AI Watch, Part 2

It asked the question, what if AI meets the most optimistic expectations?  It increases productivity, goes beyond what human beings are capable of, makes new discoveries, avoids the scary Sci-Fi scenario of trying to destroy the human race, and makes our lives better in wonderful ways.  What will AI at its best do?

Answer:  It will destroy the economy. Also, the society.

Most of the report is a thought experiment written from the perspective of a mere two years in the future.  Read it yourself.

To put it simply, the analysis says that companies will embrace an ascendant AI and make huge profits by increasing productivity while cutting labor costs.  But the labor that AI will replace comes from high-salaried white collar jobs.  As unemployment soars in those “intelligence” fields, consumer spending plummets.

White collar workers, whose training is not easily transferable to other occupations as blue collar workers, will be forced into the “gig” economy, but the massive influx of Uber drivers will send those wages even lower, as will the AI development of self-driving vehicles.

The displaced knowledge workers will no longer be able to pay their high mortgages, resulting in a housing crash, with similar catastrophes in other industries.  Large numbers of defaults will devastate the credit markets and cause banks to collapse.  Tax receipts will tumble, putting the government into crisis.

The report describes all kind of negative loops and downward spirals. “While the politicians bicker,” says the report, “the social fabric is fraying faster than the legislative process can move.”

At the end, the report gives the bottom line (Citrini’s emphasis):

For the entirety of modern economic history, human intelligence has been the scarce input. Capital was abundant (or at least, replicable). Natural resources were finite but substitutable. Technology improved slowly enough that humans could adapt. Intelligence, the ability to analyze, decide, create, persuade, and coordinate, was the thing that could not be replicated at scale.

Human intelligence derived its inherent premium from its scarcity. Every institution in our economy, from the labor market to the mortgage market to the tax code, was designed for a world in which that assumption held.

We are now experiencing the unwind of that premium. Machine intelligence is now a competent and rapidly improving substitute for human intelligence across a growing range of tasks. The financial system, optimized over decades for a world of scarce human minds, is repricing. That repricing is painful, disorderly, and far from complete. . . .

This is the first time in history the most productive asset in the economy has produced fewer, not more, jobs. Nobody’s framework fits, because none were designed for a world where the scarce input became abundant. So we have to make new frameworks. Whether we build them in time is the only question that matters.

Read this from the Wall Street Journal about the impact of this report, concluding with a good summary: “Many of Monday’s [stock market] moves roughly aligned with the situation outlined by Citrini, in which fast-advancing AI tools allow spending cuts across industries, sparking mass white-collar unemployment and in turn leading to financial contagion.”

I’ve read several reactions pooh-poohing the report, claiming that every technological innovation ends up increasing employment and bettering everyone.  But I haven’t seen any that actually engage with the specific concerns raised in the Citrini report or giving reasons  why these scenarios will not come to pass.  The underlying issue is whether AI is qualitatively different  from other technological innovations.  I guess we’ll see.

The Rural, Small Town Resistance to AI

Oklahoma, my ancestral home, may be the reddest of the red states, where Donald Trump won in every county.  Its small, rural communities are in sore need of jobs and investment.  But they are rising up against the AI data centers that the tech companies, encouraged by President Trump, are trying to build on the red Oklahoma soil.

Similar resistance is breaking out across the country.  The Wall Street Journal tells the tale in a story by Anvee Bhutani and Amrith Ramkumar entitled These Rural Americans Are Trying to Hold Back the Tide of AI with the explanatory deck, “Fearing rising utility costs, job losses and privacy violations, residents have blocked or delayed data-center projects around the country.”

The reporters tell about how Howell Township, a Michigan farming community, stopped a $1 billion, 1,000 acre data center, by pressuring the county planning commission to refuse to approve the project.  They quote a newly-minted activist: “I assumed Republicans would be for this: business, deregulation,” said Breanne Green, a community organizer who helped start one of the first Facebook groups there opposing the project. “That just evaporated.”

