This is Bloomberg Opinion Today, the muscle-grappling subversion of Bloomberg Opinion’s opinions. On Sundays, we look at the major themes of the week past and how they will define the week ahead. Sign up for the daily newsletter here.
Last spring, I wrote a column about the fears of some foreign customers that America’s F-35 fighter jet had a “kill switch” to “render it ineffective” because, well, I dunno. I never figured out exactly why anybody thought anybody would put a kill switch on a plane they are selling at a cool $100 million a pop. The whole fuss seems to have been about President Donald Trump’s behavior toward longtime allies, which ranges from ambivalent (Asian allies like Taiwan and Japan) to hostile (European allies like, well, everybody). [1] But exactly what it had to do with Trump’s resentments never became clear. If it makes little sense to put a kill switch in a fifth-generation fighter, where are they needed? There are some obvious places: motorcycles, boats, trains, smartphones, Formula 1 race cars, fake Formula 1 race cars. BUT NOT MOVIES! Unless you want a 9% Tomatometer score or a 16% Popcornmeter score, respectively:
Source: YouTube
What about chatbots? Possibly. Consider this headline from Parmy Olson: “Anthropic and OpenAI Users Still Fear an American Kill Switch.” So, if you can’t sabotage a $100 million plane, how about something five companies alone are spending $313 billion on … this year alone!
“Whatever went on in the closed-door negotiations between Anthropic PBC and the US government, it worked,” Parmy writes. “Foreign access restrictions on the company’s Fable 5 and Mythos 5 artificial intelligence models have been dropped after Anthropic resolved the Trump administration’s safety concerns.” The key words here are “whatever went on.” Because we don’t know: Commerce Secretary Howard Lutnick said the company had agreed to collaborate on “protocols and standards” for potentially dangerous models. That’s dangerously vague: Steven Seagal laughs at your protocols, and he certainly has no standards. “The standards should be public, so that enterprise customers of Anthropic and OpenAI that are already becoming more concerned about costs don’t have to worry about a kill-switch risk too,” Parmy adds. Killer AI was something of a theme for Parmy. Her other column last week was headlined: “Google’s Power Struggles Are Killing Its AI Mojo.” “The problem is likely Google’s messy history in product development. Managers are incentivized to launch new products and then move on to other teams because that’s the fastest route to career progression,” Parmy explains. “Google has become notorious in enterprise software for rolling out an array of products that compete with one another and then fizzle out. There’s even a website devoted to its graveyard of failed tools: killedbygoogle.com. That haphazard approach has now tainted AI coding.” While Google is busy killing its own products, a different tech giant has its eye on more venerable (and vulnerable) prey: Detroit. “Slate Auto, the electric vehicle start-up backed by Jeff Bezos, is a grand experiment in whether austerity sells — and a warning for the US auto sector,” writes Liam Denning. “Slate recently unveiled its flagship: A barebones, two-seater, battery-powered pickup starting at just under $25,000. It is a gray riposte — complete with hand-crank windows — to the spiraling cost of new vehicles, now more than $51,000 on average.” If Slate becomes a Detroit-killer, it may be because Motown forgot the average Joe.
Detroit also seems to be doing a pretty good job of killing itself — by killing off jobs. Ford CEO Jim Farley has predicted that AI will replace half of all white-collar workers. “Ford is hardly the only example of a company besotted with AI,” writes Gautam Mukunda. “Imagine asking any experienced executive, ‘Is evaluating your people by the amount they spend, not the output they produce, a good idea?’ You probably wouldn’t get an answer through the laughter.” But if you “sprinkle some AI pixie dust on the question,” you’ll find that “successful companies and intelligent executives make decisions that in any other context, they’d reject out of hand.” Slate’s stripped-down pickup had been billed as not only a Detroit-killer, but also a Tesla-killer. Hmmm. “Now that Ford Motor Co. is an artificial intelligence-enabler, its vehicle sales figures are irrelevant. I’m joking, obviously. With Tesla Inc., I’m only half-joking,” Liam writes. “Tesla’s quarterly vehicle sales numbers, the latest of which dropped Thursday morning, have become something like an ancient ritual: Everyone still observes the formalities, but the true meaning has become obscured.” At big banks, AI might not be a job-killer, but something even worse: a bonus-killer. “Where AI has augmented jobs and helped people with tasks, in all kinds of white-collar work, not just finance, the productivity gains are concentrated in those with the least experience,” writes Paul J. Davies. “For bankers – and almost everyone else – the main AI survival strategy is to become as good as possible at using the new tools you have and try to ensure that the rest of your job is hard for anyone else to do. Or, to accept that your workday is going to become less creative — and probably pay a much smaller bonus.” So, are C-Suites now going to come with a kill switch? Bonus Psycho Killer Reading:
What’s the World Got in Store?
- NATO summit, July 7: Poland Is Hedging Its Bets on the US — Hal Brands
- US trade balance, July 7: Clarity on the USMCA Is Essential for Economic Success — The Editorial Board
- FOMC minutes, July 8: Warsh Must Beware of Curves Flattening to Deceive — John Authers
On to the most hapless of our great institutions, higher ed. “At America’s colleges and universities, generative artificial intelligence is bringing on something of an existential crisis,” writes Justin Fox. “Large language models like OpenAI’s ChatGPT and Anthropic’s Claude ‘promise productivity while subverting the process of grappling with ideas, and with overuse they weaken the very muscles necessary to use them responsibly,’ a University of Chicago chemistry professor wrote.” Um, if the professors aren’t happy with machines doing the teaching, just think about the parents:
While the elites grapple with the new age, less-prestigious schools are giving it a muscular embrace, says Justin: “At Per Scholas, a vocational-training nonprofit that’s based in the New York City borough of the Bronx and has outposts across the country, AI is already used for tutoring, practice quizzes and mock interviews.” Justin visited Per Scholas a few weeks ago, and talked to three students about AI education. “It all sounded pretty straightforward,” he writes. “In their telling, it’s most useful for practice quizzes, study guides and answers to mid-lecture questions about difficult concepts that they don’t want to bother the instructor with.” “Soon it may be handling 70% of instruction,” says Justin. “Yet it doesn’t seem to be unleashing any existential crises.” Note: Please send square pizza and feedback to Tobin Harshaw at tharshaw@bloomberg.net
|
No comments:
Post a Comment