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

Will scaling work?

dwarkeshpatel.com

Post reply on HN
Print view View original

151–160 of 289 posts

Re: Will scaling work?

#151

Post by berniedurfee »

ctoth wrote: post #98
berniedurfee wrote: post #31

I think there’s a huge assumption here that more LLM will lead to AGI. Nothing I’ve seen or learned about LLMs leads me to believe that LLMs are in fact a pathway to AGI. LLMs trained on more data with more efficient algorithms will make for more interesting tools built with LLMs, but I don’t see this technology as a foundation for AGI. LLMs don’t “reason” in any sense of the word that I understand and I think the ab…

> I think there’s a huge assumption here that more LLM will lead to AGI. I'm not sure you realize this, but that is literally what this article was written to explore! I feel like you just autocompleted what you believe about large language models in this thread, rather than engaging with the article. Engagement might look like "I hold the skeptic position because of X, Y, and Z, but I see that the other position has…

Lol, yes, in fact, I was reacting to the article.

The point I was trying to make is that I think better LLMs won’t lead to AGI. The article focused on the mechanics and technology, but I feel that’s missing the point.

The point being, AGI is not going to be a direct outcome of LLM development, regardless of the efficiency or volume of data.

Re: Will scaling work?

#152

Post by ketzo »

herval wrote: post #90
slibhb wrote: post #49

The best analogy for LLMs (up to and including AGI) is the internet + google search. Imagine explaining the internet/google to someone in 1950. That person might say "Oh my god, everything will change! Instantaneous, cheap communication! The world's information available at light speed! Science will accelerate, productivity will explode!" And yet, 70 years later, things have certainly changed, but we're living in the…

> And yet, 70 years later, things have certainly changed, but we're living in the same world with the same general patterns and limitations. With LLMs I expect something similar. Not a singularity, just a new, better tool that, yes, changes things, increases productivity, but leaves human societies more or less the same. by what criteria do you see the world as the same today vs 70 years ago?

I mean, very broad strokes, but I can see GP’s point.

- people eat plants and animals

- people pay money for goods and services

- there are countries, sometimes they fight, sometimes they work together

- men and women come together to create children, and often raise those children together

etc, etc, etc

The “bones” of what make up a capital-S Society are pretty much the same. None of these things had to stay the same, but they have so far.

Re: Will scaling work?

#153

Post by TaylorAlexander »

HarHarVeryFunny wrote: post #96
bee_rider wrote: post #60

Earlier quoted context omitted.

The internet did change things pretty dramatically. Productivity at information communication tasks just isn’t the entire economy. I think we are massively more productive. Some of the biggest new companies are ad companies (Google, Facebook), or spend a ton of their time designing devices that can’t be modified by their users (Apple, Microsoft). Even old fashioned companies like tractor and train companies have time…

> The internet did change things pretty dramatically. For sure - I grew up in the mid-late 70s having to walk to the library to research stuff for homework, parents having to use the yellow-pages to find things, etc. Maybe smartphones are more of a game changer than desk-bound internet though - a global communication device in your pocket that'll give you driving directions, etc, etc. BUT ... does the world really FE…

Considering that I work in open source robotics I literally couldn’t do my job without the internet. So that feels pretty different!

Re: Will scaling work?

#154

Post by hibikir »

HarHarVeryFunny wrote: post #96
bee_rider wrote: post #60

Earlier quoted context omitted.

The internet did change things pretty dramatically. Productivity at information communication tasks just isn’t the entire economy. I think we are massively more productive. Some of the biggest new companies are ad companies (Google, Facebook), or spend a ton of their time designing devices that can’t be modified by their users (Apple, Microsoft). Even old fashioned companies like tractor and train companies have time…

> The internet did change things pretty dramatically. For sure - I grew up in the mid-late 70s having to walk to the library to research stuff for homework, parents having to use the yellow-pages to find things, etc. Maybe smartphones are more of a game changer than desk-bound internet though - a global communication device in your pocket that'll give you driving directions, etc, etc. BUT ... does the world really FE…

For me, it's incredibly different. I moved to the US from Spain back when the best internet we could get at home was 3kb/sec, and we liked it (yes kids, close to a million times slower than today). I recall the massive cultural and economic detachment of that move: Minimal shared culture. Major differences in food availability: Often I couldn't even cook what I wanted if I didn't smuggle the ingredients. Connecting with people with shared interests was really difficult, as discovering communities was a lot of work: Even more so in America, where I needed a car for everything, and communities lacked the local gossiping infrastructure that I relied on at home.

