This is an experiment. I don't know how many will survive contact with a lab. But the pipeline improves with every better model, and open source means anyone can contribute.
How it compares to SciAgents, FutureHouse, Sakana:
https://t.co/YwPSc0PY4J
I built a system where 15 AI agents generate scientific hypotheses by connecting fields that don't talk to each other.
Most of these are probably wrong. Here's what it proposed after 19 sessions, and why that's the point →
If you're a scientist: I genuinely need you to tell me if these are crazy.
The system generates hypotheses but can't evaluate them the way you can. That's the bottleneck: not more AI, but your expertise.
Run your own: https://t.co/SQ8lWHDjtj
"Computer scientists have demonstrated repeatedly that #AI is no better than its #datasets, and the datasets that humans produce are full of errors and biases."
https://t.co/BSmYAKxHX6
Ho appena contattato un eurodeputato italiano e ho chiesto di appoggiare l'iniziativa #fixcopyright nell'Unione europea! Contatta gli eurodeputati da https://t.co/H21hKGJhQy #fixcopyright
Beautiful recap of 10 years of evolution of the web, with a glimpse of the future, from a Chrome prospective. "The ‘Capable Web’: A 10 Year Retrospective" https://t.co/WPTyYjjF7E
Just saw "Nothing to Hide - The documentary about surveillance and you" with, among the others, Claudio @_vecna Agosti. Very interesting and well elaborated. Congratulations to everyone involved, recommended! 😉https://t.co/k1mNr8D1rz #nothingtohide
@lucacarettoni@raistolo With my team @ememory_it we recently developed a functionality to scrape all the services you use from the email senders in your Gmail account, for digital legacy purposes. I was indeed baffled by the easiness with which Google allows you to read other people full emails.