Brian Eno: “The biggest problem about AI is not intrinsic to AI. It’s to do with the fact that it’s owned by the same few people”

Eh I’m fine with the illegal harvesting of data. It forces the courts to revisit the question of what copyright really is and hopefully erodes the stranglehold that copyright has on modern society.

Let the companies fight each other over whether it’s okay to pirate every video on YouTube. I’m waiting.

Oh, and it also hallucinates.

Oh, and people believe the hallucinations.

They're not illegally harvesting anything. Copyright law is all about distribution. As much as everyone loves to think that when you copy something without permission you're breaking the law the truth is that you're not. It's only when you distribute said copy that you're breaking the law (aka violating copyright).

All those old school notices (e.g. "FBI Warning") are 100% bullshit. Same for the warning the NFL spits out before games. You absolutely can record it! You just can't share it (or show it to more than a handful of people but that's a different set of laws regarding broadcasting).

I download AI (image generation) models all the time. They range in size from 2GB to 12GB. You cannot fit the petabytes of data they used to train the model into that space. No compression algorithm is that good.

The same is true for LLM, RVC (audio models) and similar models/checkpoints. I mean, think about it: If AI is illegally distributing millions of copyrighted works to end users they'd have to be including it all in those files somehow.

Instead of thinking of an AI model like a collection of copyrighted works think of it more like a rough sketch of a mashup of copyrighted works. Like if you asked a person to make a Godzilla-themed My Little Pony and what you got was that person's interpretation of what Godzilla combined with MLP would look like. Every artist would draw it differently. Every author would describe it differently. Every voice actor would voice it differently.

Those differences are the equivalent of the random seed provided to AI models. If you throw something at a random number generator enough times you could--in theory--get the works of Shakespeare. Especially if you ask it to write something just like Shakespeare. However, that doesn't meant the AI model literally copied his works. It's just doing it's best guess (it's literally guessing! That's how work!).

Oh, and it also hallucinates.

This is arguably a feature depending on how you use it. I'm absolutely not an AI acolyte. It's highly problematic in every step. Resource usage. Training using illegally obtained information. This wouldn't necessarily be an issue if people who aren't tech broligarchs weren't routinely getting their lives destroyed for this, and if the people creating the material being used for training also weren't being fucked....just capitalism things I guess. Attempts by capitalists to cut workers out of the cost/profit equation.

If you're using AI to make music, images or video... you're depending on those hallucinations.
I run a Stable Diffusion model on my laptop. It's kinda neat. I don't make things for a profit, and now that I've played with it a bit I'll likely delete it soon. I think there's room for people to locally host their own models, preferably trained with legally acquired data, to be used as a tool to assist with the creative process. The current monetisation model for AI is fuckin criminal....

I see the "AI is using up massive amounts of water" being proclaimed everywhere lately, however I do not understand it, do you have a source?

My understanding is this probably stems from people misunderstanding data center cooling systems. Most of these systems are closed loop so everything will be reused. It makes no sense to "burn off" water for cooling.

We spend energy on the most useless shit why are people suddenly using it as an argument against AI? You ever saw someone complaining about pixar wasting energies to render their movies? Or 3D studios to render TV ads?

It varies massivelly depending on the ML.

For example things like voice generation or object recognition can absolutelly be done with entirelly legit training datasets - literally pay a bunch of people to read some texts and you can train a voice generation engine with it and the work in object recognition is mainly tagging what's in the images on top of a ton of easilly made images of things - a researcher can literally go around taking photos to make their dataset.

Image generation, on the other hand, not so much - you can only go so far with just plain photos a researcher can just go around and take on the street and they tend to relly a lot on artistic work of people who have never authorized the use of their work to train them, and LLMs clearly cannot be do without scrapping billions of pieces of actual work from billions of people.

Of course, what we tend to talk about here when we say "AI" is LLMs, which are IMHO the worst of the bunch.

In a Venn Diagram, I think your “illegally harvesting” complaint is a circle fully inside the “owned by the same few people” circle. AI could have been an open, community-driven endeavor, but now it’s just mega-rich corporations stealing from everyone else. I guess that’s true of literally everything, not just AI, but you get my point.

Well, the harvesting isn’t illegal (yet), and I think it probably shouldn’t be.

It’s scraping, and it’s hard to make that part illegal without collateral damage.

But that doesn’t mean we should do nothing about these AI fuckers.

In the words of Cory Doctorow:

Web-scraping is good, actually.

Scraping against the wishes of the scraped is good, actually.

Scraping when the scrapee suffers as a result of your scraping is good, actually.

