Superintelligence (Part 1)

When you got your first cell phone, it was relatively easy to imagine what life was going to be like if everyone had a cell phone. The cell phone had only a single use - making or receiving phone calls from anywhere.

When you got your first smartphone, it wasn’t possible to imagine what it could do in the future. None of us could imagine Twitter or Instagram or Angry Birds. We could not imagine the ability to take professional level photos and videos and send them instantly to anyone or everyone in the world

Some technologies are like that. Their capabilities are so open ended that it’s not possible to imagine what we can do with them several years later.

The hot, new technology is artificial intelligence.

The four most well known examples of products based on AI are ChatGPT, Claude, Grok and Gemini.

With these tools you can talk to them much like you would another human. You can talk (out loud), they will listen and they will speak their reply back to you. All Teslas come with Grok so I often talk to Grok once I drop off the kids or on my way to pick them up.

Or you can text messages and they’ll reply. My preferred tool for back and forth writing is Claude. Over the last 12 months, Claude and I had nearly 2,000 conversations (not a typo. I was floored when I saw how much I used it).

Some days we write software together in the same way I worked at Amazon or when I ran one of the startups I founded. We talk about what the software should do and then when Claude is done writing it, we make sure it works as expected. From October to December, Claude and I wrote the equivalent amount of code that one of my startups wrote in a year with about 20 engineers. 15 years ago that cost around $4m and took a year. With Claude it cost $600 ($200 per month) and took 11 weeks.

Other days, we do research together and Claude writes it up. For example, I have a forty some page document that is a customized guide to the colleges and majors that my eldest son is interested in. When my son has a new question or gets interested in a new school or a new major, I ask Claude to update the document based on my list of questions and we make corrections together.

Why do I do it and not him? I’m a more experienced researcher. I know how to phrase questions better. I have better instincts when something might be hallucinated. As he gets closer to college, I will stop and he will own his own journey as he should.

The current state of Claude and my knowledge is that we can work together well enough to do about 5-10 people’s work at about 10 times the speed. So 50-100 times my regular speed in a single day. Claude does that for less than $10 per day.

Claude’s current rate of improvement for the things we do together is about 4x better or faster or both every 6 months.

So in 2 years, I will be able to do 256 times as much work as I can do today.

In 3 years, 4,000 times the work. And so on.

So how does this feel emotionally?

It starts off frustrating. The technology is improving and my ability to know how to use it is improving. In the early days, you get lots of subtle errors and hallucinations.

Then elation joins the frustration. As your skills improve, you begin to feel very productive and that you’re only limited by your imagination and ability to learn to use the tool properly.

Then happiness joins the party as you realize how many things will get better for all of humanity.

So how can this apply to the world outside of me?

Let’s explore two real-life examples.

We are going to talk about two of my favorite jobs - teachers and doctors.

There are roughly 330m people in America. 4m of them are teachers. 1m of them are doctors. I think we can all agree the world would be better with more teachers and more doctors.

Today, a single teacher in America has on average 24 students. If AI could help them double the amount of work they do in a day by helping grade, meet each student where they are with a custom explanation or more sample problems, or by encouraging another student, it might make the class feel and more importantly perform as if it were a class with 12 students per teacher. Two years from now when AI makes a teacher four times as productive it would make the class feel like 6 students per teacher. It is not hard to see how AI has the potential to positively impact the life of both students and teachers in radical ways over the next 3-5 years.

With doctors, the change is potentially even more incredible. I don’t know any doctors who feel like they both know enough, could learn enough even if they had enough time, spend as much time on a case as they want AND spend enough time with a patient. My sense is many doctors wish they had 2-3 times more time to do the same work better. The potential for an AI assistant to make our doctors more productive is incredible.

These smart assistants can know 100 or 1,000 times more than a single human and that number is growing. They can work 168 hours a week. They don’t need vacations and can work 52 weeks a year. They can be awake every night., They can manage copies of themselves. They cost far less. They take less time to train. Their career can span many more decades than a human doctor. All this bodes well for doctors and their patients.

The story of the next 3-5 years will be how artificial intelligence enhances our existing workforce. It won’t replace many humans at all.

In many cases, it will take people who were maybe less capable than their peers and make them the new high performers. Over the very short term, it is hard to see this revolution as anything but positive and amazing for individuals and society.

But…

I’m saving that for tomorrow’s newsletter.

Readers can only absorb so much in a single dose.

Going forward some posts will be sent to the newsletter only and not shared on social media.

If you want to receive all posts in your inbox for free, please subscribe.