Earlier this month I published an article titled "A sobering, short-term warning about artificial intelligence and white-collar jobs". In it, I noted:
A professional is now expected to use AI to augment or supplement his training and experience, conducting searches, market research, etc. in the background while he applies himself to current problems in the foreground. It's reported that productivity improvements of up to several hundred per cent are being claimed - and those who aren't "getting with the program", learning to use AI to work smarter, are already finding their careers being sidelined or cut short.
In the past week, I've run into several more articles raising the specter of AI domination of different sectors of the economy, and what this might mean for workers. Examples:
- "Why More Farmers Are Turning To AI Machines": As the capabilities of robotics evolve, many jobs that once required human hands are being delegated to machines. Some artificial intelligence (AI) developers working on integrating this technology into America’s farms say early data support the possibility of a major farm labor force reduction.
- "Elon Musk Backs Universal High Income Fearing AI Will Take Every Job": Instead of living aimless lives without purpose, Musk believes that there’s another option: “Long term … any job that somebody does will be optional. … If you want to do a job as kinda like a hobby, you can do a job, but otherwise, the AI and robots will provide any goods and services that you want.”
- "YouTube secretly used AI to edit people's videos. The results could bend reality": YouTube made AI enhancements to videos without telling users or asking permission. As AI quietly mediates our world, what happens to our shared connection with real life? (See also my earlier article titled "What happens to trust when anything can be faked?")
- "Medicine goes AI": Medicine is precisely the sort of bounded field with measurable outcomes and complex inputs where human cognition is simply not that good and where AI can excel. And it’s coming. Fast ... In medicine the question is rapidly shifting from "Why should we risk AI hallucination?" to "Why should we risk it from humans?"
I daresay a lot of the hype is driven by technologically enthusiastic journalists, who are trying to position themselves as authority figures among the commentariat. The reality on the ground may not be as advanced or as useful as they posit, but artificial intelligence (and those who develop and sell it) seems to be trying to follow the self-hypnotic mantra of Émile Coué: "Every day, in every way, I'm getting better and better".
I'm curious. How many of us, dear readers, are actually experiencing a positive impact of artificial intelligence in our day-to-day lives? How many of us can say that we've encountered a "smart" system, or appliance, or interaction, that has shown AI to be superior to dealing with a human being? When it comes to telephone answering systems, for example, I'm finding it more and more difficult to call a company for support and get the assistance I need, because the AI "front-end" to their calling tree actively tries to prevent me speaking to a person, and instead tries to force me to interact with its limited and all-too-often inadequate menu. So far, I'd say my interactions with AI have been more negative than positive.
How about you, readers?
Peter
Well, I got mixed results when using AI, depending on the quality of the model, but generally speaking I can see how AI can be a useful tool in most jobs.
ReplyDeleteHowever, I think the industry is going WAAAY overboard in thinking that replacing most human employees with AI is a great idea. Generally speaking an AI is only as good as the model it relies upon, and even then, the scope of actions it may take is limited due to the nature of current AI systems, which is constrained due to the incompleteness of such systems (as proven by Gödel a century years ago). There's a great interview with Sir Roger Penrose, where he explains why incompleteness is such a limitation and why human factor is still going to be the "competitive advantage" in any AI-supported processes...
Given the way education has been dumbed down, AI may be arriving just in time to help add 2+2 and other mundane tasks not being taught anymore. AI will help keep the lights on and the plumbing working.
ReplyDelete"Now Chad, see the white thingy? That's the one. Now put it in the toilet tank. No- you have it upside down..."
The only AI experience I've had that is better then human, is internet searches. I use Perplexity to do research and it works much better than conventual search engines. I despise phone ai as you pointed out. Currently trying copilot in excel with mixed results.
ReplyDeleteHad to trash a bunch of code from a team that used Copilot. The code was functionally correct, but from a security perspective was something like "open doors policy"...
DeleteDon't trust AI in critical systems without checking everything at least twice!
Both. At the hospital, AI sees trends and alerts us when a patient meets criteria. The alert is called a "best practices advisory," but humans with human judgment still have the final say. Still, you may get an alert saying that your patient is meeting the criteria for sepsis, and it urges you to take action. I call that a positive.
