Here's some light weekend reading
A book about AI and some links
In this edition
Stuff you might like
Co-Intelligence: a book recommendation [repost from LinkedIn]
Stuff you might like
Sentiment seems to be gradually turning against the AI gold-rush of the last couple of years. First, an MIT professor published a study estimating AI to drive at most 0.66% productivity gain over 10 years in total. Then a Sequoia partner calculated a ~USD600B gap “between the revenue expectations implied by the AI infrastructure build-out, and actual revenue growth in the AI ecosystem”. Goldman Sachs followed with a research note arguing that there won’t be enough economic benefit to justify current investment spend. Feels like the early foreshadowing of an AI bust, similar to how so much connectivity was laid before the dotcom-bust. But that connectivity was (much) later necessary. And as the Goldman note says: “bubbles take a long time to burst.”
This NYT article about the coming war between doctors’ AI tools and insurers’ AI tools is a bit funny, and also a bit worrying, because of just how dystopian that future feels. When (not if) AI agents become real, we will have millions, or maybe billions, of AI agents negotiating with AI agents.
Someone bought a swim diaper from Amazon, then returned it in … let’s just say a clearly used condition. And then Amazon sent this soiled diaper to another customer, who (understandably) left a 1-star review complete with photos. The small business’s sales predictably fell off a cliff, while the giant tech platform carries on like nothing’s happened, no doubt protected by their standard T&Cs (which yes I used to draft when I was a lawyer). I am generally in favor of respecting commercial parties’ agreement, but seriously there is no real negotiation in these situations. Consumers usually get some legal protection, maybe small businesses should too.
Given recent events, Trump seems poised to win the upcoming presidential election, and equity markets seem to be reacting to that outcome. Yet, research seems to suggest that US presidents have either no or negative impact on stock market performance.
Co-Intelligence: a book recommendation
Do you want to build your own mental model about AI? This book helped me a lot!
Wharton professor Ethan Mollick publishes the excellent (and free) One Useful Thing newsletter, which has given me tons of great insights and powerful tips for using AI. So when I saw that he had written a book Co-Intelligence: Living and Working with AI, I was all:
Mollick proposes that we think of AI as a “co-intelligence” (i.e. some sort of partner or co-pilot), and proposes 4 rules:
🤖 Always invite AI to the table: Experiment extensively with AI to understand its strengths and weaknesses across various tasks.
🧠 Be the human in the loop: Focus on what you do best while letting AI handle tasks it excels at. Incorporate AI into your decision-making process.
👥 Treat AI like a person, but tell it what kind of person it is: Communicate with AI as if it were human to simplify interaction and optimize performance. Specify the AI's role or identity for better results.
📈 Assume this is the worst AI you'll ever use: Prepare for rapid advancements in AI technology and adapt to its evolving capabilities.
(OK I’ll admit it — I used Perplexity to generate the summary above, plus some recommended actions.)
He also suggests some useful mental models, such as the “Jagged Frontier” and “Centaurs vs Cyborgs”, that make it easier to think about AI.
Whenever I encounter someone who wants to learn more about AI, I just recommend this book as a starting point. It’s not super revolutionary or visionary, and it doesn’t do futuristic speculation, but it provides very useful mental scaffolding for people who want or need to have a point of view on how to use AI.
At the minimum, subscribe to the newsletter! It has a really good signal-to-noise ratio, and the price is right. 📧
Fun fact: I pre-ordered it and got access to some exclusive companion GPTs developed by Mollick. I’ve been using one of them for copy-editing suggestions for my posts (including this one)! But otherwise I still basically write everything, unless otherwise disclosed.