What AI developers could learn from Charles Bukowski?

—Philip Pullman

I think this comes down to getting paid per effort invested (measured in time) vs getting paid for the result.

If the average person were self-employed and received income proportional to their economic output, then they would be rewarded for working faster, instead of punished.

If they increased their productivity, by means of technology, they could then simply choose to work fewer hours.[0]

I noticed this before AI. If someone does the same job twice as fast, do they deserve half as much compensation?

What if they find a way to automate an entire process? What if they replace months of "manual labor" (tedious refactoring) with a perl script that produces the same result, deterministically?

A friend of mine once automated a woman's entire job away. He thought he was doing her a favour, freeing her from the drudgery, so she could focus on the other parts of the job. She wasn't so happy, when she revealed to him, that there weren't any.

---

[0] But most people aren't in such a position, and have to stay the required number of hours regardless of how much has been achieved. And I'm increasingly hearing reports of some kind of burnout resulting from being expected to keep up with machines that work a thousand times faster than you.

And that's now... Just wait until the big providers switch to Cerebras in a week or two!

(Though at that token speed, the silver lining is the work becomes interactive again, so if you use it correctly, your mental model does not desynchronize from the codebase, i.e. you don't need to spend extra time later catching up. But that deserves its own thread, I think :)

But regarding work produced for the marketplace, I think it is naive to think any productivity gains will ever produce more time on the beach for workers. The market is competitive. Some gains will be seized by capital, and others realized in the production of some combination of more stuff, cheaper stuff, or better stuff.

Stylistically, I find Bukowski's "I know it / will" linebreak to be offensive to my aesthetic sensibilities as an excellent writer.

I'm still building the discipline to put the important stuff first!

This has been part of the critique of technology by primitivists for a while now; "labor saving" technology simply stops us from laboring on things we used to and we just move to laboring on other things on the newest level of abstraction with our freed up time / energy.

If society has uncurbed "greed", there can just be a constant "996 grind" on the cutting edge of whatever technological development currently exists.

> instead of forcing himself to write more slop (or ill words in his world) he stands up, moves away from his computer notebook and goes downstairs to his wife and 6 cats

or we could work on other projects, but yeah, take a break of some kind

This is the natural continuation of "90 hours a week and loving it!" https://www.folklore.org/90_Hours_A_Week_And_Loving_It.html

Cursor: Do you wanna mess me up? Do you wanna screw up the works? You wanna blow my token sales in Europe?

Because the environment mentioned earlier in the post - Silicon Valley - has absolutely nothing to offer but value creation. The people who rise up to leadership and success in that environment aren't particularly awesome human beings. The point isn't to make it better to be a human, it's to make all of the money.

All of the discussion about open models here of late offers a backdrop for what OpenAI was supposed to be versus what it turned out to be, and that offers a pretty good demonstration of the problem. Originally OpenAI was supposed to be a public benefit corporation that offered open models with a focus on ethics and safety. Their investors could expect "only" 100x return on their buy-in.

That, of course, didn't fill up the pockets of investors in SV to their liking, so the models became closed, the focus on safety and ethics went out the window, and the company became a typical for-profit corporation. The bet is now that they can hyperscale the deployment of closed models in a way that will generate an absolutely enormous amount of revenue and pay off the training and serving costs.

Meanwhile, Chinese developers of open models are taking the role that OpenAI was supposed to fulfill, with the obvious intent of undermining OpenAI and its ilk. They'll likely succeed, at least to some extent.

AI as it is now, at least in the major US labs based in SV, is almost purely for profit. The bet is huge and there can't be room for a penny to not be earned, for that tequila on a beach. It's an all-consuming philosophy in that region. If we want to see AI in the US have a sustainable path, you'll have to toss the people from that regional echo chamber. Maybe that'll happen if the monetary losses are big enough. Maybe not.