Happy Bee-day to the young man whose punch packs a sting!

sigh

Bruno practicing his Elvis impression 😁

#wirehairedpointinggriffon #wpg #hounddog #elvis

Exactly!

(@sfgate.com should have a share to Bluesky icon not an X one)

Why is there a Twitter/X share button at the bottom of this (and every) article and not a Bluesky one?

The article is spot on. Ugh

I've been picking up Lenovo M Series Tiny's as Pi stand-ins, really nice.

Rush hour trains should be 10 cars

You really need a survey to gauge public opinion on a fare hike?

Fact: The more you clench your bum and wince, the less the pothole damages your car.

Super cool!

US joins Russia to vote against UN resolution condemning Russia’s war against Ukraine | CNN Politics The United States joined Russia to vote against a UN General Assembly resolution condemning Russia’s war against Ukraine Monday in a stunning shift from years of US policy.

For the US to side with Russia and North Korea to oppose a UN resolution condemning the illegal invasion of Ukraine defies all common sense and adds insult to the countless injuries suffered by the brave Ukrainian people. edition.cnn.com/2025/02/24/p...

If this was a reply to me I never argued AI models were autonomous or possessed autonomy or that AGI has been achieved.

You said "AI uses a database of known info and just regurgitates. If there's any "reasoning," it's built into the algorithm by the creators/developers"

This is still incorrect.

aiSlop on the rocks is an underrated cocktail

Great article thanks, AGI is a whole other conversation :)

I hope you agree that even this article supports the fact that what the models can do isn't programmed per se which was my original point.

Nice chatting with you.

The code that allows the model to learn, yes, but the result is a model that can do things it was never explicitly coded to do. The behavior is emergent, not prescribed. Just like a human isn't coded to ride a bike, it's able to learn through trial and error.

AI models learn from the data they are trained on, developing abilities that go beyond their initial programming. You won't find any code in a model that tells it how to write a poem, or how to translate from French to English but they are absolutely able to do so. The models become generalists.

You could argue that's true of humans too :) The behavior we see from these models is much less programmed and more emergent and we really don't know how humans reason either, perhaps it doesn't matter if the result is the same.

AI models don't work like that, they are capable of creating responses that are distinct from any of the actual training data, and the training data itself doesn't exist in the model like it would in a search engine or an expert system.

What you're describing is closer to what's known as an expert system, a big data store and logic that's been programmed to retrieve that information.

Even if it never gets any better than it is today it's still an amazing tool doing real work.

That's really not accurate

Octopus

These guys are amazing
#octopus #mandalaybay

Time magazine cover overlaying the towers of the Kremlin rising over the White House.

Well played. And frighteningly accurate.

This is going to be glorious

Bruno the Wirehaired Pointing Griffon

Happy Thanksgiving! Yes I would like some turkey.

#wirehairedpointinggriffon #wpg

Lol! These are all comedy gold