[bhouston]

We just are not thinking wide enough:

* Train on all of television history, and streaming content.

* Train on YouTube.

* I suspect at some point we'll have a recording of most of people's lives, e.g. live-streaming: https://en.wikipedia.org/wiki/Lifestreaming#Lifecasting

[mountainriver]

Exactly, put bots into the world with cameras and you have infinite training. Humans also need a ton of data to train on and have way more parameters than the biggest ML model today

[eru]

You can also gather arbitrarily more video data by just turning on some webcams and pointing them at the world.

In addition you can also feed your system from video games.

[istjohn]

Microphones, too.

[eru]

Yes. When I wrote video, I meant audio and visual.