In news to no one who actually works: AI produces nothing of economic value

[cockerpunk]

https://www.economist.com/finance-and-economics/2024/07/02/what-happened-to-the-artificial-intelligence-revolution?utm_medium=social-media.content.np&utm_source=facebook&utm_campaign=editorial-social&utm_content=discovery.content&fbclid=IwZXh0bgNhZW0CMTAAAR1zQVEGczqtJu3JKPqOLG5Cb 8ia9vtRnV4bTydgwg2NaCGxDb6K7g3Arn0_aem_ySBLxvNP3Me tppeDqSyW6w

[Arizona_928]

Neither does Google….

lol

[MikeSid]

Well at least AI is generating revenue for the Economist for the paywall that article is behind.

But on a slightly more serious note, I seem to recall the same sentiment in the early days of the World Wide Web. In the 60's and 70's the internet (as it was) was only used by military, governments, and serious scientists. But when the technology filtered down to HTTP/WWW, usage skyrocketed and a variety of new industries formed, not least of which is the ability to order pretty much anything for delivery to your doorstep.

I believe AI will have the same sort of impact, if not greater, we just don't know what it will be. We're probably 20 years from seeing what the real impact will be.

[Ayles]

I have been using it to generate code snippets for both SQL and R languages and while it's not perfect it does give some great directional recommendations.

[zakthor]

There’s dumb ai like chat gpt but there’s useful stuff too. You can deep learn so much these days and end up with a model instantly. It’s like the next level of spreadsheet.

Code? Nope it won’t do code. It looks like it can write code but so far it’s done a super ****ty job at anything I’ve asked it for. I’d say ‘high functioning hs intern’ which is to say pretty much a complete waste of time.

[Arizona_928]

Are you paying for the premium gpt?

[RANDY P]

[/sarcasm]Brilliant article [/Sarcasm] That's because they haven't STARTED using AI to replace human decision makers and intellectual workers yet. They're starting with burger flippers in CA. Next, coders, admin and recordkeepers.

rjp

[zakthor]

Originally Posted by Arizona_928

Are you paying for the premium gpt?

Yes.

I’ve heard that so many times, folks think it works for code. I suppose it depends on whether you write code that others have written before.

Ultimately I’m responsible for my code, it needs to work the way I expect and I need to know that it is correct. You need to know enough to spot it’s mistakes but when you ask it to correct them it introduces others. Clear as day the thing has no understanding and it’s totally irresponsible to let it into production without more than the usual code reviews.

Imagine you’re asking it to design a high voltage radio that will kill if there’s a boo boo. The design is simple but what must be done to ensure it is safe? Otherwise you’re tossing pasta on the wall to see it stick.

I can’t make a single request where it doesn’t make a mistake. Go ahead and ask it to prove a variant of p == np. It’lol make pages of proof before making a magic leap. Does same exact thing with its code.

Ask it to make a c++ template that builds a static const hash table of strings, or convert date to Julian microseconds using std::time. It’s code might compile but it’s still broken.

[cockerpunk]

Originally Posted by RANDY P

[/sarcasm]Brilliant article [/Sarcasm] That's because they haven't STARTED using AI to replace human decision makers and intellectual workers yet. They're starting with burger flippers in CA. Next, coders, admin and recordkeepers.

rjp

AI is not replacing burger flippers.

this whole post is just so out of touch and wrong.

[cockerpunk]

Originally Posted by zakthor

Yes.

I’ve heard that so many times, folks think it works for code. I suppose it depends on whether you write code that others have written before.

Ultimately I’m responsible for my code, it needs to work the way I expect and I need to know that it is correct. You need to know enough to spot it’s mistakes but when you ask it to correct them it introduces others. Clear as day the thing has no understanding and it’s totally irresponsible to let it into production without more than the usual code reviews.

Imagine you’re asking it to design a high voltage radio that will kill if there’s a boo boo. The design is simple but what must be done to ensure it is safe? Otherwise you’re tossing pasta on the wall to see it stick.

I can’t make a single request where it doesn’t make a mistake. Go ahead and ask it to prove a variant of p == np. It’lol make pages of proof before making a magic leap. Does same exact thing with its code.

Ask it to make a c++ template that builds a static const hash table of strings, or convert date to Julian microseconds using std::time. It’s code might compile but it’s still broken.

yup, you need to know more than the AI, because you have to debug it.

the only folks who think AI can code, are ones who think 'coding' means actually typing code. anyone who has written any code, knows that 5% of your time or less is actually writing code.

and thus refers to the title ... anyone who actually works, knows that AI isnt going to produce jack **** in value. only folks who dont work, seem to think AI is going to do something magically.

