I already do this for my weekly reports to upper management. I input what I did this week and run it thru ChatGPT and let it spit out what it does. I then pull out some of the things I like and input it back into my report. It’s better at making a dull report more interesting than I can at this point since I don’t really care that much anymore. And its fun to see what it comes back with.
Sounds a lot like a No True Scotsman. Certainly, there are tens of thousands of journalists in a plethora of fields where that isn’t the case. While you can, for example, be the type of sports journalist who travels with the team, watches every game in person, regularly interviews players, coaches and front office folks, and has notes from a 30-year career of games attended professionally. That’s not the only definition of a sport journalist and I would say that most people today have come to the conclusion that these old school beat reporters are inherently flawed in some critical ways and that writers assessing a team from a distance tend to be more adaptable, insightful, balanced and importantly, more critical.
You’ve described one (in some ways antiquated) flavor of journalist. It’s not fair or accurate to imply that these are the traits of a “real” journalist.
Note I said “a lot of journalism” not all.
They may not be able to beat the best(yet), but in the name of profit they can certainly sideline them.
I take it you are a journalist.
Twenty years ago when I first joined the Dope I was a photojournalist working at an English-language business weekly in Hungary. My experience was mostly business and sports (two fields that an AI could have a reasonable time replacing journalists.) I’ve long since left the journalism world, and do other types of photography for more money and fewer headaches. I still have a good number of friends in the industry.
Cool.
The factual journalist in me should correct that to “almost twenty two years ago” (can’t believe it’s been that long.) And I also did a bunch of feature-type stuff for automobile publications like Car & Driver and Top Gear. Those were my main areas, though I’ve done a bit of general news, as well.
And the engineers and those with other product-related knowledge will be close behind those managers on their way out the door being replaced by AI. Not all engineers, obviously, but a well trained AI could make a single engineer able to do the same work as an entire department can now. Similar to how in the 80’s, once someone loaded Lotus 1,2,3 onto their new desktop computer, they found that entire floors worth of accountants were now redundant.
It will be a while before journalists are replaced entirely by AI, but with AI assistance, I can see a single journalist doing the work of their entire team.
Keep in mind that ChatGPT is a prototype, a toy for the public to play with. It is not the final product. When this is monetized and sold, it will almost certainly be able to research up to date information about a subject, remember conversations with a user indefinitely, and be trained on data sets that are more relevant to the tasks it will be expected to perform.
If you fed it the entirety of Cecil’s work as a data set, and asked it to create a new article about a subject that it was able to research, it may be the case that we as readers wouldn’t be able to tell the difference.
AI is coming for all of your jobs. It won’t be long until the only jobs left are artisanal, supported by those who are willing to pay a premium for a good or service created by a human, even if it is inferior to the product made by the AI.
AI is not coming for blue collar workers. They took their hit in the third industrial revolution. It will be a long, long time before we have robot drywallers and plumbers. We may get to the point where an AI can write a symphony or a novel, but be utterly useless at fixing a leaking faucet. Some blue collar jobs will be aided, some lost. But in general, they aren’t going to be the ones with the greatest impact to their jobs.
Also, AIs can’t go onto customer sites and examine how to automate them. They can’t build bridges, or clean gutters. They can’t replace maids.
This time, the revolution is coming for the chattering classes and the upper middle class. That will make everyone scream much louder than if it was their housekeeper who lost a job.
Robotics is still way behind AI. General purpose robots like ATLAS exist, but are incredibly expensive and have terrible battery life. Prices will come down, but fundamental mechanical, energy and cost issues may not be solved for a long time for the mass market.
The thing accelerating ChatGPT and other LLMs is that there are almost no physical limits. And no distribution problems. Once it became available, everyone could use it right away. And the simple interface and natural language ability means a very shallow learning curve.
If we get useful robots for the home, they still have to be made by the millions to have this kind of impact. That’s a long way away.
What is the fullest potential of a chatbot like ChatGPT?
Am I the only one who thinks of Dr. Know?
~Max
I bet AI will never be able to replace a good old telephone sanitiser.
I doubt we will see many employees on the floors of factories or in the backs of restaurants for much longer. Janitorial services and garbage collection can be mostly automated as well.
What’s a “long time” to you? If you are talking years, then I agree. If you are talking decades, I’m not so sure. And once you have robots with hands like humans, there’s nothing stopping them from fixing that leaky faucet or cleaning gutters. There’s no reason they can’t do your laundry (albeit folding laundry is one of the hardest tasks currently) and vacuum your floors.
While I think you are correct that skilled trades are probably the last to go, I’m thinking that they will only hold out a few decades at most before they get replaced. The very last trade to be replaced will be dog grooming.
Instead of thinking about who these Large Language Model AIs are ‘coming for’, we should also think about who they are going to enable.
In particular, small business people or people who want to start one. Over time, it has gotten increasingly expensive to start a new business. You need lawyers, accountants, specialists in accessibility, all kinds of white collar, very expensive support people. This increases barriers to entry, especially for the poor.
But very soon, people with an idea will be able to ask an AI for advice in how to go to market, an AI to handle the legal paperwork, an AI to do the firm’s accounting and taxes, an AI to do the company’s graphics and mailings and ad work, etc. What might have cost a startup $10,000 in fees to others might be done in the future for $100 of CPU time. And it would eliminate corruption and most HR work.
Future businesses may be very, very light on office workers. We might see many new businesses formed around a small engineering team of high performers with the entire ‘backoffice’ done by AI. This is very democratizing and liberating, especially for people shut out of the startup wor;ld by lack of capital or lack of connections.
Getting back to journalism, consider SubStack. It’s already attracting journalists away from major media. With AI, now you can have good copy editing, you can brainstorm ideas, use the AI as a librarian (in subsequent releases), etc. You don’t need a newsroom and an editor and support staff. This will likely accelerate independent journalism at the expense of the old, big newspapers.
Why does this column show up only in Cecil’s Columns and not in “all categories,” even though last week’s column is pinned there?
I thought I saw it earlier and that my mind was going. Whew. Get to put that off one more day.
Y’know, an AI moderator would default to the correct option.
I don’t know if AI will replace journalists (I suspect it would replace a lot of them, because a lot of news stories are basically formulaic fill-in-the-blanks,) but I’m damn sure it could cut a path of destruction through the field of copywriting. I can see account executives type in a bunch of selling points and seconds later completely acceptable ad copy pops out ready for client approval. Ditto for most speeches. It’s only a matter of time before ArtDirector GPT comes along, analyzes the selling points and generates the proper graphics and font and a complete ad is ready in minutes.