People tell me all the time that they feel uncertain about where AI is heading. I understand it, but uncertainty is not a position - it is a state you sit in while other people decide things. It does not help you make a call about your business.
Curiosity does. And the way to be usefully curious about AI is to read both ends deliberately. Listen to the doomsayers. Listen to the AI companies telling you it is all going to be beautiful. Then understand that the truth sits somewhere between them, and that nobody in the middle is selling you anything.
The short version: read the people who think this ends badly and the people who are building it, and treat both as interested parties. This page is what I actually read and listen to. It is a signpost, not a syllabus.
The books
The first one is the most directly useful if you are running an organisation. Several of the others are a few years old now, and this field moves faster than publishing does. That is not a reason to skip them - the arguments hold up better than the predictions - but it is exactly why the podcasts below matter more for anything current.
The New Automation Mindset - Vijay Tella. Start here if you only read one. It is the practical book of the five and the closest to what we do day to day: less about whether AI is good or bad, more about how an organisation actually changes the way it works.
A World Without Work - Daniel Susskind, 2020. The clearest economic case for why this wave of automation is different from the last ones. Six years old and still the one I hand to people first, because it argues about work rather than about technology.
The Coming Wave - Mustafa Suleyman. He co-founded DeepMind and is now CEO of Microsoft AI, which is exactly why he is worth listening to: the person making the containment argument is also holding the levers at the vendor most of our clients are building on. Read it for the containment argument, and for the fact that it comes from inside the industry rather than outside it.
Scary Smart - Mo Gawdat. Former Google X, and the most human of the five. His argument is that we are raising AI rather than programming it, and that how we behave around it matters. Read it for the framing, not the technical detail.
If Anyone Builds It, Everyone Dies - Eliezer Yudkowsky and Nate Soares. The most recent one I have read, and the hardest end of the argument stated plainly. You do not have to agree with it. You should know what it says, because it is the case everyone else is implicitly answering.
Podcasts and voices, by where they sit
The grouping is the point. If everything you listen to sits in one bucket, you are not being curious - you are being reassured.
Balanced and current. Hard Fork from the New York Times, with Kevin Roose and Casey Newton. The one I would start with if you only pick one. Weekly, current, and neither breathless nor cynical.
Closest to the actual middle. Deep Questions with Cal Newport - a Georgetown computer science professor who runs a recurring AI Reality Check segment. He is deflationary about both ends, calling the labs' existential warnings doom-trolling while staying just as unconvinced by the productivity claims. He is not selling you a position on how this ends, which is exactly why he is here.
Sceptical. Better Offline - anything Ed Zitron puts out, really. He is the most consistent critic of the economics underneath all this, and he does the numbers rather than the vibes. Uncomfortable listening if you sell AI for a living, which is why I listen.
The optimists - worth hearing precisely because they are selling. Long-form interviews with the people running the labs: Sam Altman, Dario Amodei, Satya Nadella, Mark Zuckerberg and Elon Musk. The two feeds that consistently do these properly are the Dwarkesh Podcast and the Lex Fridman Podcast. Listen with the incentive in mind - every one of these people has a commercial interest in you believing them - and they are still the best available account of what is actually being built.
The doomsayers. Eliezer Yudkowsky on Lex Fridman is the long-form version of the book above. The wider argument lives at MIRI and on the book's own site.
Why this sits under Ideas
The Ideas pillar of GIVE is about keeping a pipeline of opportunities fed, and a pipeline is downstream of what the people in your organisation think is possible. If your leadership team's entire picture of AI comes from vendor decks, you will get vendor-shaped ideas. If it comes only from the sceptics, you will get none.
Read both ends. Then go and look at your own processes, which is where the actual answers are.
Sources and further reading