Alanna 🏳️‍🌈🏳️‍⚧️

I'm not anti the general field of Artificial Intelligence, it's an extremely interesting subject. I am deeply skeptical of the LLM/GPT products that are being deceptively sold to people as "AI solutions" or claiming to be the science fiction of Artificial General Intelligence. The cult-like following developing around them is even more disturbing, it is like something from Dune or W40K.

Please learn how these things actually work, even on a basic level, before making such wild claims.

#ai

Mar 21 at 11:26amWeb
Show threadTwotired1d ago
@kelpana Exactly this. I have found AI ideally suited for machine vision; eg inspecting for blemishes in surface finish or part presence/absence. It’s good at answering the question “does or doesn’t this image resemble the sample images.” It doesn’t understand what it’s looking at and it doesn’t need to. It’s just determining whether a data set statistically resembles another data set. Very useful in a situation where you know a defect when you see one, but can’t articulate or quantify a definition.
Show threadTwotired1d ago
@kelpana IMO, anything involving language, ie “meaning”, is a totally different situation.
Show threadAlanna 🏳️‍🌈🏳️‍⚧️1d ago
@Twotired I worked on computer vision stuff in 2012-2013. It's absolutely fascinating stuff and I'd love to mess around with it again. AI chatbots aren't even a new thing either, they're just being marketed as a "new" because the Transformer ANN circuitry makes them appear better at interaction, but it doesn't make them self-aware, reliable or sentient. All the Eliza effect like problems happening now have known about for decades - the difference is it is being mass marketed.
Show threadAlanna 🏳️‍🌈🏳️‍⚧️1d ago
@Twotired I worked with an artist on an art project back in 2015. It was an FSM written in Python playing a well matched video loop recorded by an actress. It was running an OpenCV face detector and a really basic human voice detector. Based on conditions it played varying sections of the video loop. It was running on a Raspberry Pi and put in a black booth with a TV. It was presented as an "AI therapist" and people genuinely thought this random video player was talking to them!
Show threadRicardo Segurado2h ago

@kelpana @Twotired

Thanks, I hadn't heard of the ELIZA effect before!

Now I know why I get sooo annoyed when the shed door blows shut on my back. (i.e. it's not doing it on purpose)

Show threadJoan of Contention 😷6h ago

@Twotired @kelpana Yeah, it seems some of these models can be very good at pattern recognition in images (distinct from pattern matching), and.. that's about it..?

But LLMs? Nope, no, no way. It's probabilistic output based on extremely fancy maths and severe environmental damage. Sin é.

Show threadAlanna 🏳️‍🌈🏳️‍⚧️5h ago
@clickhere @Twotired and they have been for years. The difference is business people have found something mass marketable in the chat bot interface plus LLM/GPT combination and are *milking* it.
Show threadJoan of Contention 😷2h ago

@kelpana Totally. The hype is off the scale, and everyone has lost their minds falling for it.

@Twotired

Show threadJordi (Tubal Hippy)1d ago

@kelpana

This perfectly encapsulates my feelings.

Show threadJoan of Contention 😷6h ago
@kelpana This sums up how I feel about all of that so much more succinctly that I can ever manage (even if I didn't veer off into extended rants..). Thank you!