>>18847708Why not? It seems that attention + scaling actually works. LLMs are getting smarter, and this happening faster than even most optimists predicted. Why would the recent massive successes in software and math not soon transfer to the other sciences?
>>18847719Astra is arguably smarter than any single human, and it didn't lead to any terminator scenario, we simply got a lot of math solved and a lot of cool new software. Doomer schizos have been consistently wrong, meanwhile giga-optimists like Kurzweil are being vindicated.