Of course AI is extractive, everything is lately
coryd.dev
coryd.dev
Social media may be extractive if I give it my attention and data to resell. Crypto is extractive to the people buying into it and losing money. But I can avoid doing those things if I want.
AI these days takes anything I've written (from literature to code) or illustrated, and retroactively extracts value from it to be repackaged and redistributed. Even if I was proactive about securing the trademark or copyright, they can seemingly just ignore it without consequence, without me ever participating or opting in to it.
It feels massively different to have social media extract value from and image I uploaded willingly, vs AI extracting value (eg making things "in the style" of me) after training on my work against my will.
You create a highly useful website full of dense hard earned knowledge. Some people will just use that knowledge, but a few create their own sites, using what they learned.
A company sweeps up hard earned knowledge from millions of sites, with automation but not permission, and goes a long way to making those millions of informational sites redundant without any compensation.
Simplifying things obviously. But these two scenarios are not equivalent.
It's basically a store-brand argumentum ad verecundiam, casting yourself in the position of authority.
It should be obvious by now that there are many, many people who disagree with you without their arguments necessarily being "ridiculous".
I mean, I too would enjoy it if everyone agreed with all my opinions all the time, but I don't expect that to actually happen.
No, it absolutely is not.
> obviously absurd claim
You are obviously and absurdly wrong.
It's clear that nothing productive is going to come of this conversation.
(Based on a true story, it took me 3 years and 2 law firms to recuperate royalties for IP theft)
Where is the focus on the foundational principles of growth? The author writes that endless growth is impossible, but that is only true in the absence of novel breakthroughs. We need creations and inventions which bring us the future we once imagined: limitless energy and abundance.
How have we completely lost focus on new physics or foundational sciences and devoted the smartest minds of our time to transformers and data scraping? There is some critical human component which is completely absent here.
Where is the renaissance, and how do we make it happen?
All these things cost a lot of money, yet the majority of it gets eaten by systems designed to minimize the money actually going towards the research work. This isn't even really getting into how the researchers are paid.
I agree that AI researchers are concerningly blasé about what they're actually aiming to do. AGI is poorly defined, and the only negative externalities they seem to pay lip service to is culture war stuff. There's near zero consideration for the real negative externalities they're perpetrating upon the world. I think partly this also manifests in how they interact with fields where even older AI developments would be useful. At the physics lab I work at, I've found that many scientists have a negative impression of AI 'researchers', where they don't care to understand the problem they're trying to solve, preferring to just treat everything - even things where the solution space can be constrained by scientific understanding - as a black box. It's almost the opposite of science in approach.
If Leland Stanford could do it…
Grants suck. What matters is having smart people together in one place.
If you pay smart people enough money, and offer them enough insulation and protection from academic politics, they will come to your institute. Grant-making to people at institutions which are already politically and socially captured does not fix the root issue.
I agree that simply applying a fourier transform isn't meaningful research, but there are fields like fourier optics, where you're effectively just doing that, with the goal of modeling wave optics.
The kinds of issues I had in mind wrt poor uses of AI are things like inverse problem solving methods which train a model as a black box and neglect to involve physics informed feedback, so the results are of questionable value, or hyperparameter estimators which need an unrealistic amount of data or produce blatantly unrealistic estimates because they're naive and the AI "researcher" has no interest in actually understanding the physical logic to the parameters.
My research hasn't been with AI, just with simulations of certain physical systems, and one thing that was constantly drilled into me by the physicists was to be really careful about how far I stretch computational techniques, because ultimately none of this digital stuff means much if it becomes detached from reality. It can even cause actual damage because sometimes expensive purchases will be made on the basis of simulations.
I am still learning, but I have so far only applied inverse methods to systems where the physics is well-defined and the model already accepted as "real". For example, CT scanner algorithms which model the photon counters and attenuation through matter (although this can get complicated...), with the output being data in well-defined format (like the distribution of matter). I do see people using inverse solvers or AI to derive something which isn't an "image" but a model of the physics (does that make sense?). The extreme version of this which I have seen once before, is a paper using AI to extract algebraic forms of PDEs from videos of fluids. To what extent does the AI user have to understand/"pre-model" the physics, and to what extent does the AI generate understanding in some way? Apologies if this is misinformed as I am still outside the research community, and don't very actively read papers.
