Analog Kid
macrumors G3
The people who think they can put their head in the sand and hope it goes away are in for the most pain.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Artificial Intelligence In Our Lives
- Thread starter Huntn
- Start date
- Sort by reaction score
Luftkopf
macrumors 65816
Yes and no. Middle management should be scared, and people in IT need to start learning quickly, but there are a great number of fields in which AI is pointless.
AI isn't the messiah come to free us from drudgery. It's going to disrupt IT and management, and other information heavy roles, then it will settle down, but not before ruining the lives of many.
Exactly !
To whit, I'm not quite sure that the seniority might become the line on which employability might hinge. Juniors, unburdened by archaic processes and pre-ai scoping mentality might have a leg up.
Some orgs are already shedding senior layers in preference for junior headcount.
The experience premium in specialization along language lines has already all but dissolved. Whereas knowing what to build (that is having an embodied understanding the business context might become an enabler).
I do believe in a blurring of the lines between (PrM , TPM, SWE) but only for those that have an innate understanding of the new work requirements.
We (as in the market) are actively working to make all back office functions automated, and I do expect software engineers to fare much better than the telephone secretaries in re-skilling.
The layer of the infra on which the value will materialize seems to be unclear. Power seems to be a good candidate, HW is just bottlenecked, to see if the value remains once production catches up (see solar panels and china).
Models seems to be vulnerable to distillation and the gap between closed source and open source is closing fast (although the close source providers seem to want to sell you State of the Art as the absolute premium. Harnesses ... hard to know the prevailing paradigm has changed at least 5 times in the last 4 years.
Time will tell, but I am definitely keeping my savings rate much higher than before.
Good luck everyone, take time to enjoy your loved ones in turbulent times, community is the ressource.
Last edited:
throAU
macrumors G4
Anyone in content production needs to figure out an exit strategy or at least a pivot towards direction because you can already do proper ads and decent quality short movie content for pennies on the dollar vs. getting real actors and sets for it.
It's not just IT, and coding.
Luftkopf
macrumors 65816
It's going to be interesting how the market accepts AI content. So far, it's being rejected as slop. I think I read the other day that AI submissions make up 30% of new Apple Music submissions, but they just aren't being listened to.
Personally I think AI can excel as a suite of technologies to make better tools for artists to use. As a generator of end product though. I’m not buying it, and the pushback from regular people is getting stronger.
Pacific1972
macrumors regular
I think 95% of people have no idea just how far AI has actually come today (and that is most dangerous)...
Development is advancing at an ever-faster pace, and AI can now instantly provide answers to questions
that would take humans days to think through - even on topics that are normally quite simple !
When I read that in Singapore, 9-year-olds are already being trained to use computers so that AI can verify
question-and-answer results - because in that country, many jobs are set to be replaced by AI (resulting in
a loss of prosperity for thousands of people as well as a loss of tax revenue for the state of Singapore) -
then this will apply (early or later) to all nations of the world !
AI will not bring prosperity to everyone (that's the biggest lie) - rather, it will lead to a redistribution
of wealth from the bottom to the top - humans will no longer be the first choice when it comes to job
allocation, but only the second choice !
Last edited:
eltoslightfoot
macrumors 68040
Everyone that says this cannot show clear evidence of a qualitative change in LLMs. All they show is quantitative improvements, per se. It is just refinement. My biggest fear isn't even the impact of AI, it's the powers-that-be thinking it can have more of an impact than it possibly can. In other words, it's the difference between laying off 25% of your developers and having the rest use AI for certain things, while being careful not to add vibe coding debt, etc., and laying off 100% of your developers and saying the subject matter analysts can do it all with vibe coding because they know everything about how our processes work anyway.
Then when tokens start going up 11x overnight (already happened) and keep on increasing, suddenly they need those people back.
Look at it this way. No company has ever made money selling their LLMs. Not Google, not Microsoft, not Apple, not OpenAI (definitely not them. They are buy out candidate within 6 months), not Meta, not xAI, and not Anthropic.
After some major f-ups, everyone is going to realize that AI can't be held responsible for anything like a human worker, and that, combined with ever present drop in quality, will make everyone realize what a study in radiology showed years ago.
