3 ms·

I think you may be right that a simulation of intelligence—in full—either is intelligence, or is so indistinguishable it becomes a purely philosophical question

I think you may be right that a simulation of intelligence—in full—either is intelligence, or is so indistinguishable it becomes a purely philosophical question. However, I do not believe the same is true as a simulation of limited, superficial hallmarks of intelligence. That is what, based on my understanding of the concepts underlying ChatGPT and other LLMs, I believe them to be.

naasking 4y ago
> However, I do not believe the same is true as a simulation of limited, superficial hallmarks of intelligence Sure, but that's a belief not a firm conclusion we can assert with anywhere near 100% confidence because we don't have a mechanistic understanding of intelligence, as I initially said. For all we know, human "general intelligence" really is just a bunch of tricks/heuristics, aka your superficial hallmarks, and machine learning has been knocking those down one by one for years now. A couple more and we might just have an "oh shit" moment. Or maybe not, hard to say, that's why estimates on when we create AGI range from 3 years to centuries to never.