Rescuing the Future | AI will not take over? | Alex Kantrowitz and ‘Jobs are for Humans’

Today I read Jobs are for Humans – Why ChatGTB hasn’t ignited the employment apocalypse that so many predicted, ‘ by Alex Kantrowitz, republished on Slate News and Politics. Kantrowitz finds solace in the fact that 8 months after ChatGTB was introduced few jobs have been taken over by AI:

A.I. is being widely adopted, but the imagined mass firings haven’t materialized. The United States is still effectively at full employment, with just 3.5 percent of the workforce unemployed.

While I’m on the side of AI, believing that in the long run it is more likely to save us than destroy us, I see the dangers and find this a little naive. It would be hard to find an AI researcher who thought Chat GTB was designed to take jobs.

But in the Lex Fridman  interview 3 with physicist/AI researcher Max Tegmark  (see my post on Fridman/Tegmark), Tegmark warns us that the ChatGTB bots are ‘baby AIs’ – the real danger, and the reason he and many other scientists have called for a 6 month pause on AI research, will be the adults, whose capabilities will be far beyond those of the infants.

But Kantrowitz does have good points. For example:

Legal work, for instance, was supposedly squarely in A.I.’s sights, but law firms enthusiastically incorporating A.I. aren’t using it to replace lawyers. Allen & Overy, a firm that employs more than 3,000 lawyers worldwide, started working with a generative A.I. tool called Harvey last year and hasn’t replaced a single person with it.

That’s promising, and interesting. If this is happening, I suspect it’s because the Harvey program has allowed Allen & Overy to expand their business, their employees presumably able to handle more files. However, that may be putting other law offices out of business, or at least reducing them in size. The employee reduction may be happening elsewhere,

But there’s also a lot being overlooked here. Some industries were dramatically affected long ago. The automobile industry employed far more people 4 decades ago than it does today, though far more vehicles are now produced.

When I entered the insurance claims industry in the late ’60s, half the employees in a large office were typists. They’ve all been gone for a long time. Some people did experience an employment apocalypse.

Finally though, I think the transition to AI employees is simply beyond the imagination of most industry managers. For example, when I re-entered insurance offices in 1996, after 16 years doing investigation work outside them, I was shocked to find that claims handlers no longer wrote their own letters.

Instead, each company had a set of form letters – for the accident benefit claims I was then handling, there would be 50-60 letters, each of them requiring the handler to chose various possibilities from check lists, with a resulting set of at least 500 possible letters . It often took some time to figure out what you should check. Pressed for time, under heavy workloads, file handlers routinely checked the wrong thing, often with disastrous results.

Despite the crude computers we were using at that time, I immediately saw that with the assistance of a good programmer it wouldn’t have been particularly hard to have those computers take over the form letters. They would have checked the boxes almost instantly, with far more accuracy. That alone could have cut the number of employees in half.

But it never happened. Keep in mind that this was during the 1980s move by corporations to be ‘lean and mean’, when staff numbers in almost all companies were cut severely, often ignoring the dysfunctional chaos that resulted. Fewer employees was routinely assumed to be a sign of increased efficiency.

The managers of those companies refrained from using computers that way, but not because of concern for employees. I think they just didn’t have the imagination to see the possibility. I suspect most of them won’t see the possibilities arising with coming AI. The AI revilution will probably develop more slowly than we expect.

Max Tegmark thinks AI can be safely controlled. But near the end of the Fridman/Tegmark interview 3 mentioned above (about the 2hr mark), Lex Fridman presses him to respond to the pessimistic views of AI researcher https://en.wikipedia.org/wiki/Eliezer_YudkowskyEliezer Yudkowsky (also at MIT along with Fridman and Tegmark), who has been warning us for some time about these safety issues.

Yudkowsky believes we are currently on track for the complete replacement of human beings by AI and robots. Tegmark is forced to admit the strength of Yudkowsky’s thinking, and tries valiantly to oppose it. That part of the discussion is very compelling. Tegmark warns that if we give up trying to keep control of AI, we will lose control by default and only have ourselves to blame.

Yudkowsky,  by the way, is pessimistic because  he thinks the chance to control AI has probably passed, that we have run out of time.

Alex Kantrowitz says humans are underrated, that he thinks we’ll have a good future with AI. Max Tegmark hopes he’s right, but fears the explosive moment we are in. To me, it looks like the future is up for grabs.