It’s a longer read, and the tangent on copyright while interesting and something Cory is passionate about is slightly off topic, but it’s the rare piece bringing up the key issue with AI. The argument goes like:

- The valuations are only reasonable if they are going to enable mass worker replacement. Yes there is the machine god argument, Wall Street doesn’t buy that.

- The tooling doesn’t have to be capable of replacing workers. The sales people just have to be able to convince execs it is.

- Even ignoring to fact that lots of people would lose their jobs this replacement would make everything worse because AI isn’t capable of replacing jobs.

- The bubble is based on the assumption everything will get better.

- We need to convince people things will get worse before they actually do.

These tools aren’t useless. They are remarkable. But that doesn’t mean they will meet the hype nor the valuations. In order to avoid an economic cataclysm it’s important for a realistic and measured narrative to take hold fast.

The article covers a lot of ground, but where he was explaining artists, rights, and copyrights, I couldn't stop thinking that in good part, what artists (and programmers, and a lot of other professions that are threatened by AIs) is more or less the kind of work that AI does.

We are meme machines, we mix, combine, reshuffle or just ensemble different memes and patterns we saw somewhere, mostly System 1 work, and then put a "it's only mine" label to it and proceed to sell that with exclusive rights. We can go further away with AIs, as we can mix things from very different fields, extrapolate trends that are just not there, make System 2 play here, and not go much more farther than that, in the sense that other humans can follow and understand it.

And the lion's share of what is around, those "owned" mixes that everyone else is forbidden to think or express on their own, don't go much further than what a current AI can do. A lot of music, books, and other pieces of "art" are just an algorithm result (performed by humans or not).

There are things with the use of AI that may be wrong or a bad trend at the very least, but don't lose sight of what we essentially are. We can take a step forward, but most of the time we are not so different.

At its core, a case for allowing sectoral bargaining which would give creatives the massive BATNA they need to overcome the imbalance of power.

An interesting read with good points regarding copyright discussions and how pointy haired bosses will hope to do a neo nafta to white collar employees like the original did to blue collar ones.

I suspect businesses will attempt to do compilation copyright if not individual asset copyright to get around tfas recipe for centaurness for artists.

If I have a story that I write and use AI to render the images, soundtrack, motions, potentially voice acting, when does what the AI did stop and where does my work begin?

If I write and voice act the thing, those pieces are copyrighted. Does that constitute enough of the creative work that while I can't copyright a particular screen grab, I can copyright the work as a whole like a phone book?

Maybe, I genuinely don't know. What if I have the AI do the voice acting as well? Probably, that is a public domain work to my uninformed opinion.

It is not a question in my mind if I also just throw in a concept and have the AI produce the script that is then rendered out as well.

AI has shown that for creative art, it all begins with the writing, even if the final work produced is entirely visual or audio, the whole point of Art and copyright is to create a shared impression, hallucination, experience from Human to human. To communicate.

If the AI rendering can properly give things like tabletop RPG Recaps (and the content that the AI is rendering is original and under copyright) how much of that original concept (such as someone playing a tabletop RPG session with a story they wrote themselves) does the final rendered work get transient copyright protection? None? Even though its based on an underlying concept that is copyrightable if written down?

Businesses will be wanting to know these sorts of questions too, and that sill help shape how artists / coders / knowledge workers at large are transformed into centaurs or the reverse, ceteris paribus.

I get the metaphor they're going for with reverse-centaur but I can't help believe that there isn't the name a animal headed, human bodied mythological creature that could be used instead.

I love Doctorow, but I think his bit about hallucinating library names needs a rethink.

In a world where you're on the hook for the code your AI writes, the job is code review, but it's just as much about wiring up tests and linters and type checkers so that errors of that kind are noticed and fixed by the AI before you even see the draft that got abandoned.

If I had to share an example about the subtle and then suddenly not so subtle ways that AI is gonna disappoint a coder it would be about the time I asked it to update my flake.nix such that kustomize 4.5 was installed instead of 4.6, and instead of sourcing the older code it patched the current version such that the output of --version was "4.5".

> and it's based on the manuscript for my next book, "The Reverse Centaur’s Guide to Life After AI," which will be out from Farrar, Straus and Giroux next June:

Yea nah not worth reading as its just advertisement.

I spent some time experimenting with AI art. He's right that it doesn't look right somehow most of the time. But in my case, if I started with a photograph I took and extracted its silhouette lines as the basis for the image I was generating, I got something a lot harder to distinguish from manually generated art.

He's right about the motives of tech Bros and executives here. But I don't think he's right about where AI will eventually go because I think it will become pervasive because it almost already is.

It's ridiculously optimistic to think you can run a 1 trillion parameter model on a phone anytime soon. So those data centers have a power generation problem and their limitation will be the extent to which the power generation problem is solved.

I just don't think AI is going to crash. But if you look at ridiculous valuations and PE ratios higher than 100 or so, timber! But crashing to 20 to 50 (CSCO after the crash from 200) is just the market pulling its own head out of its butt so I suppose there's the bubble. Good luck, however, figuring out when. Also, CSCO is still here. He really ought to look around at how many people have outsourced their thinking and writing to large language models already.

I'm a bit reminded of this Outer Limits episode minus the BCI:

https://www.imdb.com/title/tt0667957/plotsummary/

I don't think senior talent will get replaced anytime soon, but I also think that one or two fundamental innovations down the road on par with the Attention is All You Need paper will make AI dramatically better at tasks that don't require embodiment. Those innovations seem to occur every couple years across the sector. But again, good luck. predicting when and how.

See also: https://medicushcs.com/resources/the-radiologist-shortage-ad...

Before we fire all the radiologists, we'll have to centaur the crap out of them just to meet demand. But that doesn't fit his narrative so he left that out because I'm sure he's aware of it.