Technology
by Henry Kissinger, Eric Schmidt and Daniel Huttenlocher
What happens when machines know?
A former secretary of state, Google's former boss and an MIT computer scientist wrote this together. Henry Kissinger, Eric Schmidt and Daniel Huttenlocher argue that AI reaches real answers by routes humans can't follow, and that we've barely started to think about what that means.
Hear a sample · read by Laura
DeepMind's AlphaZero was given only the rules of chess and taught itself by playing against itself. It then beat Stockfish, a leading chess program, playing in ways grandmasters had never seen. The authors see a new kind of knowing, where we can check the result but can't follow the reasoning.
The authors' name for services like Google search, Facebook and TikTok, used by huge numbers of people across borders and run on AI. They shape what billions see and read. No government designed them, and even their makers can't always say why the AI shows one thing over another.
Nuclear weapons can be counted and watched from space. AI can't, because it's software that is easy to copy and useful in peacetime too. The authors, one of them a veteran of Cold War arms talks, argue that rival powers still need to talk about limits.
Since the Enlightenment, people have seen reason as what sets them apart. If machines can drive, diagnose and write, that picture wobbles. The authors worry people may defer to AI without thinking, and want humans to decide on purpose which choices stay with people.
AlphaZero
AlphaZero taught itself chess in hours and found moves no human had tried.
Halicin
MIT researchers used AI to find halicin, an antibiotic that humans had missed.
Knowing
AI can reach right answers without being able to explain how, which changes what it means to know something.
Platforms
AI-run platforms now shape what huge numbers of people see and believe.
Security
AI weapons may act faster than people can think, so nations need limits agreed in advance.
A story from the book
Researchers at MIT trained an AI on molecules known to kill bacteria, then had it search thousands more. It picked one nobody had thought of as an antibiotic, and it worked against bacteria that resist existing drugs. They named it halicin, after HAL in 2001: A Space Odyssey.
Hot take · true or myth?
How AI is used is best left to engineers.
Myth. The authors, a diplomat, a businessman and a computer scientist, say these questions are too big for technologists alone. They want philosophers and political leaders deciding too.
One thing to try
This week, when an app or chatbot gives you an answer you'd act on, ask what it's based on. If you can't find out, check it another way first. The authors expect us to lean more and more on answers we can't trace, so build the habit now.
Knotd has the key ideas from 250+ non-fiction books, each in a few minutes, in plain words. The Age of AI comes with a Hot Take round to make it stick.
Get the whole book on KnotdOpen in the appWhat happens when machines know? A former secretary of state, Google's former boss and an MIT computer scientist wrote this together. Henry Kissinger, Eric Schmidt and Daniel Huttenlocher argue that AI reaches real answers by routes humans can't follow, and that we've barely started to think about what that means.
Answers without reasons; Network platforms; Arms control for code; What's left for humans.
This week, when an app or chatbot gives you an answer you'd act on, ask what it's based on. If you can't find out, check it another way first. The authors expect us to lean more and more on answers we can't trace, so build the habit now.
Yes. Knotd has the key ideas from The Age of AI in a few minutes, in plain words, plus a Hot Take round to test them. It's one of 250+ non-fiction books in the app.