Currently I am trying to learn background to understand electromagnetic properties of materials with an interest to apply this to computer graphics. I eventually want to learn what processes are currently done for extracting material properties from scans, etc., along the lines of the classic gonioreflectometers. In this space, from the papers I look at, it looks like AI is unavoidable and I am sure an extremely useful tool but I definitely want to be careful not to get lost in misunderstanding.
The conclusion that I reached is that you need to understand at least enough to be able to handle basic scrutiny from people in the same field. So, for example, if you come up with an AI for deriving some model of the physics of a CT scanner, you should at least understand the problem you're tackling enough to make meaningful comparisons to any popular competing approaches that might exist. That understanding should automatically encompass things like understanding how your variables are bounded/how they behave.
A contrived example could be with an AI intended to guess 2 prime factors of a given number. Imagine someone making such a tool without even knowing what a prime number is, where the only benchmark they use for correctness is if its outputs fit the dataset of products of prime numbers given. The tool would be clearly useless, because while to the AI researcher it'd seem fine that it outputs 36, which is close to the intended 37, to anyone who knows about prime numbers it'd be an immediate warning sign needing further elaboration and explanation. The researcher could protest that as an AI expert, they don't know anything about prime numbers, but that doesn't really matter in research.
Similarly, due to my non-physics background, while I was (and to an extent still am) given significant leeway in terms of my ability to respond to detailed physics questions about my work, by now I am expected to be able to handle questions about things that are common knowledge among physicists in this field (so, for example, knowing what performing a Fourier transform means in our context, understanding the high level functioning of the key components of the machines we work with, having some intuition for when a simulation result is unrealistic etc). If I were to talk about how I've simulated some of these components, but drew a complete blank when asked how my work improved upon the common knowledge means of simulating them, I'd obviously look like a complete idiot :P
It may seem like a pretty low and obvious bar, but surprisingly I have seen many cases of researchers who didn't even care about reaching that level.
Universities absolutely encourage independent exploration.
Every year, they mint ~10x more grad students than there are research professor positions for them.
You want universities to produce even more of them?
https://cra.org/cra-wp/research-labs/
Just because Boeing sucks these days doesn't mean there isn't good research happening in industry.
Sure, they wanted to produce things that Ma Bell could monetize—but it was abundantly clear that they were not being micromanaged to that effect. Today, at least from the outside, it seems like any industry-funded research is cut off the moment it looks like it's not going to produce a clearly-profitable product or service within the next few years.
I can not disagree more with you on this.
The primary thesis of your original post appears to be universities' shift toward getting people ready for careers that require advanced knowledge (and away from pure learning/research).
That has very little to do with research freedom and the race to publications. Those are both real phenomena, but a) it is arguable whether the former is actually a problem, and b) they are only very loosely connected.
The number of grad students being trained, by contrast, is much more likely to be related to the degree to which universities still value and practice learning for learning's sake—although I think it is somewhat less so than vkou wished to imply, since there are, in fact, plenty of non-academic jobs that do require or prefer a Masters or even a PhD.
My premise is, purely and simply, that you are conflating multiple things that are not the same.
Change everything about the research grant system, and increase funding for it by 10x, and it'll address your concerns, and universities will happily make even more grad students. They aren't currently the bottleneck, because they aren't the sources of research funding - they are just a landlord (with reasonably okay terms) for the research lab that spends that funding.
A more advanced Hadoop will not change your life. It will not change the way you read, listen, watch, communicate (fundamentally). The Internet did.
Especially at a time when the "legacy" internet is falling apart, a new way to find information is rather a necessity than some breakthrough.
The "good" AI needs to fight it out with the "bad" AI before we see if the net positive is even there. A podcast can be translated by AI into a different language, you say? The price is super-realistic fraudulent phone and video calls from your "family".
If you look at the enablement of features, AI is a new (and, currently, often flaky) medium.
It would be analogous to looking at the Internet as limited to making it simpler to order from a catalog over the phone. While technically true, there is more that it can do besides.