AI combined with a radiologist (with the radiologist in charge of the AI) found more instances of cancer than either could alone.
The other problem here is that the difference between models isn't stark enough to ever charge enough per token to justify the cost of AI (and that's without getting into power or water costs). In other words, say Anthropic 10x (again) their token costs. Everyone just switches to OpenAI. Then OpenAI switches, and people start figuring out how to run Deepseek on their local computers.
Eventually these costs will rise to where profits need to be made and everyone will realize they can only afford to use AI for actual important things--such as figuring out what compounds could make new antibiotics, etc.,
And as far as entertainment, everyone wants uniqueness and care. They do. not. want. AI slop. It will eventually be used in small ways to speed up digital work, but that's it. We want connections in our art, and AI has none.
Edited to add: So it appears I was incorrect about that oft cited radiology study. Here is an interesting follow up:
Why Radiology AI Didn’t Work and What Comes Next | Out-Of-Pocket
FDA Approvals, PACS Pain, Selling Point Solutions, and Why Radiology AI Needed a Rethink
www.outofpocket.health
BSDnostalgia
macrumors 6502
Thank you for this concise description. Unfortunately, it fits my preconceptions perfectly.
My thought exactly. In engineering, in general, there is a phase where people learn job-related reality by starting at the junior level and working your way up. In traditional engineering, someone becomes hire-able when they contribute more to the mid- and senior-level engineers than the time they use up by being a beginner. Traditionally, that was at, say, the BS or MS level.
It isn't clear that there will exist a practical educational path for those coming up. Basically, what you experts are saying is that you already have to know how to do, e.g. "test definition, test code, verification test and documentation", so that the AI agent can do this for you, without your actually ever having had to do these tasks. It appears that AI is creating a situation where only mid- and senior-level engineers are needed and hireable, without an educational path for people to get to that level.
And, it appears that AI-enabled engineering is somewhat better than, say, law, where AI may pretty much abolish the need for assistants. So, where are the entry-level jobs going to come from?
eltoslightfoot
macrumors 68040
This is already backfiring and changing. Token costs are going up so much, junior devs are cheaper.
skodises
Contributor
The legal profession is experiencing a great deal of trouble in the use of AI. The issue is that the use of AI to research case law for precedent, or prior art in patent prosecution, is fraught with peril: hallucinations are all too common, and reviewing the output to separate the real from the slop is absolutely critical. The courts have demonstrated very little patience with counsel that carelessly bring slop into the proceedings, and rightly so.
AI absolutely can produce ungodly amounts of research output, arguably much more than the average team of attorneys and paralegals, but culling the slop requires very detailed checking line-by-line. And that will continue to take trained, knowledgeable, and experienced legal talent. I don’t see that changing soon…
I’ve served as an expert witness in patent prosecutions as a side gig for some years now, and the legal teams I’ve worked with have some deep concerns here, and well they should. Here’s a quick taste of some of it, for the interested student:
AI on Trial: Legal Models Hallucinate in 1 out of 6 (or More) Benchmarking Queries | Stanford HAI
A new study reveals the need for benchmarking and public evaluations of AI tools in law.
A legal practitioner’s guide to AI & hallucinations
Using AI carries both responsibilities and risks for legal professionals. Understand how generative AI works, what it does and does not do well, and how to use it responsibly.
www.ncsc.org
AI Hallucination Cases Database – Damien Charlotin
The most comprehensive database of AI hallucination cases in law: legal decisions from courts worldwide, searchable by country, party, AI tool, and outcome. Updated daily.
www.damiencharlotin.com
Last edited:
Pacific1972
macrumors regular
Everyone that says this cannot show clear evidence of a qualitative change in LLMs. All they show is quantitative improvements, per se. It is just refinement. My biggest fear isn't even the impact of AI, it's the powers-that-be thinking it can have more of an impact than it possibly can. In other words, it's the difference between laying off 25% of your developers and having the rest use AI for certain things, while being careful not to add vibe coding debt, etc., and laying off 100% of your developers and saying the subject matter analysts can do it all with vibe coding because they know everything about how our processes work anyway.
Then when tokens start going up 11x overnight (already happened) and keep on increasing, suddenly they need those people back.
