OpenAI employee Vahid Kazemi stated, "In my opinion, we have already achieved AGI... what we have is 'better than most humans at most tasks.'" This claim highlights the unconventional definition of AGI being utilized. Kazemi explains that while the AI may not surpass expert-level human performance in specific tasks, its ability to perform a wide range of tasks distinguishes it. Critiques suggest that this broad definition is not enough to classify the technology as true AGI, given that it's not uniquely better than individuals who specialize in those tasks.
Kazemi muses on the nature of large language models, arguing, "Some say LLMs only know how to follow a recipe... But even if you believe that, the whole scientific method can be summarized as a recipe: observe, hypothesize, and verify." Here, he defends the capabilities of LLMs by equating their operations to foundational scientific processes. This perspective suggests that even complex AI could develop insights and understanding similar to human intuition, albeit through different methods.
He added, "Good scientists can produce better hypothesis based on their intuition, but that intuition itself was built by many trial and errors. There's nothing that can't be learned with examples." This emphasizes the learning process inherent in both science and AI, advocating that machine learning systems evolve with experience and iterative feedback, thus allowing them to approach tasks with increasing competence.
Collection
[
|
...
]