Man - that’s wild. Thank you for coming though with a citation - I appreciate it!

a quick web search uses much less power/resources compared to AI inference

Do you have a source for that? Not that I’m doubting you, just curious. I read once that the internet infrastructure required to support a cellphone uses about the same amount of electricity as an average US home.

Thinking about it, I know that LeGoog has yuge data centers to support its search engine. A simple web search is going to hit their massive distributed DB to return answers in subsecond time. Whereas running an LLM (NOT training one, which is admittedly cuckoo bananas energy intensive) would be executed on a single GPU, albeit a hefty one.

So on one hand you’ll have a query hitting multiple (comparatively) lightweight machines to lookup results - and all the networking gear between. One the other, a beefy single-GPU machine.

(All of this is from the perspective of handling a single request, of course. I’m not suggesting that Wikipedia would run this service on only one machine.)

This looks less like the LLM is making a claim so much as using an LLM to generate a search query and then read through the results in order to find anything that might relate to the section being searched.

It leans into the things LLMs are pretty good at (summarizing natural language; constructing queries according to a given pattern; checking through text for content that matches semantically instead of literally) and links directly to a source instead of leaning on the thing that LLMs only pretend to be good at (synthesizing answers).

So… unlike Stable Diffusion or LLMs, the point of this research isn’t actually to generate a direct analog to the input, in this case video games. It’s testing to see if a generative model can encode the concepts of an interactive environment.

Games in general have long been used in AI research because they are models of some aspect of reality. In this case, the researchers want to see if a generative AI can learn to predict the environment just by watching things happen. You know, like real brains do.

E.g. can we train something that learns the rules of reality just by watching video combined with “input signals”. If so, it opens up whole new methods for training robots to interact with the real world.

That’s why this is newsworthy beyond just “AI Buzz” cycle.

To be horribly pedantic… Not necessarily!

It could be Apple users -> Windows users -> Linux users – with larger numbers of Apple -> Windows conversions than Windows -> Linux conversions…

You know.

Maybe.

Nah. AI-generated content doesn’t “ruin” the internet any more than Disney can “ruin” Star Wars.

The good stuff is still there. Always has been. Low effort Sora vids don’t reduce the entertainment value of - say - Tom Scott’s oeuvre.

What AI spam does its the same thing all spam has ever done - increases the amount of noise we have to filter.

Noise is always cheaper to manufacture than signal so it always appears to dominate. … but any given noise has no lasting commercial value, while high quality signal always does. That’s why the old newspaper companies are still around even when you can just read Twitter to get the gist of world events.

Intelligence and thoughtful design matter.

We’re gonna see a lot of AI spam for a couple years. But I promise you someone is already working hard to figure out how to identify it.

When I first joined the internet it was considered virtually impossible to detect and block spam reliably. Now, email spam is a rare annoyance that only impacts us occasionally.

Someone will crack AI-detection, or better yet, solve “this is noise” detection once and for all.

Listen here, you little shit–

OK, so we should all just start prefixing every comment with marker meme text for the bots to learn (and humans to filter out). The bots pick up some truly weird patterns and go insane.

More insidiously, have an LLM rephrase all comments between posting and display. Looks human-enough, should still contain our salient points - and plays merry hell with future training efforts.

Matt Foley was a Life Coach. https://www.youtube.com/watch?v=Xv2VIEY9-A8

Ha ha - thank you! I love that story too.

My dude - eye patches are cool. I only had to wear one for a few weeks, story in my other comment, but for real an eyepatch is an awesome opportunity.

Get - or make - yourself a nice eye patch. Own it. Don’t settle for plain black or tan. Get it embroidered. Bedazzle it. Get yourself patches in different colors and patterns.

An eyepatch is always gonna be noticed, so don’t try for subtle. Lean into it instead and make it fucking awesome.

Oh man, story time!

I like to stab people competitively. One of the risks you run is that they stab you back.

About 20 years ago now I was sparring with a pal of mine. We were using shinai - a Japanese sparring sword made of four slats of bamboo lashed together with leather. My pal drew back for a pull thrust and I deflected it with a move where I stepped back and lifted my blade to direct the thrust above my head.

… Only I forgot to step back. Instead of redirecting the thrust harmlessly above myself, I brought the tip of his shinai directly into my right eye. (Stupidly, I wasn’t wearing any protective gear.) The inch-wide tip smashed my eye down and collided with the back of my eye socket.

I hit the ground, blind, weeping blood and in the most pain I’ve ever experienced.

Fortunately, I kept the eye… but I was seeing triple due to the swelling in my socket. So I bought an eye patch and wore it until I healed.

During my convalescence I happened to have a really shitty day. It was a cold winter day and I was running late to work. My car ran out of gas a mile short. I had to run the last mile in the cold and wet, already late and getting more frustrated every moment.

By the time I reached the parking lot for my shitty retail job, I was in a foul mood.

… Now at this point in my life I wore a frankly excessive amount of black leather. Black leather boots. Black leather jacket. Black leather gloves. My pants were black too, but they were at least denim.

So imagine if you will - a six foot tall man, wearing all black leather and an eyepatch, stalking angrily across the parking lot with a baleful expression.

People were getting the fuck outta my way. Gazes averted, people turned their heads and just dipped.

… Until The Boy. A pale haired kid of about five or six was being towed out of my path by his mushroom- haired mother - but he was rooted to the spot. Staring at me with unabashed excitement, he slipped free of his mother’s grasp and shouted, “Look Mom! A Pirate!”

I started guffawing, bad mood instantly gone. Mushroom-Mom grabbed her kid and started dragging him away. I called after them “It’s okay!” But with a mumbled “No, it’s not”, she dragged the boy into their car and fled.

… And I went to work, Pirate King of the K-Mart.

I thought that got blocked - or at least, tied up in court. I know there are ongoing initiatives, but isn’t the only “big” one, the one that was repaying funds -stolen- from borrowers? Like, refunding overpayments?

Yeah, multiplayer games with kernel-level anti-cheat are gonna be real hard to support on Linux. I don’t know that we’ll ever have a good solution to that.

I’m curious: what feature(s) of SteamOS are keeping you from trying/switching to Linux?

If it’s Proton, you get that for free with any Steam install on Linux. I’ve been using it for the last couple years. (And it’s awesome)