CIO Talk: Collies on the Moon, or, Are We Done with LLMs?

by
Barry Briggs

Before joining Directions on Microsoft in 2020, Barry worked at Microsoft for 12 years in a variety of roles, including... more

Earlier this month, Berit Anderson of the Strategic News Service wrote, persuasively, that “we have already reached maximum LLM utility.” She rests her argument on the increasingly clear and evident fragility and error-prone nature of LLMs. As often as not they get things wrong: they “hallucinate” to use the term currently in vogue. Given that Microsoft has made enormous bets on AI — not only is it the foundation of their technology strategy but it’s at the core of everything marketing (which worries me, because I think they may be ignoring other products)…is the AI (and Microsoft’s) bubble about to burst?

As Elon Might Say, “Concerning…”

I recently downloaded Meta’s much hyped Llama 3.1. Running it on my desktop using Ollama (if you haven’t tried this, you really must), I conducted a test in which I asked it a single question then shut it down, restarted the app, then asked it again. I did this three separate times.

The question: “Who is Barry Briggs who used to work at Microsoft?”