Thursday, January 4, 2024

A.I. concerns

Programming note: I am sorry for today's complaining. I did not think very hard before writing it. Also, Today's Learning will be off Friday and Saturday, back Sunday.

Representative Ro Khanna, one of the more interesting members of Congress because of his median-voter, widely ranging opinions, published an op-ed in The New York Times today: "Democrats Must Not Repeat the Mistakes of Globalization." He begins to list some errors of globalization:

You remember the story that those Davos conferences broadcast to the world: Everyone will be able to get a knowledge job. Consumer goods will become cheaper. Globalization coupled with the internet will lead to prosperity for everyone.

Well, it didn’t quite work out that way.

What these Davos participants missed was how unfettered globalization hollowed out the working class here at home. We are all familiar with the consequences now: shuttered factories and rural communities that never saw the promised jobs materialize. As the American dream slipped away from them, many people developed deep and justified resentment. They saw the obscene concentration of wealth and opportunity in districts like mine in the heart of Silicon Valley. The evangelists for the new economy were prescient about the wealth generation that globalization and the internet would unleash but wrong that it would increase economic opportunities for all Americans.

It is just not clear that "unfettered globalization" a) happened and b) hollowed out the working class here at home. The focus on here at home is an interesting one. If we look beyond home, at say, China, Vietnam, or Bangladesh, they have undoubtedly become wealthier and better places to live for the people who live there. 

GDP per capita is can be affected by other factors, so here is median income:

Both of these metrics show that since these Davos conferences, poor countries outside of the United States have indeed become wealthier. That is good! Khanna says that the working class has been hollowed out at home. Real median incomes have risen in the United States, too. 

I couldn't find a good chart, but looking at household income quintiles, from 1990-2019, the lowest quintile of households have almost doubled in real income while the fourth quintile has only risen by 28%. Yes, there has been a lot of reporting on increased income inequality, which is bad, but there has even been research contradicting that. It is a bit weird that a representative from Silicon Valley (median income: $166,489) is telling us about his constituents from Silicon Valley and telling us that wealth is concentrated in Silicon Valley because of globalization (it does not strike me that Facebook, Google, or Netflix were waiting for NAFTA or Davos in order to start exporting software).

Ro Khanna brings up self-driving trucks and a California bill to require truck drivers to sit in one when they roll out. In the long term, I don't think truckers are the backbone of the American economy. I think that we should change the way our transportation works in order to privilege more sustainable transportation. This might mean less trucks and more trains. Does that mean I want to kill trucker jobs? No. Is it really about jobs, or safety? There has been a ton of news about Cruise stopping operations nationwide after being told off by San Francisco's government. Cruise cars crash way less than regular old human driven cars! If trucks, which just drive endlessly between white lines, could be permanently more alert, that would be incredible, I think. 

He brings up the AI provisions in the new Writers Guild contract, that prevents AI from being used to write scripts. I support workers! But if AI can replace your job, a contract is not going to save you! It is not like AI can write a screenplay in present. Research from Google's Deepmind showed that researchers were able to make ChatGPT spit out training data with some code and $200. Is AI actually creating new content? I don't know. I have found ChatGPT personally to be very useful for just searching things that are hard to search on Google. If there is some kind of super-ingenious AI that can write movie screenplays, the world has bigger problems. Film companies that try to put up an AI generated movie (it would perhaps look somewhat like any of the last 5 Marvel movies no one heard anything about) against an actual good, interesting movie written by a sentient screenplay writer will find themselves a loser. 

Khanna brings up employer ownership of firms they work in. Employer ownership is a good idea, but is certainly not an end all be all. It also has, uh, nothing to do with AI? Khanna links to an Newsweek article, claiming it says that tens of thousands of employees will be replaced by AI. The article, unsourced, says that Google is about to lay off 30,000 salespeople and replace them with AI. The source is an internal announcement from the head of ad sales in the Americas. The Newsweek article does not even link to the original article from The Information (which is paywalled, so I used a Reuters source), which does not even try to claim that Google will fire 30,000 people (15% of the company). (If Google lays off 30,000 people, I will be very owned.)

Trying to regulate AI in some kind of attempt to make sure we keep jobs for people is a bit difficult to do right now when we have literally no idea what AI will do. McKinsey likes to say that 70% of jobs will be automated by AI or that 400-800M jobs will be destroyed by AI. Whether or not you believe that a bunch of Excel models made by 22 year olds at McKinsey can accurately predict the job market in 10, 20, or 30 years, how would you even start to figure out that 70% of jobs can be replaced by AI? Every job being affected by AI is great for McKinsey's business model. "Hey, there's this new super ambiguous technology that affects your entire business, hire us and pay us a ton of money and we will help navigate the ambiguity for you and your business."

AI will undoubtedly be transformative. I just think that attempts to figure out entirely what AI will do right now in order to regulate it will find themselves outsmarted by just regular old people, let alone the superpowered "artificial general intelligence" that people in Silicon Valley are scared of. Comparing AI to other things like globalization is just a bit challenging, especially because things like globalization have generally had good results for material outcomes for actual people. I genuinely believe that life could be better, and that there a lot of changes that can be made to actually do that (the Californian house is doing a lot of them, like SB 50). But saying that AI will further hollow out the working class is a fascinating claim. You would have to actually substantiate that the working class has in fact been hollowed out.