ReplyDeleteOn the negative side, I had a system try to tell me that the drug concentration of a bag of phenobarbital was 130mg/ml. The bag contained 1300mg of phenobarbital in 100 ml of IV fluid. That doesn't math, as the computer was claiming a concentration that was 10 times what it actually was.
I don't think that we are there yet. I don't think that AI is ready to take over- there are still errors thanks to GIGO, and AI simply allows the mistakes to happen at computer speeds while encouraging humans to stop critically looking at the data- it's making people complacent and lazy.
Precisely. Discernment is still vital. AI has uses, but it remains unthinking.
DeleteI use it to create business process flow maps for SAP system integration and it cuts out 90% of the work. It builds a product close enough that you can finish it easily with some simple edits.
ReplyDeleteI've found some uses for it. But as I wrote on my substack not enough to warrant paying much for the service.
ReplyDeleteI get maybe $20/month value out of it. It probably costs the AI companies $40/month or more to provide it though
I have not been affected by AI one way or the other. As far as I know. I'm retired, so there is no employment requirement, although I am pretty sure the huge multinational corporation I last worked for has done so by now. I don't feel the need to use AI to write, having a good old-fashioned education. As in, I learned how to write well, and have a large vocabulary from extensive reading. Time will tell, but I don't anticipate much interest on my part.
ReplyDeleteTHey said the same thing about machines on assebly lines, and about the first computers:
ReplyDelete"Yer gonna either build em, fix, em, program them, ....or push a broom around 'em."
And while there was some truth to that, people adapted in the workplace. The machines changed things, but so did the people doing the jobs adjust to the changes.
The same thing will happen to AI. Someone will still do the wiring, someone will still do the welding, someone will still conceive of the bridge that the AI will do the grunt work on designing.
Ai will change the workplace, but not all that much. Lower level employees may or may not become obsolete.
The google AI that pops up with any generic search with a quick summary is mostly useful in the general sense, as in it's generally correct but not always very detailed. I have a boss from Germany and he doesn't always understand our Texas ways, so I've been known to cut-n-paste the google AI summary on various topics and send to him. The problem occurs when I need an exact detail on something obscure - the AI is no help at all in that situation. Also, I don't trust it with anything political as that's easily manipulated.
ReplyDeleteAs for the medical, I generally do the base research already and know what's wrong with me before I ever make a doctor's appt. I'm only going in so he/she can prescribe what I know I need. So AI in the medical field is acceptable to me as long as a REAL doctor can override the AI decision in border cases.
I don't see AI as the answer to everything. I think after the initial flurry of EVERYTHING MUST BE AI we'll probably scale back to something reasonable.
My own thought is that AI as current enables people to research things more thoroughly and more widely. It's still up to you to check the output, verify the references etc.
ReplyDeleteI'm a physicist/engineer working in healthcare, and will be making the AI plunge in a week or two. I got the idea from a friend and colleague who showed me what he's been getting done with a paid version of ChatGPT- effectively the AI is the assistant he needs but that his employer won't provide him, and for now it's only about $20 a month. The demo that sold me was using AI to go through a stack of 50 resumes and job applications for a technical position. You describe the problem to the AI what you want to do, in this case apparently the AI generated some python code to extract the desired info from a bunch of PDF file resumes (you drag and drop the resume files into the AI dialog window to operate on them) and then sorted and did evaluations based on the things he told it - ie educational level, experience. You have to tell it what you want and it probably takes some iterations to home in on what you want. But the AI doesn't get tired and with dozens or more of things to go through, it will do all of them more consistently than a human will, not to mention much much faster. I have a similar thing coming up this Fall and will give it a try, for $20 a month it's well worth it and to be honest I won't learn how to use AI without trying something like this.
ReplyDeleteMy friend did this to start but now he's finding all kinds of other applications for the AI as well, like evaluating his fitness and exercise routines and suggesting changes for various goals. He says so far it's working out really well.
I refuse to call it AI - I usually refer to it as an LLM or an overgrown chatbot. Yes, it can do some basic tasks for me, but all it is really doing is running a Google search faster than I can. I consistently find errors in "AI" summaries of searches and documents, we already know that art, music, and books 'created' with these tools are trash. One major company has been subpoenaed by the Justice Dept regarding their "AI" recruiting tool that consistently discriminates based on DEI/AA checkboxes.
