Winning at revenue and profit does not mean Apples closed source system beat out open source . 80% of people chose open source 20% chose closed source . Profitability has no bearing.
I just wanted to say that the blog was terrific! The insights you shared about AI Hallucinations were incredibly valuable. Oh, and I came across another article on the same topic that I thought you might enjoy, too!
Something in my head is that there are so many zeros in those parameter matrices. Seems we will think we are doing things the wrong way. And it’s unclear to me the better way comes from more of the same.
> "Most talkers are playing catch-up, hallucinating a self-serving worldview that’s a mix of despondence (mope), denial (cope), and salesmanship (hope)."
Above sentence captures the essence of what is going on excellently.
> Incremental model updates cannibalize the complementary infra stack that has been built around them. If only a handful of LLMs dominate, then a robust infrastructure ecosystem matters less
I think that is only true if the model updates focus on eliminating the model's weaknesses or adding auxiliary services (rather than scaling them up, which seems to be what LLM model providers are mainly focused on)? I don't see why we won't see increased specialization (as with any other technology or service industry).
Winning at revenue and profit does not mean Apples closed source system beat out open source . 80% of people chose open source 20% chose closed source . Profitability has no bearing.
a lot of the AI "godfathers" sure look like boomers to me
I just wanted to say that the blog was terrific! The insights you shared about AI Hallucinations were incredibly valuable. Oh, and I came across another article on the same topic that I thought you might enjoy, too!
https://expertguru.ai/blogs/tech-insights/preventing-pitfalls-mindful-testing-for-early-openai-hallucination-detection
Lovely piece. Thanks for having written it.
Something in my head is that there are so many zeros in those parameter matrices. Seems we will think we are doing things the wrong way. And it’s unclear to me the better way comes from more of the same.
> "Most talkers are playing catch-up, hallucinating a self-serving worldview that’s a mix of despondence (mope), denial (cope), and salesmanship (hope)."
Above sentence captures the essence of what is going on excellently.
Position luck will play biggest role in sorting out winners https://twitter.com/mtrajan/status/1652363726585036800
> Incremental model updates cannibalize the complementary infra stack that has been built around them. If only a handful of LLMs dominate, then a robust infrastructure ecosystem matters less
I think that is only true if the model updates focus on eliminating the model's weaknesses or adding auxiliary services (rather than scaling them up, which seems to be what LLM model providers are mainly focused on)? I don't see why we won't see increased specialization (as with any other technology or service industry).