Apr 25, 2023
In 1970 economist George Akerlof published his groundbreaking and highly influential paper “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism.” His work netted him a Nobel Prize in 2001. In a nutshell, Akerlof’s framework explains how information asymmetry between buyers and sellers in a market can lead to a decline in the quality of goods sold, with buyers being unwilling to pay higher prices due to their inability to differentiate between high-quality and low-quality products. This can result in a “market for lemons” where only low-quality goods remain in the market.
We have been talking about Akerlof’s insights in the context of our Hourglass Economies framework, pointing out that due to the Internet, we have lost many of the previously existing information asymmetries (e.g. sites like Yelp give you a decent information edge when you consider which restaurant to dine at). But recently, driven by the cacophony of voices surrounding ChatGPT and its friends (or rivals?), Akerlof’s insight rings true in a different way: With the seemingly breathtaking and never-ending release of news covering the latest advancements in large language model-based AIs, we also get inundated with “experts” proclaiming the rise of AGI (artificial general intelligence) and, essentially, the end of humanity (or maybe just “the singularity”).
It strikes me that when you listen to the people who are actually building most of the systems in question, such as Stanford’s Andrew Ng or Meta’s Chief AI Scientist Yann LeCun, you hear a much more nuanced message: Yes, the advances in AI are stunning — but we are still miles away from AGI, a lot of the hype is just that, and we have a lot of unsolved problems ahead of us.
Meanwhile, the AI circus focuses on its perceived information advantage (typically, our AI gurus just know enough about the underlying technology to sound “smart” to the average business person) that they keep selling lemons to their unsuspecting clientele.
When it comes to any hot technology, we are typically always best served by listening to the people who are building the core technologies – not the ones who are either merely putting the Lego bricks together or, worse, the ones who just talk… (via Pascal)
