I feel rather nervous about the bar for software development being raised over my head by ML - something that has already happened to human language translators, where the entire industry went away, more or less. All the lower paying customers just switched to Google Translate almost overnight, taking away almost all the business - the lower tiers of the pyramid of customers.

This is currently happening to digital artists and writers. I think that the only real barrier to this happening in software is that problem specification is hard, deployment is hard & maintenance is hard. But look at chatgpt - I can see a very short line between that and a system that can produce software systems on demand like stable diffusion produces images - then auto-deploy the winner to the cloud.

And it's probably cloud providers who will end up owning all this - AWS & GCP will integrate this and have it deploy to your cloud account. There will be plenty of edge cases where it doesn't work and plenty of screw-ups, but cheaper & easier almost always wins.

You're a founder with an idea. Do you need a technical co-founder? No - you just chat with AWS AutoCode until it makes you a prototype that matches what you wanted and works - and off you go. You get some ML thing to write your SEO & marketing copy, paste it into the box.

Like most tech, there are lots of advantages to this, but also lots of disadvantages. If you haven't read this, you should: https://marshallbrain.com/manna1 - it's not happening as described here, but quite a bit of that stuff has started happening, one way or another.

So, short term, white collar, creative, developers slowly/rapidly have the bar raised above their heads - or the cost dropped below their feet and are left out to dry.

Longer term, what happens when Capital doesn't need Labour any more? Is it worth teaching my kid to code? Is it worth teaching them _anything_?

I don't know, but I do know that it fills me with foreboding.