Eh… “Robin Li says increased accuracy is one of the largest improvements we’ve seen in Artificial Intelligence. “I think over the past 18 months, that problem has pretty much been solved—meaning when you talk to a chatbot, a frontier model-based chatbot, you can basically trust the answer,” the CEO added.”
That’s plain wrong. Even STOA black box chatbots give wrong answer to the simplest of questions sometimes. That’s precisely what NOT being able to trust mean.
How can one believe anything this person is saying?
(insert here the “The TikTok at home” meme format) So actually I did my own with PeerTube (self hosted server side) and Latrix (mobile client to live stream) and you can see the result at https://video.benetou.fr/w/p/hfPcHz1kCgnM6zKhfPrS4b (playlist of 6 short videos with progress over time).
I’d argue it… works. Is it necessary or useful? Well I didn’t keep up with the format but it potentially can be. My point being… we already have quite a few tools in place.