Today, I got to do some miniature painting while hanging out on video with someone in England. I get to buy books digitally the same day they are published, and I don't have to travel a suitcase full of them, plus a cd collection for a 1 month vacation. My son can talk to his grandma, on video, whenever he likes: Too cheap to meter. Food? I can find an importer that already has what I want most of the time, and if not, i can get anything shipped, from anywhere. A boardgame from germany, along with some cookies? Trivial. Spanish TV, including soccer games, which before were impossible. My hometown's newspaper, along with one from Madrid, and a few international ones.

An immigrant in the 90s basically left their culture behind with no recourse. Today I can be American, and a Spaniard, at the same time with minimal loss of context by being away. All while working on a product used by hundreds of millions of people, every day, with a team that spans 16 timezones, yet manages to have standups.

A lot of people's lives haven't changed that much, because their day to day is still very local. If you work at the oil field, and then go to the local high school to watch your kid's game on friday night, and all your family is local, a big part of your life wouldn't have been so different in the 90s, or even in the 60s. But I look at the things my family did week that I couldn't have possibly done in 98, and it's most of my life. My dad's brain would have melted if he could hear a description of the things I get to do today that were just sci-fi when he died. It's just that the future involved fewer people wielding katanas in the metaverse than our teenage selves might have liked.

Re: Will scaling work?

#155

Post by berniedurfee »

lkbm wrote: post #51
berniedurfee wrote: post #31

I think there’s a huge assumption here that more LLM will lead to AGI. Nothing I’ve seen or learned about LLMs leads me to believe that LLMs are in fact a pathway to AGI. LLMs trained on more data with more efficient algorithms will make for more interesting tools built with LLMs, but I don’t see this technology as a foundation for AGI. LLMs don’t “reason” in any sense of the word that I understand and I think the ab…

I guess it's an "assumption", but it's an assumption that's directly challenged in the article: > But of course we don’t actually care directly about performance on next-token prediction. The models already have humans beat on this loss function. We want to find out whether these scaling curves on next-token prediction actually correspond to true progress towards generality. And: > Why is it impressive that a model t…

The conclusion that AGI will happen in 2040 is what I’m arguing against. I think 4020 is maybe a better estimate.

I don’t feel like we’re anywhere close given that we can’t even yet meaningfully define reasoning or consciousness… or as another commenter put it, what is it that differentiates us so significantly from other animals.

Re: Will scaling work?

#156

Post by dullcrisp »

ctoth wrote: post #98
berniedurfee wrote: post #31

I think there’s a huge assumption here that more LLM will lead to AGI. Nothing I’ve seen or learned about LLMs leads me to believe that LLMs are in fact a pathway to AGI. LLMs trained on more data with more efficient algorithms will make for more interesting tools built with LLMs, but I don’t see this technology as a foundation for AGI. LLMs don’t “reason” in any sense of the word that I understand and I think the ab…

> I think there’s a huge assumption here that more LLM will lead to AGI. I'm not sure you realize this, but that is literally what this article was written to explore! I feel like you just autocompleted what you believe about large language models in this thread, rather than engaging with the article. Engagement might look like "I hold the skeptic position because of X, Y, and Z, but I see that the other position has…

Why does it matter?

Re: Will scaling work?

#157

Post by throwup238 »

slibhb wrote: post #49

The best analogy for LLMs (up to and including AGI) is the internet + google search. Imagine explaining the internet/google to someone in 1950. That person might say "Oh my god, everything will change! Instantaneous, cheap communication! The world's information available at light speed! Science will accelerate, productivity will explode!" And yet, 70 years later, things have certainly changed, but we're living in the…

By that standard, nothing has meaningfully changed since agriculture and domesticated animals. We're still killing each other, forming hierarchical societies, passing down stories, eating, drinking, sleeping, and making families - except now we're killing each other from afar with gunpowder, forming those hierarchies using the guise of democracy or whatever, passing down stories in print rather than speech, can use condoms to control when we make families, and so on.

Human civilization has accumulated many layers of systems since then and the internet changed all of them to the point that many are barely recognizable. Just ask someone who's been in prison since before the internet was a thing - there are plenty of them! They have extreme difficulty adapting to the outside world after they've been gone for forty or fifty years.

Re: Will scaling work?

#158

Post by corethree »

HarHarVeryFunny wrote: post #96
bee_rider wrote: post #60

Earlier quoted context omitted.

