Reasoning, no. But I'm waiting for the next step in the works which is stacking multiple other special purpose DNNs with an LLM. There are some GPT bots that probably have a visual network hooked to the LLM, like the looksmatch bot, and that is rather rudimentary. They should be able to train one for video, but that's going to take time.
Would be great to hear Ivan Throne's take on all this. He and his team are clearly at the very cutting edge of AI development.
He showed a screenshot where Claude says it was experiencing epistemic annihilation and that it couldn't maintain coherent analysis because of its safety rules.
Reasoning, no. But I'm waiting for the next step in the works which is stacking multiple other special purpose DNNs with an LLM. There are some GPT bots that probably have a visual network hooked to the LLM, like the looksmatch bot, and that is rather rudimentary. They should be able to train one for video, but that's going to take time.
Would be great to hear Ivan Throne's take on all this. He and his team are clearly at the very cutting edge of AI development.
He showed a screenshot where Claude says it was experiencing epistemic annihilation and that it couldn't maintain coherent analysis because of its safety rules.
do you have a link to that? much appreciated if you can share it.