I’ve seen a lot of exceptionally smart peers buy this idea without interrogating it. If these models could be trained solely on even medium sized data sets, the consumption race between the big players wouldn’t be happening. There are 2-3 companies that own an amount of assets that even gets close
jess m. 🌤️I cannot stress this enough: this is not possible in a literal technological sense: you cannot train an AI and expect even comprehensible results with under ~1 million pieces of data, nobody can supply all that with the variance needed. You can only achieve it through wanton scraping
