wellewds

18+

[wellewds @wellewds]

@BDBSoCal @KatieMiller I agree, but I think there's a very good chance Trump would have sold Taiwan out if it weren't for that reliance.

[wellewds @wellewds]

@CBonkosky @basedaf77 @politico Why am I an asshole for using GIFs? Do you not have any room in your heart for whimsy?

[wellewds @wellewds]

@orlandopedro I agree for AI agents that are like Codex. But Alexa with Codex-level automation and mobile device integration?

That would take off, at least with a decent portion of the public.

[wellewds @wellewds]

@CBonkosky @basedaf77 @politico

[wellewds @wellewds]

@CBonkosky @basedaf77 @politico

[wellewds @wellewds]

@CBonkosky @basedaf77 @politico You too.

[wellewds @wellewds]

@Tur713Turtle @bouncyclown @coreybrickley Hallucinations are only a big issue when you are trying to use it for legal or medical work without reviewing it or something nowadays.

The rate is low enough now that it is not a real concern for casual "research". At least with the frontier models on max reasoning.

[wellewds @wellewds]

@CBonkosky @basedaf77 @politico

[wellewds @wellewds]

@CBonkosky @basedaf77 @politico That's a hyper local issue. Do you know how little water data centers use compared to just what we use for non-edible corn (ethanol production)?

[wellewds @wellewds]

@Zenul_Abidin @kimmonismus AGI = around human-level in every domain, ASI means anything beyond that technically.

But generally people mean orders of magnitude smarter and more capable than humans. Basically an AI that can engineer just about anything that is physically practical...

[wellewds @wellewds]

@CBonkosky @basedaf77 @politico I was replying to @CBonkosky, not you. They were saying that the size of the facilities is the issue. Though afaik the hyperscale ones are all in rural areas where that isn't really a problem.

[wellewds @wellewds]

@CBonkosky @basedaf77 @politico So if AI data centers were decentralized but used more power and land overall, that would be better?

[wellewds @wellewds]

@OscarMDavies @chiefofautism Set practical goalposts/tasks that any AI can never meet or your constant statements like this mean nothing.

You can twist yourself into as many knots as you need to avoid admitting there's intelligence there if it's just philosophizing and semantics.

[wellewds @wellewds]

@Tur713Turtle @bouncyclown @coreybrickley You haven't really said why you can't treat LLMs as one source of many too. Maybe most won't but that's nothing new.

LLMs definitely can do that research in an intellectually rigorous and productive way.

But you do need high tech and media literacy beforehand to pull it off...

[wellewds @wellewds]

@Tur713Turtle @bouncyclown @coreybrickley This is just an issue of media/tech literacy.

Yes, it is going to make things worse for some but it can enable learning as well. It's a double-edged sword.

Idk why you need to check GPT 5.6 Pro more than, say, The New York Times.

But I don't think very highly of the NYT! xD

[wellewds @wellewds]

@TheMindScourge I also think cars have concentrated tourism into fewer cities. All the top destinations are walkable, have extensive public transit, high density of residential and retail, etc.

This might also be me applying my pet peeve where it doesn't belong but I think I'm on to something,

[wellewds @wellewds]

@the_lettuce_cat @Ash243x @red0neharvest LLMs are not like the metaverse. Tons of people and businesses are using these tools heavily, and are paying for it.

And businesses pay per token with a big markup (it's profitable). The only question is whether the investments will pay off before there's some sort of shock.

[wellewds @wellewds]

@AutismCapital @signulll I wouldn't say that car ownership makes the magic of youth. I think kids in cities who grow up in places like NYC taking the subway are way more free than kids stuck in the suburbs, it is what has enabled helicopter parenting.

[wellewds @wellewds]

@turnip_turns @Leninistwarrior @Lux_R_Spiral_II @onligrikdefne These models are trained on a lot of synthetic data with reinforcement learning now. Not that they don't use any new human-generated data still but it's not literally storing strings of text. It's a neural network, not a database.

[wellewds @wellewds]

@joshm We also don't have the compute for everyone to be running frontier LLM agents 24/7. At least not yet.

And LLM agents on mobile are way behind desktop, most mobile agents in use are really shitty like Gemini Nano/Apple Intelligence.