Much like we don't use copper in our walls for dialup anymore, I don't anticipate the initial architectures of LLM and MoE to last terribly long, but they do enable proving that the concept works.
OP disagrees with GP. OP's main thesis is that AI enables a lot new applications. OP claims that GP is simply looking at it as if it were training data.
I stated that current AI techniques ARE indeed just reflections of the data used in training. I agree with GP that the current "AI"s are simply not transformative in the same way the Internet was.
If you change the training data for the current generation of AI, you get different behaviours. The training data forms a manifold - which you can think of as a landscape with features forming valleys and hills. What the current generation of AI does is that it tries to find a shape that fits the landscape - think of it like taking a very large sheet of cloth to cover a landscape. The stiffer the cloth, the less well the cloth fits to the landscape. The "stiffness" of the cloth is the amount of parameters that a neural network has. Modern deep nets are highly overparameterized - imagine a very soft pliable cloth - of course it fits to a landscape well.
So if you have a different training data - the neural network will fit to this different landscape as well. Hence the response will be different.
It's unfortunate that the training data is the entire internet for a few reasons:
1. Only the rich can train a vaguely competent AI. You're at the whims of those well-resourced enough. 2. There's no "alternate" training dataset anymore. (Though a clever thing people at OpenAI are doing are Mixture of Experts models, where you train multiple NNs using different subsets of the full training set, so you get multiple competencies)
Calling all AI LLMs is like calling all of the internet the web. Of course if I am mistaken, corrections are welcome.
...Because that's easily and widely understood to be what people mean in recent times when they're talking about "AI", referring to the stuff that's in the news, without further qualifiers.
If you want to talk about something more specific, you are going to need to be explicit about it, rather than expecting everyone else to understand what you've got in your head without actually saying it.
This is like saying "but "crypto" means so much more than just cryptocurrency! there's a whole cryptography field out there that does lots of good stuff!" It's true, but it's not helpful, because it's ignoring the obvious (at least to the other participants in the discussion) context. In this particular case, the context should be even more obvious because it's so clear that's what the article is talking about.
When people read our comments in 5 years, they will read "AI" and have a much broader topical take than the present excitement about LLMs.
Take computer vision for example - a "hello world" version of object recognition would use ImageNet, which is 14 million hand annotated images. Or Cifar10 which is 80 million images. That of course but sets the stage for training data differentiation. Google's image recognition algorithm is far superior to other search engines'. Why? Because of Google's data set.
Any Tom Dick and Harry can go create their own image recognition AI and train it based on all the public datasets (COCO, CIFAR, ImageNet) but that's considered pretty baseline nowadays. The differentiator is what _other_ datasets you have.
Different datasets yield different results. It doesn't matter the network. More data is better (usually).
The only real difference between then and now is that OpenAI's models are significantly better than my models from 2015, and they have that because well, they can afford to pile on more data. TBH, I never even considered using a large proportion of the whole internet as a training set as even remotely possible due to the sheer mind boggling costs.
Even now, to go through about 10% of The Pile would cost me way too much money.
THIS part of the AI sector is just a giant pyramid scheme - impress the investors so they shovel trillions your way. That's not exactly new in Silicon Valley - keep the valuation of a hot potato going up until someone is left holding the bag.
AI's most useful applications are not being a generalist.
This can be seen as a “shift left” in data capabilities. The rise of Data Products ultimately democratizes access to data.
I'm not smart enough to be a physicist, but I like listening to Eric Weinstein[1]. He thinks string theory is essentially a dead-end honeypot doing exactly what you describe with our smartest minds.
I like his podcast appearances — they are fun to listen to — but the solution to political machinations destroying established institutions is not to focus on the politics! We need to escape that frame entirely, and focus instead on building new institutions that are sufficiently reverent of smart minds and brilliant people.
Yes, we need new physics! We get there by escaping the current career trap which stops brilliant people from trying new approaches. Give the top, boldest, most daring researchers an alternative to tenure — $10m vested in a secure position at a new research institute. Then they won’t have to be scared of string theory boogeymen.
Many people think about the diretion this is going but tend to get dismissed as singularuty cranks if they think far enough ahead.