Look at it this way. No company has ever made money selling their LLMs. Not Google, not Microsoft, not Apple, not OpenAI (definitely not them. They are buy out candidate within 6 months), not Meta, not xAI, and not Anthropic.
After some major f-ups, everyone is going to realize that AI can't be held responsible for anything like a human worker, and that, combined with ever present drop in quality, will make everyone realize what a study in radiology showed years ago.
AI combined with a radiologist (with the radiologist in charge of the AI) found more instances of cancer than either could alone.
The other problem here is that the difference between models isn't stark enough to ever charge enough per token to justify the cost of AI (and that's without getting into power or water costs). In other words, say Anthropic 10x (again) their token costs. Everyone just switches to OpenAI. Then OpenAI switches, and people start figuring out how to run Deepseek on their local computers.
Eventually these costs will rise to where profits need to be made and everyone will realize they can only afford to use AI for actual important things--such as figuring out what compounds could make new antibiotics, etc.,
And as far as entertainment, everyone wants uniqueness and care. They do. not. want. AI slop. It will eventually be used in small ways to speed up digital work, but that's it. We want connections in our art, and AI has none.
Edited to add: So it appears I was incorrect about that oft cited radiology study. Here is an interesting follow up:
Why Radiology AI Didn’t Work and What Comes Next | Out-Of-Pocket
FDA Approvals, PACS Pain, Selling Point Solutions, and Why Radiology AI Needed a Rethink
I have a BROTHER-IN-LAW who works for a Chinese B2B company in Germany.
He is highly skilled (in his field) and, after 15 years, he is considering leaving the company and
his “secure” job because his Chinese B2B company doesn’t offer any guarantees or warranties
in Europe for AI graphics cards that cost business customers tens of thousands of dollars - and
he lost arguments for B2B sale (items perhaps also overpriced) - AliExpress Mindset !
But when he says, “I'm good in my job and could even sell Donald Trump an AI graphics card
with 24GB or more because I'm no longer afraid (of high-rise-persons) to take risks !” 🤪
I can only say that he's an 😉 IDIOT (multiple) 'cause he understands nothing about life !!!
I look at what he’s achieved in life — divorce, debt, and child support payments to the children
he fathered. His only lifeline at 50 is the prospect of retiring in another 15 years to what will
hopefully be a government-guaranteed pension 🙄😬🤮 !
Anyone who can't take care of themselves and doesn't make things a little easier for their own
children has already lost the race against AI, because the level of education for the most at 50
(+ muddy mindset) isn't the same as it was at 25 - Game Over !
"I know that I don't know !"
Last edited:
skodises
Contributor
And then, of course, there is this nice, cautionary tale for the excessively lazy “content creator”. Hoist with his own petard, this one.
www.nytimes.com
Book on Truth in the Age of A.I. Contains Quotes Made Up by A.I.
Steven Rosenbaum, author of “The Future of Truth,” said he had started his own investigation after The New York Times asked about the fake quotes.
Last edited:
eltoslightfoot
macrumors 68040
throAU
macrumors G4
You can parrot that line, but the reality is that real businesses are using AI to solve real problems today. doing things that were not possible 2-3 years ago.
I know this because I'm both working and and next door to companies that are doing this. Today.
Last edited:
Analog Kid
macrumors G3
You are holding to the mindset that junior level means transactional grunt work. It’s that way today because there is a lot of transactional grunt work to be done so it goes to those with least seniority, power and connections. There’s no reason that has to be true in a world where a lot of the drudgery is automated away. Vision doesn’t necessarily come from experience. There are a lot of junior employees who have extraordinary vision but are delayed in showing it because they need make their way out of the boiler room first.
Take a step back and realize how paradoxical it is to say only senior people are hireable. Where are the entry level jobs going to come from? Where they always do: a need to meet demand in a world where senior people age out.
A reality check on the AI jobs hysteria
What do the numbers really say about the impact of artificial intelligence on the labor market? The answer might surprise you.
Mousse
macrumors 601
In the early days of personal computing...
