When something new comes along, everyone suddenly has to have an opinion about it. Often the same one.
When crypto went mainstream, people who had never owned so much as a share were suddenly holding forth on bitcoin. When Ozempic arrived, people who had never once expressed a view on weight loss were just as eager to share the very same thoughts. Now it is AI’s turn. No conversation about it seems complete until someone has solemnly used the word “hallucination.” AI is utterly transformative. A vastly bigger deal than almost any innovation that came before. As big as the invention of writing, or the industrial revolution — only all at once. Whether you run a business, a club, a school — or a think tank, as I do — it is going to upend a great deal of what we take for granted. Who you hire. How you delegate. What outcomes we should even expect from our employees. It lets you work almost at the speed you can think — fast enough that knowing when to stop becomes a discipline in its own right. My own journey with AI started where everyone’s does — marveling at how my newly opened account acts as such an awesome search engine. Call it the “Does anyone still use Google anymore?” stage: you put your new tool to work running glorified Google searches. Great fun — but if that is as far as you get, you could be forgiven for thinking the whole thing overhyped. The second stage of AI is when you start using AI as a writing tool. This is where those who talk about hallucinations tend to be. And it is, frankly, the least interesting thing about AI. Sure, it is excellent for editing and polishing what you have written — but not the other way round, with you editing what it writes. That is the surefire route to slop. The third is the realization that you can do things techies spent years learning to do. Figure out how to program an agent to perform one complex task, and it can then perform that same task again and again — at a scale, and a speed, no human team could match. Information and expertise that once belonged only to those at the top of society are coming within reach of everyone else. The social implications of that are at least as large as the economic ones.
Perhaps my all-time favorite TV show is Young Sheldon. The running joke throughout every episode is that Sheldon, the brilliant boy genius, is able to figure out extraordinary things, yet is forever tripped up by the mundane. Treat AI as you might Sheldon: super smart, but dumb enough to misread the obvious.
The fourth stage of AI comes when you realize an agent can handle a great deal of your existing workflow. Instead of getting you or your team to do the donkey work, you farm it out to your new AI employee — one who never tires, and shows up on time, every time, ready to put in a full day’s work.
The fifth stage arrives once you start asking what you might automate that you never imagined could be automated at all. At MCPP we are beginning to do precisely this, in all sorts of weird and wonderful ways.
This week it was reported that AI software built by Palantir has saved hundreds of lives at a hospital in Florida by catching the signs of sepsis before doctors could. No miracle cure: the system simply reads the data already flowing through the hospital and prompts preemptive action.
Now picture the same logic applied more widely. Imagine AI scanning the records of every American over a certain age, picking out each one whose cholesterol sits in the danger zone, and ensuring they are offered statins — a few dollars a day — to bring it down. AI need not conjure something out of nothing to be transformative, but to merely take the knowledge already sitting in front of us, ignored, and put it to work.
The sixth stage of AI follows when you work out how to automate the autonomous agents themselves, so that they more or less manage one another. Ask me how that is going in a few weeks…
So where might all this lead?
I am an optimist. I think Jeff Bezos is right. Far from causing mass unemployment, AI could leave us with a shortage of workers.
Yes, there will be disruption. Anyone who fails to adapt could lose out. It may not be much fun for those whose only skill is coding. But there will also be an enormous increase in output. Every one of us has some sort of comparative advantage, and AI lets us do far more with what we have than ever before.
The result could be far greater prosperity for all — provided we let energy producers produce, rather than strangle them in red tape, and provided we steer well clear of European Union-style control from the top down. The EU’s GDPR goes a long way to explaining why so much AI innovation is happening on this side of the Atlantic, and so little on the other.
One final thought. Energy policy is about to matter more than almost anything else. AI runs on power — vast and growing amounts of it — and the states and nations that produce energy cheaply and abundantly are the ones that will get to build the future. Those that hobble their energy producers will be left watching from the sidelines.
Douglas Carswell is the President of the Mississippi Center for Public Policy and author of this post.
This post is sponsored by the Mississippi Center for Public Policy.
2 comments:
"A vastly bigger deal than almost any innovation that came before. As big as the invention of writing, or the industrial revolution — only all at once."
Bigger than fire, the wheel, and penicillin, all rolled into one! Better than sex! More important that mitosis! It's huge, I say!
And people are already sick to death of having it rammed down their throats. What a disaster.
Post a Comment