Reflections on AI

This is an interview from way back in 2021 that I happened to listen to recently. It’s an interesting conversation to listen to now, given how rapidly the AI landscape has changed since the release of ChatGPT.

I actually listened to this interview about a week ago, so my memory of it is pretty limited. I will just cover two main aspects of the conversation that have stuck with me.

Klein begins by referencing an article by Altman called Moore’s Law for Everything. (Hello, reading list!) The basic idea behind this is that AGI could lead to progress in almost countless dimensions in the same way that Moore’s Law famously projects steady, predictable progress in transistor technology (the number of transistors in an integrated circuit doubles about every two years).

This is an illuminating glimpse into Altman’s mind, suggesting one of the primary ideas motivating him to want to introduce AGI to the world. I must add this it is also peculiar to me how at one point in the interview Altman talks about how there is “no upper limit” to what AGI should be able to achieve. I find this peculiar because it seems to me that Moore’s law does have an upper limit, namely the physical limits of how small a transistor could possibly be (I’m no expert here, but surely no smaller than an atom?). And from what I know, much of the progress in AI research to date has come from doubling the resources dedicated to producing large language models (LLMs). I feel like if you are doubling your investment at every major milestone, once again surely there is some practical upper limit, even if it is very far away. (For example, you can only build as many GPUs as there is enough of the necessary materials on planet Earth.)

The other thing I remember from the conversation is that Klein raised concerns throughout regarding the distribution of power. He wondered how people like Altman could ensure that the immense value unlocked by AGI benefits humanity as a whole rather than only a small, privileged circle. I recall that Altman acknowledged this concern and expressed the belief that government regulation will be needed.

An interesting little debate happened around the middle of the interview, when Altman said he didn’t see any reason why a trillionaire shouldn’t exist one day. Klein pushed back on this, wondering why an individual person should need that much wealth and power. I wouldn’t say they dove particularly deep into this question; but again, it provided an interesting glimpse into the mind of the man who is currently leading arguably the most successful and influential AI company in the world.