Scraping to train machine-learning models is good, actually.

Scraping to violate the public’s privacy is bad, actually.

Scraping to alienate creative workers’ labor is bad, actually.

We absolutely can have the benefits of scraping without letting AI companies destroy our jobs and our privacy. We just have to stop letting them define the debate.

And also it's using machines to catch up to living creation and evolution, badly.

A but similar to how Soviet system was trying to catch up to in no way virtuous, but living and vibrant Western societies.

That's expensive, and that's bad, and that's inefficient. The only subjective advantage is that power is all it requires.

I don't care much about them harvesting all that data, what I do care about is that despite essentially feeding all human knowledge into LLMs they are still basically useless.

Dead Internet theory has never been a bigger threat. I believe that’s the number one danger - endless quantities of advertising and spam shoved down our throats from every possible direction.

Could also put up:

• Massive collections of people are exploited in order to train various AI systems.
• Machine learning apps that create text or images from prompts are supposed to be supplementary but businesses are actively trying to replace their workers with this software.
• Machine learning image generation currently has diminishing returns for training as we pump exponentially more content into them.
• Machine learning text and image generated content self-poisons their generater's sample pool, greatly diminishing the ability for these systems to learn from real world content.

There's actually a much longer list if we expand to talking about other AI systems, like the robot systems we're currently training to use in automatic warfare. There's also the angle of these image and text generation systems being used for political manipulation and scams. There's alot of terrible problems created from this tech.

Power usage

I'm generally a huge eco guy but on power usage particularly I view this largely as a government failure. We have had to incredible energy resources that the government has chosen not to implement or effectively dismantled.

It reminds me a lot of how Recycling has been pushed so hard into the general public instead of and government laws on plastic usage and waste disposal.

It's always easier to wave your hands and blame "society" than the is to hold the actual wealthy and powerful accountable.

Power usage probably won't be a major issue; the main take-home message of the Deepseek brouhaha is that training and inference can be much more efficiently than we had thought (our estimates had been based on well-funded Western companies that didn't have to bother with optimization).

AI spam is an annoyance, but it's not really AI-specific but the continuation of a trend; the Internet was already drowning in human-created slop before LLMs came along. At some point, we will probably all have to rely on AI tools to filter it out. This isn't something that can be unwound, any more than you can undo computers being able to play chess well.

Here we are using recycled bags, banning straws, putting explosive refrigerant in fridges and using led lights in everything, all in the name of the environment, while at the same time in some datacenter they are burning kwh’s by the bucket loads generating pictures of cats in space suits.

That's, #1, fashion and not about environment, #2, fashion promoted because it's cheaper for the industry.

And yes, power saved somewhere will just be spent elsewhere. Cheaper. Cause that means reduced demand for power (or grown not as fast as otherwise).

Here we are using recycled bags, banning straws, putting explosive refrigerant in fridges and using led lights in everything

lol, sucker. none of that does shit and industry was already destroying the planet just fine before ai came along.

My biggest gripe with AI is the same problem I have with anything crypto crypto: It's out of control power consumption relative to the problem it solves or purpose it serves.

Don't thrown all crypto under the bus. Only bitcoin and other proof of work protocols are power hungry. 2nd and 3rd generation crypto use mostly proof of stake and ZKrollups for security. Much more energy efficient.

I mean, it's our work the result should belong to the people.

This is where "universal basic income" comes into play

That's what all artists have done since the dawn of ages.

Techno-Feudalism

I'll say it, yet again. It's just feudalism. "Techno-Feudalism" has nothing different enough to it to differentiate it as even a sub-type of feudalism. It's just the same thing all over again, using technological advances to improve the ability to monitor and impose control over the populace. Historical feudalists also leveraged technology to cement their rule (plate armor, cavalry, crossbows, cannon, mills, control of literacy, etc).

So long as there are big players releasing open weights models, which is true for the foreseeable future, I don't think this is a big problem. Once those weights are released, they're free forever, and anyone can fine-tune based on them, or use them to bootstrap new models by distillation or synthetic RL data generation.

Welcome to every technological advancement ever applied to the workforce

The system in place is “open weights” models. These AI companies don’t have a huge head start on the publicly available software, and if the value is there for a corporation, most any savvy solo engineer can slap together something similar.

most LLM's for chat, pictures and clips are magical and amazing. For about 4 - 8 hours of fiddling then they lose all entertainment value.

As for practical use, the things can't do math so they're useless at work. I write better Emails on my own so I can't imagine being so lazy and socially inept that I need help writing an email asking for tech support or outlining an audit report. Sometimes the web summaries save me from clicking a result, but I usually do anyway because the things are so prone to very convincing halucinations, so yeah, utterly useless in their current state.

