If there is any chance that we could, even unintentionally, create super intelligence the only ethical answer is to not create it at all. It would be a species ending mistake at our current level of knowledge.
If there is any chance that we could, even unintentionally, create super intelligence the only ethical answer is to not create it at all. It would be a species ending mistake at our current level of knowledge.
How are you measuring intelligence? How are you defining intelligence? Formal measures of intelligence like ARC still show even SOTA LLM’s doing incredibly poorly.
As Chollet said, people are misinterpreting LLMs as looking or seeming intelligent, therefore they are, just in the same sense that they think there must be lemon in this lemon flavored candy since it tastes like lemon, when there’s no lemon at all.
This just seems like more Yudkowsky BS about how AGI is right around the corner, despite industry leaders (Ng, Chollet, etc.) saying we aren’t much, if any closer to AGI despite recent developments. There’s no indications that LLM -> AGI.