I'm not smart enough to understand the 3D chess of this administration, but war with Venezuela seems at odds with an America First agenda.

That's ok, the gap will be made up by us welcoming immigrants who will help fund Social Security for future generations! /s

As I got older, I've moved further left on issues where I think many people assume are talent or virtue related, as opposed to the luck to be born to certain parents in a good situation. I've gotten more conservative on simple rule following -- poverty isn't an excuse to throw trash on the floor.

If there is one thing I have objectively gotten better at as I age, it's growing ear hair.

This is what they took from us.

There’s a reason the movie “Office Space” resonates. A *lot* of jobs are crank turning.

Ask yourself: would you advise a young person to know how to use AI effectively? Why or why not?

People can choose whether or not they want to engage with new technology. I’ve worked in universities, government, and the private sector and have seen what many people do day to day. Based on my experience, a great deal of white collar work can and will be automated. You disagree. 2/2

I am not pushing AI onto scientists. We process a huge amount of data that benefits from automation. I don’t push it on any scientists (though my colleagues benefit from it). Through this work, I realize just how much of our back office work can be automated. 1/n

You must have me confused with someone else. That is literally not my job.

I'm not saying there isn't. But of the 300,000 Deloitte employees (to pick a company that employs a lot of white collar workers), how many work on "bespoke shit" vs. document review, spreadsheets, and preparing slide decks? Bespoke work is hard to automate but most work is not bespoke.

How many NMR technicians/managers are there in the world? Specialization has advantages. Most jobs are just not that specialized.

The world will always have edge cases that need to be escalated in some way. But huge swaths of white collar work doesn't involve edge cases.

It's not to say that everything will be solved (NMR magnets are more complicated and less common than dryers), but for many problems, an LLM + professor youtube will get a solution. 3/3

Type this into an LLM: "I'm having trouble with my clothes dryer. Can you help me come up with a plan to diagnose what's wrong and possibly fix it?"

Claude does a pretty good job of walking through the process.

For many tasks, finding a video to implement a solution is a youtube search away. 2/n

I said nothing about physical labor. LLMs or AGI (if achieved) will not supplant people that work with their hands.

That said, I'm not sure that LLMs are as incapable at helping problems/generating solutions as you allude to. 1/n

... but don't make decisions are probably in trouble. A related issue is the training gap: if the work of junior software developers and attorneys can be automated, how will we get senior ones? That restructuring will be interesting. 2/2

I'm a better thermodynamicist than a kineticist. Altman is predicting 2-3 years, which is hype IMO. I'm guessing people will offload parts of their work and management will find ways of justifying RIFs. People that spend a lot of time analyzing spreadsheets or building slide decks... 1/n

A fair question is how much of a day's work could be outsourced to a machine? As I've become proficient with AI tools, I find that much of what takes me time (but not deep cognitive effort) can be automated. PhDs are hired, at least in part, for their thinking skills. Not all jobs require them.

My observation is that many PhD scientists dismiss the utility of AI by arguing that it will never replace us. However, in many areas, AI can produce A-/B+ level work at little cost. Many people work in jobs that can be largely done by AI, so corporate America will keep investing in it.