4 ms·
Hopefully this will incentivize LLM designers to stop heavily censoring and limiting output and acceptable answers.
by lend000 2y ago
Hopefully this will incentivize LLM designers to stop heavily censoring and limiting output and acceptable answers.
- Renaud 2y agoWhy? If censoring s a problem in these applications, they’ll just make two versions: one lawful-adjacent for the government and one with all the chains that will be mandated for the pleb.
- lend000 2y agoIf it's mandated by the government that's one thing, but if they could save training costs by only using one superior version, that's the obvious business move.