OpenAI Is Slowing Down Its AI Training

The motivated reasoning of people who have dismissed AGI capabilities and risks year after year are endless, even after they have been proven wrong at every step of the way.

Every time doomers have been vindicated in their accurate predictions, denialists will come up with the closest thing they can grab to rationalizing their pre-existing belief. "LLM's aren't that useful, It's all marketing, capabilities are slowing down, they want to regulatory capture, etc."

I very much agree. There is need for more innovative llm training methods than just give it more data.

theyve been talking about their upcoming astra models for a while, this is neoliberal self-regulation talk