Yeah, it’s in AI companies interest to hype how all powerful their product is. I’m not necessarily talking about that.

When I say potentially, I’m not a doomer, so not trying to say there’s no hope, I do think we can make the world better than it is.

That said, one example of dystopian future: one of my clients makes networked security cameras. I saw a presentation where they were showing off their AI search capability. They had a static shot of an expressway, with hours of footage, and they said “I want to find a person in a bus wearing a brown shirt” and just like that pulled up the relevant moment. Not something you want an authoritarian to have. They’re going to get it anyways. But would love to have similar power available to the people.

Likewise when you think about accumulation of capital, which is shorthand for accumulation of power. AI will allow elites to accelerate that accumulation and have less use for people. “Who cares if you go on strike? I don’t need you.”

Having open source AI can turn the tables: “Who cares if you raised the price? I’ll make it myself.”

I don’t think you’re engaging with what I’m saying sincerely, and I’m trying to do that with you.

You’re right, we don’t “need” most modern conveniences. But they save us time and provide an improved quality of life. Why do we need power tools? Why motors? Why electricity, running water, artificial light? I joke, but AI is one of a long chain of tools designed to save us time, or extend our capabilities and reach. Why are you particularly opposed to this one over others? I return to my argument: AI is concerning because of its potential for change. I don’t think we can prevent it from existing, but we can make it accessible to normal people and not solely the domain of the rich.

It doesn’t need to be our new cyber god for it to matter. All the software developers I know use AI as a tool. One person I know who heads a department at a publicly traded company you’ve probably heard of said they’re not hiring more junior developers, and is worried about his long term career outlook.

That’s concerning right? Capitalists who have a choice between hiring ten people or hiring one person who supervises an AI doing the work of nine are going to choose the cheaper option, and pocket the difference.

To say that LLMs are not useful to me indicates a lack of familiarity. I’m not taking about bullshit “write this email for me” type stuff, I mean like “write this web app for me” “find this type of document in this massive trove of documents” “troubleshoot this technical issue”. And it can do it.

LLMs are not replacing humans, but they are reducing labor required for many tasks. Who receives the benefit from that reduced labor? Right now, owners. That’s why I support open source AI or preferably democratic-entity-only open source AI.

I don’t see it as fear mongering. What do you think of open source software like Linux, Firefox, VLC and so on? Because a similar argument is made there - we need to use these tools day to day, and the only option being to pay a corporation to use them would make us literally poorer, and we also wouldn’t have the features and quality we see in Windows, Mac, Chrome etc if there weren’t open source competitors.

Is lack of open source AI currently a civilizational existential threat? No. Could it be if current trends continue? Potentially. Things could get pretty dystopian if only the mega rich control tools with ultra-intelligence.

There are some open models right now, but they’re mostly ones created by private enterprise that were released to the public. Creating models is more resource intensive than making open source software. So it makes sense to pool resources for the public good.

You’re right that global warming is an existential threat. And right now the power grid is contributing to it.

At the same time, we’re seeing constant innovations in solar power and battery technology, while the cost of both are going down. Solar is now the cheapest source of energy per unit. There is enough energy from the sun to cover multiples of all current energy use by humans.

AI models started out as extremely energy intensive, but are becoming increasingly more efficient. You can now run models on your phone with abilities which previously required a server running dedicated graphics cards.

You’re also right that we shouldn’t outsource critical thinking. Still to say AI has “all the worst use cases” is wrong. Yes, it has plenty of bad use cases, by many great ones. I use it in my day to day work to save time.

Something like a law banning the use of AI will never pass, and if it did, other countries wouldn’t also pass it. A country which uses AI will outperform one that does not (economically, although probably militarily as well). In that scenario, leaving AI capability in the hands of private enterprise makes us subject to the whims of rich elites. For an analog see the case of Elon Musk turning Starlink on/off in Ukraine. Which again is why I do think there’s value in open source AI, and more specifically AI only available to democratically controlled entities - nation states but also things like worker cooperatives to compete with corporations.

I bet me and the person I’m responding to have a ton in common. Likewise me and the person who wrote that post. But I don’t relate to the position of “all AI should stop.”

I think there should be democratically accountable AI development. I’m interested in open source but the problem with it is that private corporations take advantage of it until they get big enough to try to lock out competitors. (See Google with Android) I would love to see an “open source” license which allows use by individuals and democratic entities, but forbids use by private enterprise.

It’s not going to. Saying all AI should stop is like saying “all gunpowder should stop”. Would everyone be better off if no one was using it? Maybe. But that’s not going to happen at this point, and will give you a strategic disadvantage if you unilaterally disarm.

And to be clear I’m not talking about using AI to write a screenplay or something creative, I mean like using it to write software, to optimize industrial processes etc.

Wow if its rumoured it must be true

If the instance is in the same country, a government can much more effectively exercise leverage against some guy running a server than a publicly traded corporation with lawyers on retainer.

If it’s not the same country maybe easier just to ignore.

The best case in terms of privacy is just not collecting data. Something like mullvad vpn where they don’t keep the sort of records that governments ask for.

Unfortunately anti spam measures often involve collecting identifying information (like email addresses).

Sorry for a casual, what do you mean cap at 60hz?

I just use Firefox on Ubuntu, which fifteen years ago seemed like enough.

Which also doesn’t seem that casual, but this shit is too much to keep up with. Today my engineer dad was complaining about search engines having too many ads and I asked what he used, and he said besides Google on the one computer he uses Bing on the other.

Thanks for sharing. Although I’m an enthusiastic open source user, I haven’t written any code of significance, so I’m not aware: has anyone made a license where use is restricted to individuals and democratically controlled organizations? I’m picturing that would allow for some degree of profit motive while encouraging things like worker co-ops and excluding venture capital controlled entities.

This is helpful, thanks.

This is a good post.

Thinking about it some more, I don’t necessarily mind if someone said “I googled it and…” then provides some self generated summary of what they found which is relevant to the discussion.

I wouldn’t mind if someone did the same with an LLM response. But just like I don’t want to read a copy and paste of chatgpt results I don’t want to read someone copy/pasting search results with no human analysis.

Pretty concerning that a “western democracy” is doing this, because it gives cover for the next one and the next one.

It’s easy to say “oh I’ll just stop using such and such a service” but what happens when there are no more legal services to switch to?

I think you’re misunderstanding the argument. I haven’t seen people here saying that the study was incorrect so far as it goes, or that AI is equal to human intelligence. But it does seem like it has a kind of intelligence. “Glorified auto complete” doesn’t seem sufficient, because it has a completely different quality from any past tool. Supposing yes, on a technical level the software pieces together probability based on overtraining. Can we say with any precision how the human mind stores information and how it creates intelligence? Maybe we’re stumbling down the right path but need further innovations.

Agreed. We don’t seem to have a very cohesive idea of what human consciousness is or how it works.

Other people replied, but to rephrase: the USB will not work with a car, and there will never be an adapter that will allow it to.

Instead you need a way to send the audio from the ipod headphone jack into your car stereo. Something like this.

It’s from the show “I think you should leave.” There’s a sketch where someone has crashed a weinermobile into a storefront, and bystanders are like “did anyone get hurt?” “What happened to the driver?” And then this guy shows up.