The Great LLM Delusion - Part V: Bjarnason's Take on the AI Illusion and Dishonest (Mis)Use of Terms
THE previous part, which dealt with what looked like compromised publishers, has attracted a lot of attention. It was published shortly after Microsoft had faked "success" by claiming "clown computing" gains fueled by "hey hi" (AI) or something to that effect.
We should note that Microsoft got rid of over 20,000 employees (mostly ousted last year), not counting perma-temps and contractors (probably well over 30,000). If one counts them all, combined, together with this year's layoffs, it's not difficult to imagine that 40,000 people who worked for Microsoft (under some kind of contract) no longer work for it. Microsoft-sponsored media will try to say "hey hi" replaced them, but this is an excuse we've heard from Microsoft since 2020 when it said MSN editorial staff would be replaced by "hey hi".
This fifth part will likely be one before the last in this series (unless we come up with an important enough update) and it comes from a contributor who wishes to remain anonymous.
"By the way," he said, do you know about this guy?" (Bjarnason)
It is a book we mentioned here earlier in the series. "He wrote a good book, "Intelligence Illusion," and posted a lot of great stuff criticizing LLMs and "AI" hysteria," the contributor noted. "There is a very useful summary of issues with generative models compiled by him on a separate site." (needtoknow.fyi
)
"Regarding Bjarnason, he wrote an excellent overview of the "AI" con in action on his other site."
"He also tries to describe what is wrong with terms like "training" to give a reader a sense of just how badly they are misleading: he wrote that calling the construction of LLM from a dataset constitutes as much "training" as "construction" of a baby in the woman's womb during pregnancy."
As we shall show in the next (and likely final) part, the insincerity goes a long way. A lot of it is tied to financial interests, which severely compromise honesty and professional integrity, even in academia.
This part is relatively short, but it is an essential lead ahead of the 'finale'. █
“I feel like I have worked hard to comply with the agreement. It said you must avoid any conflict of interest or the perception of a conflict of interest. It's hard to control that last part. Perception of a conflict of interest could be anything.”
--Dwight Murphy