@deanwball - Some reflections on Paris and our current moment 1. The

Some reflections on Paris and our current moment 1. The odds that we are living through a takeoff to transformative ai seem high—much higher than they did to me one year ago—and most evidence I see reinforces, rather than complicates, this basic intuition. 2. Transformative ai is not social media, the computer, the smartphone, electricity, or steam power. It is mechanized intelligence, the mechanization and productization of problem solving itself. An invention for inventing inventions. Physical and digital intelligences (drones, robots, surveillance cameras, things we do not have words for today) that are “smart,” that know what is “going on” the way you do, or better. And ones that improve exponentially. This is what we are building, in some meaningful sense the culmination to date of post-gutenberg western civilization, the black monolith, or as my friend Michael toscano has called it, “the big one.” 3. This means that what we say and do now matters immensely. This should be no license to catastrophize, but neither should it be a license to sit back, relax, and enjoy the ride. 4. If you are, as I am, sympathetic to open source software, and want to durably extend that tradition to AI, please know that we have a difficult case to make to the world: what does the world look like with powerful and open source AI? I haven’t heard many good, concrete answers to this question. It is important to make them, because even if it turns out that open source powerful AI is undesirable in various ways, it is probably inevitable. 5. The EA/ai safety/rationalist communities were brought into the halls of power by Joe Biden and others because they appeared to be the experts on ai. They were summarily removed after their ideas, predictions and frameworks proved unworkable and often politically disastrous. Just because the ai optimists “won” the first round of ai policy does not make us immune from being completely discredited in the future. 6. Most people are overindexing in various ways on the lessons of deepseek. Others are extracting lessons that truly are not there. 7. Like any macroinvention, AI will both augment and replace humans. We do not know what the ratio will be. 8. You can be skeptical of “doom” and “X-risk” scenarios while acknowledging that advanced AI could have unprecedented catastrophic risk potential. 9. Electricity started out as a way of ornamenting public buildings with artificial light and ended up with assembly lines, white paint on interior walls, women’s suffrage, wires in the walls of every American factory, office, and home—and eventually, computers and the ai that runs on them. Every prediction about long term ai outcomes is a stab in the dark, and that includes things I have said here. 10. It is not wrong to be optimistic, as I largely am, but please be kind and understanding to your fellow humans who are worried, and remember that they too have a right to advocate for the policies they want with the resources they have.