From Hype to Sobriety: Revisiting Ken Griffin on AI

In March 2023, I watched an interview with Ken Griffin on Bloomberg. Being interested in finance myself, Griffin is a well-known and respected figure. At the time, he said¹:

This branch of technology has real impact on our business… It will take an enormous amount of work that’s done today by people, and do it in a distinctly different, highly automated, efficient way.

His enthusiasm puzzled me.

I had been following developments in machine learning (LLMs, diffusion models, DALL·E—which had been publicly available for months but gained little traction, many of us had seen the “armchair in the shape of an avocado” without much amazement) and I was not particularly impressed by next-token prediction. Tools like Google Translate had seen dramatic improvement years earlier², but I was still skeptical that this represented a general breakthrough.

The strange episode involving Blake Lemoine at Google (claiming sentience in an AI system he worked on³) or the fallout around Timnit Gebru (related to her research on the risks of large language models⁴) hinted that significant, potentially groundbreaking developments were occurring within major AI labs. These were not trivial incidents, given the stature of those involved. However, the resulting narratives often became entangled in broader societal debates, sometimes obscuring the underlying technical shifts. Then, when ChatGPT was released in November 2022, it publicly showcased the type of powerful generative models that had, in hindsight, clearly been maturing inside these compute-rich companies, waiting for deployment.

I remained somewhat reserved—perhaps, influenced by Yann LeCun, not quite as enthusiastic as the average observer. But I assumed Ken Griffin knew more than I did, and I began looking into the technology more seriously. (One of my early talks on the topic from Nov 2023: 619.io/blog/2023/09/30/introduction-to-llms.)

Fast forward to last week, during a Stanford interview (youtu.be/GuF14oKon8A), Griffin said:

Do we use [GenAI] in our investment business? A little bit… I don’t think it’s going to revolutionize most of what we do in finance.

I was not entirely surprised, but still, a bit.

Meanwhile, the Wall Street Journal reports that AI investments have boosted efficiency, but profits remain elusive: “It’s time for AI to start making money for businesses. Can it?”⁵

Are we slowly shifting from hype to realism, or is the real transformation still ahead?