Memes

Turns out, most people think their stupid views are actually genius

Same as it ever was...

depends. for "AI" "art" the problem is both terms are lies. there is no intelligence and there is no art.

Technology is a cultural creation, not a magic box outside of its circumstances. "The problem isn't the technology, it's the creators, users, and perpetuators" is tautological.

And, importantly, the purpose of a system is what it does.

Always has been

Careful! Last time I sarcastically posted a stupid AI idea, within minutes a bunch of venture capitalists tracked me down, broke down my door and threw money at me non stop for hours.

Robots figuring out that without humans releasing gas their job is a lot more efficient could cause a few problems.

They're not really comparable. Crypto and blockchain were good solutions looking for problems to solve. They're innovative and cool? Sure, but they never had a widescale use. AI has been around for awhile, it just got recently rebranded as artificial intellectual, the same technologies were called algorithms a few years ago... And they basically run the internet and the global economy. Hospitals, schools, corporations, governments, the militaries, etc all use them. Maybe certain uses of AI are dumb, but trying to pretend that the thing as a whole doesn't have, or rather already has, genuine uses is just dumb

Line go up 🤓

Nowadays you can actually get a semi decent chat bot working on a n100 that consumes next to nothing even at full charge.

As with literally every technical progress, tech itself is no problem, capitalism usage of it is.

In what sense does a small community working with open weight (note: rarely if ever open source) llm have any mitigating impact on the rampant carbon emissions for the sake of bullshit generators?

In my country this kind of AI is being used to more efficiently find tax fraud and to create chatbots for users to understand taxes, that due to the much more reliable and limited training set does not allucinate and can provide clear sources for the information given.

Machines will do exactly what you tell them to do and is the cause of many software bugs. That’s kind of the problem, no matter how elegant the algorithm, fuzzy goes in, fuzzy comes out. It was clear this very basic principle was not even considered when Google started telling people to eat rocks and glue. You can’t patch special cases out when they are so poorly understood.

The relative number here might be more useful as long as it's understood that Google already has significant emissions. It's also sufficient to convey that they're headed in the wrong direction relative to their goal of net zero. A number like 14.3 million tCO₂e isn't as clear IMO.

Miles is chill in my book. I appreciate what he is tackling, and hope he continues.

It seems that there are much worse issues with AI systems that are happening right now. I think those issues should be taking precedent over the alignment problem.

Some of the issues are bad enough right now that AI development and use should be banned for a limited time frame (at least 5 years) while we figure out more ethical ways of doing it. The fact that we aren't doing that is a massive failure of our already constantly-fucking-up governments.