ReplyDeleteIt might take me longer to find the data you want, but at least I will make sure it IS the data you want.
My husband uses it for finding the right parts, or repairs. The AI has given him direct links to videos, with correct time stamp where to start, instead of him spending an afternoon searching through websites or videos himself. It’s given him the correct part number, dimensions, for our 40 yr old plow truck. It provides links for any question he has, and saves a lot of searching on his own.
ReplyDeleteSouthern NH
I have a family member with a bunch of different medical scripts and we had an AI evaluate their interactions. Found some interesting trends that explained some symptoms they were experiencing. Yes, could have found them ourselves but would have required wading through ALL the safety data sheets and related pharmaceutical information. I've started calling it the 'data digester' ....
ReplyDeleteWork's started pushing us to try it. I've had poor results but I can see it maybe working better for people who write reports or something instead of my workflow. In one of the trainings, the trainer said:
ReplyDelete"We want you to use AI to get 80% of the way there and clean it up. We want to change your work from a creative task to an editing task which is faster and easier."
I am a Google API and Mapbox hobbyist, creating maps for trip journals that were all HTML based and now being converted to WordPress pages. If it were not for the free ChatGPT assistance that I have received I would not have been able to created the two journal pages with Mapbox maps on WP that I have done so far. I received much better coding assistance from ChatGPT that I could get from the Mapbox AI Assistant. Brave browser AI is helpful to do internet searches, better than the Brave Search feature.
ReplyDeleteDaughter sinus's are filled to the brim with polyps and some are pressing on the brain. Took her in yesterday to get first shot in nose. Ugh! She has to have them spaced precisely so many days apart. Well, best laid plans, doc come in and says sorry, one extra month for surgery due to receiving an AI machine for much better placement of polyp's to see every "whichway" the knife needs to go in ml with being able to remove the one's pressing against the brain. So, AI to the rescue. A doc is coming in with the machine to calibrate daughters brain vs polyps. That's what a person with no medical experience understood. Polyps bad, brain good.
ReplyDeleteI work in civil construction (roads, airports, etc.), haven't seen a use for it in my field yet. I've heard from folks in the medical field reporting similar things to what the folks here have said, same with using it to replace search engines. Other then that the only uses I've seen are like Mikey above, what he wrote appears to be in English but I have no idea what he said. (-:
ReplyDeleteI am using ChatGPT as a software assistant. It is like having an intern in the office, one who never interrupts, but who has memorized all of the manuals. Saves me a lot of time. But there is no way it could write the code I am writing.
ReplyDeleteI gave it a serious problem once: Given a noun phrase, return that phrase with the correct indefinite article; e.g. "dog" -> "a dog". First it gave me the stupid simple solution of testing whether the first character is a vowel. Nope, try again; "unicorn" --> "a unicorn". Next it gave a slightly less stupid solution involving the first two charcters. Finally, it hallucinated a solution involving an IBM software library which does exist, but which does not have the APIs in its "solution".
It can do a lot of grunt work. Bad news for grunt programmers, I guess.
The government neds to ban AI or give us all UBI checks of 4k a month. Time to choose gov't dudes.
ReplyDeleteA friend who farms in Nebraska said they had tractors going walk-about when GPS went down. The kid "driving" the tractor was either talking with his girl on his phone or asleep in the cab.
ReplyDeleteA car hitting a farm tractor is going to lose every time.
Pilot-less tractors are a non-starter unless they have backup guidance systems.
I'm retired as an HR exec. I saw an article yesterday that says that IBM is using AI to replace most of its HR function. For rote work it makes a lot of sense. Most employees do their own data entry and AI can flag outliers for human review. Discipline, promotions, and grievances will remain Human for legal protection purposes. My personal experience is only to use AI as a better way to do an internet search; I find that I can focus my request better. However I have literally been scolded by A.I. for being a racist: I asked: "Do Haitians eat cats?" I had to complain to the A.I. that I only wanted objective data - no opinions about why I wanted the information. By the way, the A.I. "apologized" and THEN coughed up that Haitians DO eat cats but only some Haitians, and only around Christmas.
ReplyDelete