There is an article on the HN homepage about leaded petrol lowering IQ; I bet that in 100 years TikTok etc will be considered far more detrimental than that.
I will go so far as to say that millions of small businesses exist because social media provides them with global reach. As an example, if you forge, say, artisan chefs knives made out of meteorite, no one in your town/village/etc likely cares, but there are thousands of people worldwide that will gladly buy from you.
TikTok/Reels is simply the next evolution in information condensation. It's entertaining but also educative. I might not care to spend 20 minutes learning about the intricacies of, say, forging Damascus steel, but I'll gladly watch a couple TikToks on it.
Learning and understanding requires expending some kind of effort, and easy access to condensed information actually precludes you from experiencing that. If we don’t digest what we learn we are no better than LLMs — mindlessly regurgitating small bits of non-integrated information back and forth to each other.
When the alternative is not spending any time learning those topics at all, I'll take the 30 seconds.
I'm not going to spend minutes to hours learning something I'm barely interested in to begin with. As far as general knowledge goes, TikTok/Reels are great.
"At best dubious benefits" ??? I feel like there is a whole essay missing there, because my analysis of the situation is that it provides at minimum minor benefits to million of people using it on a daily basis, and at best it saves lives to some of them.
That it not to say there is 0 harm anywhere; we definitely need to do some cost/benefit analysis. But the initial premise is so obviously flawed that it's hard to consider the rest of the arguments.
Copilot saves me quite a bit of time writing documentation and boilerplate.
For hobbies AI is nice because I can ask my beginner questions and get results, where often you end up on low quality blogs or reddit threads arguing about things you shouldn’t bother thinking of when starting out.
Maybe someone somewhere asked GPT how to cook pasta and had no idea boiling water was hot, and GPT's warning not to get boiling water on themselves saved them from massive burns?
I mean, the phrase sounds like "I started my business with ChatGpt", which is true of all the recent trendy things and so would be less unique than people imagine or "ChatGpt cured my depression", where someone really shouldn't do that.
Not lately - everything is extractive and has always been. This is the very foundation of economic development and creation. If we were against all extractive things then nobody would work for a living and our entire society would crumble.
Modern development, for all its flaws, has brought immense progress in living standards and quality of life that shouldn't be romanticized away.
>Modern development, for all its flaws, has brought immense progress in living standards and quality of life that shouldn't be romanticized away.
Of course but it's imbalanced. The whole "infinite growth at any cost" thing combined with intense individualism is a recipe for solitude and mental health disaster. I'm not surprised populations are dropping.
Only 37% of people surveyed in the US believe AI has more benefits than drawbacks
AI is curiously well-received in countries with high levels of corruption
Everything constructive is just refining entropy into something.
>extractive industry can be defined as a processes that involve different activities that lead to the extraction of raw materials from the earth (such as oil, metals, mineral and aggregates), processing and utilization by consumers.
AI seems the opposite. It takes nothing from the earth and outputs vast quantities of iffy writing photos and similar.
> AI [is] costing a whole lot and providing, at best, dubious benefits
A wild take. Technology like GPT-4 being available today was unimaginable to us just 5 years ago. Our company saves hundreds of thousands of engineering-hours a month using generative AI. It has let me automate the humdrum boilerplate bullshit out of my workflows, both professional and personal.
It's also essentially replaced Google for me. Just today I've already used GPT-4 at least 5-10 times. Immensely useful, the $20/mo is a drop in the bucket for the value it provides to me.
> Ride-hailing services? Those flouted regulations, hooked users on cheap prices, drove down wages and made employment more precarious.
Has this guy ever stepped foot outside a major city? Has he ever gotten drunk and needed a ride home without friends in the immediate vicinity? Uber/Lyft provide immense value that simply wasn't there before.
Has he even ever spoken to any Uber drivers, or ex-taxi Uber drivers? Does he realize that people often prefer working for Uber as opposed to a "regular, stable job" and that customers actually like using it over taxis? Compared to taxis, Ubers tend to be cleaner, the drivers tend to be nicer, and it's just a easier, better experience.
The taxi cartel is even more powerful and "extractive" than Uber in many cities, and they exert undue market power, especially at airports and resorts, not Uber.