The internet did change things pretty dramatically. Productivity at information communication tasks just isn’t the entire economy. I think we are massively more productive. Some of the biggest new companies are ad companies (Google, Facebook), or spend a ton of their time designing devices that can’t be modified by their users (Apple, Microsoft). Even old fashioned companies like tractor and train companies have time…

> The internet did change things pretty dramatically. For sure - I grew up in the mid-late 70s having to walk to the library to research stuff for homework, parents having to use the yellow-pages to find things, etc. Maybe smartphones are more of a game changer than desk-bound internet though - a global communication device in your pocket that'll give you driving directions, etc, etc. BUT ... does the world really FE…

It's because the change happened slowly. So it feels like nothing has changed.

Another thing that's changed is engineering. The US has moved up the stack. Engineering is now mostly software development and within that it's mostly web development. Engineering and manufacturing has largely moved overseas to Asia and that's where most of the expertise lies. The only thing off the top of my head that the US still dominates in engineering is software/aerospace/defense. In general though everything else is dominated by Asia, if you want the top hardware technology the US is no longer the place to get it. In Silicon Valley there used to be a good mix of different types of engineers, now everyone is SWE, and most likely doing web stuff. But here's the thing, you most likely wouldn't have noticed this unless you thought hard about it because either you're too young or because the change happened so slowly.

The same will be for AGI if it comes into fruition. A lot of jobs will be replaced, slowly. Then when AGI replacement reaches saturation most people will be used to the status quo whether it's better or worse. It will seem like nothing has changed.

Re: Will scaling work?

#159

Post by ctoth »

berniedurfee wrote: post #151
ctoth wrote: post #98

Earlier quoted context omitted.

> I think there’s a huge assumption here that more LLM will lead to AGI. I'm not sure you realize this, but that is literally what this article was written to explore! I feel like you just autocompleted what you believe about large language models in this thread, rather than engaging with the article. Engagement might look like "I hold the skeptic position because of X, Y, and Z, but I see that the other position has…

Lol, yes, in fact, I was reacting to the article. The point I was trying to make is that I think better LLMs won’t lead to AGI. The article focused on the mechanics and technology, but I feel that’s missing the point. The point being, AGI is not going to be a direct outcome of LLM development, regardless of the efficiency or volume of data.

I can interpret this in a couple different ways, and I want to make sure I am engaging with what you said, and not with what I thought you said.

> I think better LLMs won’t lead to AGI.

Does this mean you believe that the Transformer architecture won't be an eventual part of AGI? (possibly true, though I wouldn't bet on it)

Does this mean that you see no path for GPT-4 to become an AGI if we just leave it alone sitting on its server? I could certainly agree with that.

Does this mean that something like large language models will not be used for their ability to model the world, or plan, or even just complete patterns as does our own System one in an eventual AGI architecture? I would have a lot more trouble agreeing with that.

In general, it seems like these sequence modelers that actually work right is a big primitive we didn't have in 2016 and they certainly seem to me as an important step. Something that will carry us far past human-level, whatever that means for textual tasks.

To bring it back to the article, probably pure scale isn't quite the secret sauce, but it's a good 80-90% and the rest will come from the increased interest, the shear number of human-level intelligences now working on this problem.

Too bad we haven't scaled safety nearly as fast though!

Re: Will scaling work?

#160

Post by majkinetor »

ketzo wrote: post #152
herval wrote: post #90

Earlier quoted context omitted.

> And yet, 70 years later, things have certainly changed, but we're living in the same world with the same general patterns and limitations. With LLMs I expect something similar. Not a singularity, just a new, better tool that, yes, changes things, increases productivity, but leaves human societies more or less the same. by what criteria do you see the world as the same today vs 70 years ago?

I mean, very broad strokes, but I can see GP’s point. - people eat plants and animals - people pay money for goods and services - there are countries, sometimes they fight, sometimes they work together - men and women come together to create children, and often raise those children together etc, etc, etc The “bones” of what make up a capital-S Society are pretty much the same. None of these things had to stay the sam…

VERY broad strokes. We also still have a Sun, and the stars.

Internet and the last 30 years tech did change things dramatically. I bet that most people would feel handicapped if they were teleported just 50 years back. We got into this type of life progressively, so people didn't notice the change, even though it was dramatic. The same phenomena with gradient changes happen on physiological level too, this is not different.

Post reply on HN