenAI the way it's being run. Sam is trying to create AI which causes the greatest benefit for humanity. That's openAI's mission. Open sourcing everything now would not achieve the mission. In case the "why" is not obvious: AI progress is limited by a) great research talent; b) money -- specifically being abl…

AI progress is not limited at all, it's the fastest moving research field in the world (5x improvement in efficiency / year for training a task to the same precision for the same cost if I recall correctly, far better than Moore's law).

OpenAI is opening the world to AI and helping people just like Google's doing ,,no evil'', Facebook is connecting people. At the point when an organization gets big enough to not keep its original values (being open for OpenAI), it's not better (less altruistic) in ,,making the world a better place'' than any other organization. Competition and having the power of AI distributed in more companies is good though (until they acquire each other).

Re: Moore's Law for Everything

#246

Post by qeternity »

maxsilver wrote: post #242
qeternity wrote: post #206

Earlier quoted context omitted.

These arguments always ignore the fact that the only reason $X GDP is generated is because people are incentivized. Children don’t produce, so someone else is producing for them. If you tell people they will get $12k no matter how hard they work, they won’t work. We’ve tried socialism.

> If you tell people they will get $12k no matter how hard they work, they won’t work. This is obviously not true, since in every capitalist society, the hardest working people already make the least amount of money, and the laziest people employed already get given the most amount of money. Capitalism has already proven that financial incentive has no correlation to how hard someone works.

What?? It’s the exact opposite. Those people HAVE to work that hard precisely for the reason I mentioned: they won’t be paid otherwise. This is precisely the issue. If you paid people irrespective of how much they work, they won’t work.

Your example is evidence of my statement, not refutation.

Re: Moore's Law for Everything

#247

Post by boring_twenties »

aksss wrote: post #72
bpodgursky wrote: post #61

Earlier quoted context omitted.

Well, a few decades ago, it took maybe $200 to save a life. We're certainly trending right.

That's fascinating.. so human lives are worth more, or there's more friction to intervention these days? Hoping it's the former. But curious what you think the explanation is for this. Just a reflection in standard of living, and so the cost to save has a higher standard?

I think it's more that the lowest hanging fruit have already been picked. In other words, all the lives that could be saved for $200 have already been saved. If I'm right about that it would seem to be an unambiguously good thing.

Re: Moore's Law for Everything

#248

Post by Viliam1234 »

stickfigure wrote: post #237
ralph84 wrote: post #7

> This revolution will generate enough wealth for everyone to have what they need We had enough wealth for everyone to have what we need a long time ago. No matter how much we have humans always want more. AI won't change that.

> We had enough wealth for everyone to have what we need a long time ago. That's obviously not true today: There isn't enough coronavirus vaccine to go around. There almost certainly will be enough eventually, but human beings live in the now. There will be another pandemic someday. Or some other natural disaster that creates localized or temporal scarcity. We can't just spin up a new lifesaving drug or a million new…

> There isn't enough coronavirus vaccine to go around.

That is mostly a question of regulations. The part that takes so much time is getting the vaccines approved; many researchers don't even try because they know they wouldn't have enough money to get their vaccine approved. Also, most governments negotiate hard to reduce the prices, despite the fact that economic damage from lockdowns is much greater.

https://de.wikipedia.org/wiki/Winfried_St%C3%B6cker

Re: Moore's Law for Everything

#249

Post by joe_the_user »

nindalf wrote: post #123

> We will discover new jobs–we always do after a technological revolution CGP grey put this well > Imagine a pair of horses in the early 1900s talking about technology. One worries all these new mechanical muscles (cars etc) will make horses unnecessary. The other horse reminds him that everything so far has made their lives easier - remember all that farm work? Remember running coast-to-coast delivering mail? Rememb…

I agree that the "new jobs are generated automatically" claim isn't supported by any good argument.

But the horse analogy isn't good way to show this imo. I hate to be "specie-ist" but human beings have a remarkable ability to learn new skills and attain new abilities. Horses don't. The human consumption of human's time, attention and so-forth is at the center of the economy. Human consumption of horse time isn't all that high a priority.

Human flexibility makes human activity something that can satisfy a vast number of human wants (and it means that there are a vast number of those wants to satisfy). That still doesn't mean you'll always have people employed but it makes employment a more plausible thing.

Re: Moore's Law for Everything

#250

Post by dzink »

Sam, you have a lot of great ideas and a lot of assumptions in this essay. Just like the parameters in deep learning models, we can’t know what will work in different scenarios until we can model it reliably. Time to build a worldwide game, version of the Sims if you will, to test different assumptions and global activity based on them. Happy to help on this too.

Some big assumptions: - Childcare can or cannot be handled by robots (very likely not if you need to raise healthy humans).

- Healthcare can or cannot be handled by robots.

- Humans will enjoy lack of employment without mental damage and how to retrain or provide for those needs.

In the game each player should be assigned a random individual with different roles in society - so they can see all angles (from different speeds of income accruing, to health, to time demands of their job, to responsibilities that need handled, etc). You will see all kinds of bugs that way - from individual liquidity crunches, to mental breakdowns, to industries that need more innovation/AI.

You can run any assumption in different epochs and get the answer by humans who play along worldwide.

That way when the rules are about to change due to an innovation, we can run a fun game simulation instead of running the risk of anarchy which kills a lot of people.

It is critically important to think about how we structure the future based on the technology we unlock. My pull request on this essay would be to propose we build a test suite for that first. Maybe we can use AI to simulate outcomes too after a good number of human runs.

Post reply on HN