> streaming platforms that are actively rolling back up into something analogous to cable
Spotify still provides me with far more value/$ than having to buy every single song from iTunes or whatever. Apple TV is still quite useful to me and I far prefer it to ad-riddled cable with set showtimes for what you want to see.
Tale as old as time: people seeing the past with rose-tinted glasses so they can have a pessimistic "hot take" about today vs the good old days.
If you are tired of hearing the emperor has no clothes, getting mad at the crowd for saying it is not likely to help.
Except the emperor does have clothes, this pessimism simply isn't based in reality, it's based on a rose-tinted view of the past and a misunderstanding of why people use these services over the older alternatives.
The author seems to have such limited life experience that he can't even fathom why someone would prefer Uber over the taxis of old, or find great value in using generative AI every day, or prefer Spotify over something like the iTunes Store. The value proposition is so obvious and clear, and yet the author can't even see it.
Cabcharge and its associated monopolies in many cities is a blight that I am glad uber had a go at. "Flouted regulations" I hate when people associate words on paper, often well past their usefulness, with something physical or tangible or useful that can be damaged.
AI is doing so much to help a lot of people. Give it 5 years and hating on AI is going to go from mainstream to the most boomer of takes.
One thing is obviously immediately concerning though. It is unbelievable how fast ChatGPT has been normalized. Within a year the new people have come to expect something like ChatGPT to always be provided for them, for the rest of their life. Of course people are rushing in to define what that something is, and provide an AI service which people will use/misuse like the first page of Google results. I think almost certainly this new "first page of Google" component of people's thoughts will be abused. Do you think the first page of Google is good and formed by good incentives and empowerment of the user, considering how much of a cultural force it is (Googling something being just a normal part of your thought process)?
(I do have limited life experiences. Anecdotally, I know a high school teacher who has been teaching for almost 40 years. Over time students went from hiding their phone in the back of the class to scrolling through it in class freely, vaguely attempting increasingly simplified homework, and have now jumped on the new cool AI website to complete even basic tasks. This is extremely normal, why would a kid not use it?)
Yes. Google is an immensely useful everyday tool available to everyone for free. I think the negative externalities it causes are laughably trivial compared to the utility provided.
I would argue software is an infinite resource. Therefore, we can have infinite growth in a world of software.
If the rate of economic growth merely slows, we get a recession. What kind of hell happens if the rate goes negative and the economy shrinks? For too long? I don't feel obligated to point at any one person who says "yes I believe growth can continue infinitely forever" because it's the most basic assumption baked into economic models on how prosperity works. An assumption of infinite growth, and assumption that "progress" is also eternal, that there is always something more.
There is infinite distance between infinity and a negative number.
1. https://en.wikipedia.org/wiki/Self-replicating_spacecraft#Vo...
One can can instead use resources better to produce more valuable stuff (intensive growth).
The question is if there is a limit to the value of things we can produce with finite resources - and the answer is likely no - at least not in any kind of human timescale.
It's only confusing when seen from the bottom. The purpose of growth, seen from the perspective of the wealth-weighted-people served by capitalism, is to deliver a return on capital, to pay them for being rich in proportion to how rich they are. It's possible for everyone to win if the growth is real, but in a pinch fake redistributive growth will work (see: Cantillon pump). In either case, the expectation of growth rationalizes the outcome and the pulling of any policy levers needed to get there.
I think people are satisfied as long as stuff they dont like remains absent from their new sources.
I hate this saying.
Why is unending growth impossible? Nobody seems to every try to explain it.
Economic growth is generally improvement of society in some way. Usually it's about new inventions and ideas that spread more. Better ways to make things that people want, more efficient ways etc. if you come up with a more efficient fuel then that's still economic growth and the upside is that resource usage can decrease for the same benefit.
Math is essentially proven to be unending. There's always going to be something more to be discovered. Doesn't this mean that unending growth is possible?
Of course, in practice "unending" only means "for a very long time". We are nowhere near a point where we have exhausted new advancements.
Unless technological progress ends, we will continue to produce more and more valuable things. Even if the amount of money never changes or even if money disappears, growth still happens.