The battle is part of a nationwide resistance to the AI boom, challenging President Trump and others who have embraced the technology. Fearing rising utility costs, job losses and privacy violations, locals blocked or delayed about 20 projects around the country representing nearly $100 billion in combined investment in the second quarter of last year, according to Data Center Watch, a research firm tracking the fights.

Many of them belong to tech giants including Meta and Amazon.com, fueling battles from Indiana to Oklahoma. Trump’s industry-friendly AI strategy has received criticism from other Republicans including former strategist Steve Bannon and Florida Gov. Ron DeSantis, who fought a recent executive order aimed at challenging states that pass AI regulations.

“This is a fundamental contradiction that is emerging from the Trump coalition,” said Michael Toscano, a senior fellow at the right-leaning Institute for Family Studies. More than 70% of consumers in the think tank’s recent survey in six Republican or battleground states said they were either concerned about AI or felt it was a big threat. “It’s potentially a very large problem for Republicans that are running in reliably red states” where MAGA voters don’t want to be sidelined in favor of Silicon Valley, he added.

The tech companies are telling these folks:  This will benefit you! Development! Prosperity!  Jobs!  Trump wants it!  But that’s less important to these folks–whom I know well–than the fact that they just don’t approve of AI.

UPDATE:  One problem:  a “high-pitched whine.”

(This, of course, ties into this weekend’s discussion topic.)

The Next Election Campaigns Will Be Rife with AI

In the upcoming elections, candidates from both parties are planning AI strategies.

You can expect AI-generated political ads, AI-written campaign literature, AI-crafted speeches, AI-concocted policy papers, and AI-manufactured opposition smears.  And consulting companies are gearing up to manipulate AI queries and chats to make politicians look good.  (As in yesterday’s item at this blog entitled  “How to Hack AI.”)

Politico tells about one of these consulting firms in an article by Jessica Piper entitled A tech group is launching a new effort to keep Democrats from falling behind on AI, with the deck “Tech for Campaigns plans to partner with Democratic groups to test what works.”

According to the article,

Campaigns across the political spectrum are grappling with how to take advantage of the rapidly evolving technology. Major Republican groups have embraced AI-generated content for ads more than their Democratic counterparts in the past year, although some Democratic campaigns have used AI imagery. AI-generated ads tend to be less expensive for campaigns, but strategists are still figuring out how voters feel about them — Rep. Jasmine Crockett’s Senate campaign came under fire this week amid online accusations that her latest ad featured an AI-generated crowd image, although her campaign said it “was created through hundreds of hours of real craft and collaboration between creatives and union labor” without commenting on whether AI was also used.

As for Tech for Campaigns, which the Democratic Party is planning to pay millions of dollars, here are some of the projects in the works:

Among the challenges the group hopes to tackle: Shaping how candidates show up in output from large language models such as ChatGPT, a practice known as answer engine optimization. Outside researchers have found that AI chatbots can be effective at political persuasion, with voters shifting their opinions on candidates or issues after a short conversation.

Alter said campaigns need to ensure they are well-represented in chatbot results about them, lest the chatbot basing their response more on an opponent’s research and messaging. While major companies are prioritizing shaping chatbots’ response, she said, campaigns so far have been more hesitant to work on it.

The group also hopes to study whether AI tools can help with personalized communication and how Democrats can make better use of platforms, such as Reddit, where the party has generally had less of a presence.

So if you get a personalized communication from a politician in the upcoming campaigns, you can expect it to be from AI.  And if you get into a political discussion on Reddit, you might be arguing with an AI chatbot.

And “shaping how candidates show up” and making sure candidates are “well-represented” on chatbots is exactly what we were writing about yesterday in “How to Hack AI” on how information on AI can be manipulated, simply by posting what you want to show up on the internet.

Wouldn’t you say that giving AI free rein in politics–thus driving out authentic human interaction in favor of fakery and manipulation–is a threat to our democracy?

Photo:  What Are the Power Requirements for AI Data Centers? via Hanwha Data Centers [a builder of data centers, headquartered in Korea], CC BY-NC 4.0