[jyl]

AI will be great for customer service. It doesn’t need to produce economic value, just give people the runaround.

Seriously, I think half of all contact center employees will lose their jobs in the next five years, and eventually maybe the great majority. Contact centers employ about 3MM in the US (and about 350,000 in India and a similar number in the Phillipines).

More generally, I think functions that involve having verbal interactions (voice or text or video), answering questions, providing information, where the stakes are low and mistakes can be tolerated, will go to AI. Like, I would hope to never call 911 and talk to an AI, and I hope a pilot never radios the flight controller and talks to an AI, but a consumer with a warranty question? No expensive human for you! Making minor, low-stakes business decisions can also go to AI (if an airline occasionally lets customers change a flight for free or gives an upgrade when it didn’t need to, that’s acceptable; if it occasionally overcharges for that when it shouldn’t, that’s also acceptable - to the airline).

[URY914]

I heard a story this morning on the radio about an AI engineer that has made a AI clone of his mother. His DEAD mother. He can talk to her as if she was alive and she responds like a normal conversation. So weird.

[pavulon]

A lot of radiologists are rightfully worried about being replaced.

[cockerpunk]

Originally Posted by pavulon

A lot of radiologists are rightfully worried about being replaced.

they shouldnt be.

imagine what **** an AI would find in a print when it cant even make fingers properly. lol.

[1990C4S]

Originally Posted by pavulon

A lot of radiologists are rightfully worried about being replaced.

I think 'cross-checked' more than replaced. It's a slippery legal slope. Radiology errors lead to a lot of lawsuits, the move to AI will be slow.

I also think that the reports of AI 'radiology success' are biased and wildly optimistic.

[pavulon]

make fingers properly?

They're nervous because computers have been used to screen medical images for some time already.

https://www.cancer.gov/news-events/press-releases/2020/automated-dual-stain-cervical

Originally Posted by cockerpunk

they shouldnt be.

imagine what **** an AI would find in a print when it cant even make fingers properly. lol.

[RANDY P]

Originally Posted by cockerpunk

AI is not replacing burger flippers.

this whole post is just so out of touch and wrong.

Yes Punk, things will never change. They'll never outsource you to a machine (who can spell properly)

LOL, says the boy who quoted an article that is so premature it's laughable.

rjp

[RANDY P]

Originally Posted by jyl

AI will be great for customer service. It doesn’t need to produce economic value, just give people the runaround.

Seriously, I think half of all contact center employees will lose their jobs in the next five years, and eventually maybe the great majority. Contact centers employ about 3MM in the US (and about 350,000 in India and a similar number in the Phillipines).

More generally, I think functions that involve having verbal interactions (voice or text or video), answering questions, providing information, where the stakes are low and mistakes can be tolerated, will go to AI. Like, I would hope to never call 911 and talk to an AI, and I hope a pilot never radios the flight controller and talks to an AI, but a consumer with a warranty question? No expensive human for you! Making minor, low-stakes business decisions can also go to AI (if an airline occasionally lets customers change a flight for free or gives an upgrade when it didn’t need to, that’s acceptable; if it occasionally overcharges for that when it shouldn’t, that’s also acceptable - to the airline).

I just hope it's better than their 1st shot at automation- the automated attendant. You know, the one everyone cusses at.

rjp

[GH85Carrera]

Originally Posted by pavulon

A lot of radiologists are rightfully worried about being replaced.

So who do the patients sue for a bad diagnosis from AI?

One of my friends is a radiologist. He said he get about 30 bucks to look at an x-ray from a medicare patient. From that he has to pay his billing agency, and his malpractice insurance. So if he misses a minor detail, he gets sued for millions for doing a job that pays him minimum wage or less.

[Arizona_928]

Originally Posted by GH85Carrera

So who do the patients sue for a bad diagnosis from AI?

One of my friends is a radiologist. He said he get about 30 bucks to look at an x-ray from a medicare patient. From that he has to pay his billing agency, and his malpractice insurance. So if he misses a minor detail, he gets sued for millions for doing a job that pays him minimum wage or less.

Honestly, it might not be a bad thing. Since DEI has infected healthcare… these incompetent providers need all the help they can get. E.g. va sent me out to a neurosurgeon (care in the community) for broken pedicle screws. The provider couldn’t read a MRI, ct or x rays. When i showed her in the MRI where they were broken, she didn’t believe me and kicked me out of the office. Wrote a very nasty medical note for my records indicating that i was malingering.

Ignorant at best. Incompetent more than likely.