[...] which would reduce the capability trajectory of OpenAI/​Google LLMs immensely, and unfortunately this is not good news from a safety perspective.
Why is this not good news? It would be great news.
Why is this not good news? It would be great news.