A.I. is a force multiplier that can make our mistake 10,000x worse. Maybe in 20-30 years A.I. will become useful. Right now, I trust Al [Bundy] over AI.😏
eltoslightfoot
macrumors 68040
Luftkopf
macrumors 65816
And further to your point, I have always hated these LLMs, and the nefarious way most of them have been trained. I called them ‘plagiarism engines’ since I learned of them. But I recognised that they’re not going away, big business desperately wants them to succeed, and investors are willing to burn stupid piles of cash in hope. So I figured it was important that I learned to use them so my opinion would be more informed.
Now I have a tiny daughter, I find it even more important to learn this new world of BS, so I can successfully teach her to navigate the mess as she grows up.
The only AI product that has shifted my opinion in a positive direction is Perplexity, which gave me much better search results than Google has done for a long time. I had two assignments to research for with very similar subject matter, I needed to search for the same things, but with a different focus. It took me ages to turn up relevant links with Google. Dozens of search terms, a couple of days reading through obscure websites, before I had enough. Perplexity turned up better results when I started using it on the second assignment. I had all the information I need in a morning of reading the links it provided!
I seldom trust the answer Perplexity writes though. It too often has little errors.
My free year of Perplexity is about to end though, and I don’t know what the free tier is like, so I’ve started experimenting with getting something similar run locally. I use Ollama + Chatbox with an Ollama free API for web search capabilities. Using Ministral 3 8B, it’s been ok, it’s been fun as a nerd to try different offline LLMs, but they’re of little practical use.
In addition to that, I have used chatGPT, Claude, Mistral’s Le Chat, Gemini, a Meta’ AI. Both online, and their local models that fit in my 18GB MacBook Pro.
I have not seen anything that indicates that these are any form of tech messiah come to make life better for all. They’re flawed technology that big business is desperately trying to funnel us into. A LOT of people hate AI chatbots being wedged in everywhere, so the industry response is to wedge more of them in, and tell us we’ll love them this time.
I can see that these tools have good uses in certain circumstances, but they’ve been only borderline useful to me. Aside from Perplexity, the only other scenario AI has been helpful to me has been when I’m stuck on an assignment, and have run out of ideas, I’ll query the AI about the question I’m stuck on. The reply is essentially useless, but it will have one or two ideas that will give me an idea, enough to get me unstuck.
Hardly a compelling argument for generative AI.
Sorry for this wall of text. My daughter is 5 months old and I’m a bit delirious.
TLDR: I haven’t buried my head in the sand, my negative opinion has been reinforced by hands-on experience.
eltoslightfoot
macrumors 68040
Exactly. So has mine (come from negative hands-on experience). Ever notice that most of the really pro-AI folks kind of sound like desperate salespeople? Okay, then leave me behind. Again, I use it within specific constraints and for specific reasons.
But I see way too many young people that are literally abdicating all thought over to AI chatbots. Like they literally have to ask their AI everything. At the rate they are going, they won't even be able to do creative prompting in a few years. It'll be a prompt to get a better prompt. It's getting gnarly out there.
My kids are going to own the world someday because they will be creative without AI, and only use it for specific reasons. I would like the think they won't be alone though. I would like to think of the examples of graduation speakers all getting booed for mentioning AI. AI is a solution in search of a problem.
turbineseaplane
macrumors Penryn
And this is on top of what is already an insane wealth gap right now.
turbineseaplane
macrumors Penryn
It's sadly quite true, and we are already seeing early data supporting the "brain drain" from becoming dependent upon it.
Luftkopf
macrumors 65816
My nephew is 16, and heaps into coding. He has an assignment at school to make a little app. He has his idea, but there wasn’t enough time, so he asked his teacher for time, but was told to use AI. He did, and it worked, but he was so mad about it! He has a good rant to me about how bad he felt doing it because he didn’t learn anything from it!
He’s going to be one of the kids that grows up and has actually learned to think. He’s going to be ok.
turbineseaplane
macrumors Penryn
Sort of flies in the face of the purpose of "school".
Clearly this moment we're in can't last like this.
This is nonsensical to have students have AI "do stuff" for assignments and then the students don't learn anything.
It's basically like having another student just do your work for you.
Luftkopf
macrumors 65816
Yeah, it is totally stupid. He was so mad about it, he just wants to learn more! He’ll be right though. He’s got the right attitude.
Something will eventually have to give. Hopefully it happens before it takes out too many people!