I usually get some angsty reply when I say this by some techbro-AI-cultist-singularity-head who starts whinging how it's reshaped their entire lives, but in some deep niche way that is completely irrelevant to the average working adult.

I have also talked to way too many delusional maniacs who are literally planning for the day an Artificial Super Intelligence is created and the whole world becomes like Star Trek and they personally will become wealthy and have all their needs met. They think this is going to happen within the next 5 years.

Well I'm on board for fuck intellectual property. If openai doesn't publish the weights then all their datacenter get visited by the killdozer

Eno does strike me as the kind of person who could use AI effectively as a tool for making music. I don’t think he’s team “just generate music with a single prompt and dump it onto YouTube” (AI has ruined study lo fi channels) - the stuff at the end about distortion is what he’s interested in experimenting with.

There is a possibility for something interesting and cool there (I think about how Chuck Pearson’s eccojams is just like short loops of random songs repeated in different ways, but it’s an absolutely revolutionary album) even if in effect all that’s going to happen is music execs thinking they can replace songwriters and musicians with “hey siri, generate a pop song with a catchy chorus” while talentless hacks inundate YouTube and bandcamp with shit.

Yah, I'm an AI researcher and with the weights released for deep seek anybody can run an enterprise level AI assistant. To run the full model natively, it does require $100k in GPUs, but if one had that hardware it could easily be fine-tuned with something like LoRA for almost any application. Then that model can be distilled and quantized to run on gaming GPUs.

It's really not that big of a barrier. Yes, $100k in hardware is, but from a non-profit entity perspective that is peanuts.

Also adding a vision encoder for images to deep seek would not be theoretically that difficult for the same reason. In fact, I'm working on research right now that finds GPT4o and o1 have similar vision capabilities, implying it's the same first layer vision encoder and then textual chain of thought tokens are read by subsequent layers. (This is a very recent insight as of last week by my team, so if anyone can disprove that, I would be very interested to know!)

But the people with the money for the hardware are the ones training it to put more money in their pockets. That's mostly what it's being trained to do: make rich people richer.

It's moronic. Currently, decision makers don't really understand what to do with AI and how it will realistically evolve in the coming 10-20 years. So it's getting pushed even into environments with 0-error policies, leading to horrible results and any time savings are completely annihilated by the ensuing error corrections and general troubleshooting. But maybe the latter will just gradually be dropped and customers will be told to just "deal with it," in the true spirit of enshittification.

The real issues are capitalism and the lack of green energy.

If the arts where well funded, if people where given healthcare and UBI, if we had, at the very least, switched to nuclear like we should've decades ago, we wouldn't be here.

The issue isn't a piece of software.

Yeah, the AI absolutely is a problem.

AI is noto a problemi by itself, the problemi is that most of the people who make decisions in the workplace about these things do not understand what they are talking about and even less what something is capable of.

My impression is that AI now is what blockchain was some years ago, the solution to every problemi,which was of course false.

Dunno, the part about generative music (not like LLMs) I've tried, I think if I spent a few more years of weekly migraines on that, I'd become better.

That is a big issue, but excessive power consumption isn't intrinsic to AI. You can run a reasonably good AI on your home computer.

The AI companies don't seem concerned about the diminishing returns, though, and will happily spend 1000% more power to gain that last 10% better intelligence. In a competitive market why wouldn't they, when power is so cheap.

Large power consumption only happens because someone is willing to dump lots of capital into it so they can own it.

Ollama is FOSS, SD has a proproprietary but permissive, source-available license, but it is not what most people would associate with "open-source"

most of the replies can be summarized as "the biggest problem with AI is that we live under capitalism"

Both AI and social media are a shit show because it's owned by a few people.

Unironically, the best social media is Fetlife. Not that it's perfect by any means--not by far--but it is designed to facilitate bringing people together.

Totally right that there are already very impressive open source AI projects.

But Eno doesn't control diddly, and it's odd that you think he does. And I assume he is decently well off, but I doubt he is super rich by most people's standards.

Apply it to healthcare, science, finances, and the world will become a better place, especially in healthcare.

That's all kind of moot if we continue down the capitalist hellscape express. What good is an AI that can diagnose cancer if most people can't afford access? What good is AI writing novels if our homes are destroyed by climate change induced disasters?

Those problems are mostly political, and AI isn't going to fix them. The people that probably could be replaced with AI, the shitty "leaders" and such, are not going to voluntarily step down.

None of it is currently useful to those right now