AI Is Starting to Threaten White-Collar Jobs. Few Industries Are Immune
wsj.com
wsj.com
Let me provide a counter point: I work for one of the largest corporations in the US, inside an organization dedicated to advancing AI solutions across the enterprise. I would know of anything sufficiently advanced to "threaten" anyone's job white collar or not. This is simply not the case.
(going to be a little vague here for the sake of anonymity)
Here's a few examples where I see the opposite of the claimed effect.
1. In the call centers: at best improved search capabilities can help agents do work faster, maybe ultimately such productivity gains could reduce staffing, but that'll probably never happen. Search is improved a little with AI, but LLMs are kind of garbage. We've even partnered with a Big Tech Company With Famous AI People and the results are unusable.
2. Back office: call it the shuffling of papers. Think of the series of transformations of, say, new orders and things like that as they pass through various systems. These are already automated to the hilt at most every company! AI is maybe assisting with quality control a bit but whatever jobs exist in this area already survived being automated. They aren't going anywhere.
3. Fraud detection: Any company operating at any scale already has great (almost all automated) systems in place for detecting whether they are getting ripped off. Again, already automated to the hilt and anyone with a job in this area isn't going to lose it to any amount of further automation.
Articles like the one linked to need to explain, at the very least, exactly which jobs in 2024 have already escaped 50 years of automation efforts. It should beggar belief that somehow these corporations that ruthlessly cut and automate already somehow have jobs left which don't actually require a person!
Anything short of actual AGI (that is, a perfect human simulacrum that can be infinitely cloned) isn't the threat to human employment that these articles want us to believe.
The real question is what motivates these outlandish claims? My best guess is that it's paid PR firms seeding these articles at respectable news outlets like the WSJ just to keep pumping up the bubble a bit more. Keep it going and let the early money get a sweet exit before this whole thing collapses.
Joking aside, I can't even get ChatGPT to find an import or sort a list without dropping items. Everything it does has to be triple-checked. Imo, AI has the same problem as website generators did in the past. It's all fine until you need to change some specific details and have to go into the Spaghetti code.
Is this asking chatgpt to do the actual operation "in its head", or are you asking it to write python code that you then execute? Because my understanding is that it confabulates freely, but when its code that you can run and check, it tends to do better - you might have to pass back the error stack trace and run a second version.
Using your examples of call centers, back office and fraud detection as examples: - what happens when five humans can do the work of seven, because there is less shuffling paper and more automation in the call center?
Companies will staff for those five people, two folks will be out on their can.
The economy may or may not make more jobs to replace them, though we hope they do, there is no guarantee that that back office person will find similar employment. Over time, more and more of what is done will be automated away. Death by a thousand cuts.
On the other hand, this idea is in the collective zeitgeist. Its the duty of respectable newspapers to share stories that are on people’s minds. People on the margins always have concerns about losing their job, or affording rent next month. This feels like a worthwhile article to me
those unfortunate two lost their jobs 10 years ago to existing automation and process improvements.
call centers, it could be noted, are already ahead of the curve in terms of cost pressures on a business: human wages in India and the Philippines will not grow faster than what an AI company will want to charge to validate their absurd ($100 billion? $3 trillion?) valuation.
yeah, it’s cynical sounding, but the truth is that exploiting humans will continue to be cheaper than paying for humanless tech for a lot of these supposed AI use cases.
AI also needs to compete with the previous decades’ main cost cutting measure of offshore business process outsourcing.
The lessons that are AI generated are obvious and many times are completely incorrect.
On the other hand, they won the Super Bowl with a push notification timed to their TV commercial.
Me: <specific billing question>
AI: Would you please repeat that?
Me: <that same question, copy-pasted>
AI Would you please repeat that?
...<bemused/amused, suspecting AI, I repeated a dozen times with no change in the AI response>...
Me: Are you just ChatGPT? May I please talk to a real person?
<system immediately connects me to a person>
I'm bullish, and I think we'll get there, but there are precious few jobs that kind of AI can actually replace at present. Maybe a few customers give up but stay subscribed, so the ROI is good? The tech is piss-poor though if you don't have a human in the loop to work around the 10-90% error rate in common menial tasks.