I am asked more and more to "find ways to implement AI into the company" where I work and I laugh. There isn't much real-world use case to eliminate functions and/or reduce process as most think. I agree with that 100 %. This whole AI hype really is a bubble, it seems to me, at the present moment anyway (much like the dot.com bubble was a bubble, even though the internet indeed became huge, and companies eventually prospered off of it later on).
The only real "AI" use case in one's private life currently is GPT. (Or Gemini, or Claude, or whomever one prefers.) I think it will stay that way for a while, too.
Mainly, it's "just" a better, faster Google, IMO.
Of course you have specialized AI/Machine Learning for doctors, meteorologists etc. – but they have used AI for a long time, that's nothing new (although breakthroughs seem to be happening more and more exponentially there too the last couple of years).
So we have two quite distinct parts of is currently being hyped as "AI": LLM's and ML.
I don't know much about ML.
I don't know that much about LLM's either, although I do work with helping develop them. I would say that real progress is still happening there too, and will for a long time to come – although it is slowing, in the same way that eventually, AUM slows your CAGR.
Maybe this is a unique thing to where I live (Stockholm, Sweden), but GPT has a near monopoly here. Almost every single student I see at Stockholm's public library has GPT open. Have never seen or heard of anyone that uses Gemini or other LLM's, except for specific work purposes/reasons. Almost everyone who uses LLM's a lot (here and elsewhere in the world) seems to prefer GPT, it seems to me.
I think that LLM's will be a winner-takes-all market, much like search was. GPT will take 90 %, Gemini or someone else will take the 5 % that is equivalent of Bing in search, and a few others will get some crumbs.
JMO and I am often wrong :) (as Paul Sr. says) |
