Alex Wiltschko

The actual experience of a spring morning is not in the data set.

I don't know if we're going to be able to get into that.

Try to tackle that in five minutes.

Here's what I think.

All technology here to date, and I don't think that what we're seeing right now is any different, progresses on an S-shaped curve where it doesn't get better faster really at all.

And then all of a sudden there's something that changes and it gets better really vast.

And then it saturates and it basically plateaus.

Everything is like that up until now.

So internal combustion engines, generating light via a candle, generating light via natural gas, generating light via electricity.

Now those S-curves stack, right?

So candles, there's actually technological breakthroughs in candles while natural gas was coming online.

And candles were straight up better for like a decade because people were like improving candle geometry.

Like you don't think about candles as tech, but it was straight up tech back in the 1800s.

But those S-curves stack, right?

And when you stack them really well, it creates an exponential in some capability, like the amount of light you can produce for a dollar, right?

But it requires a bunch of different breakthroughs that each have their own S-curves.

Text generation has its own S-curve using transformer-based models.

So we call those LLMs.

I think we're getting close to the elbow and plateauing off because the limiting factor is the internet.

Right.