GeorgiAg said:

---

1981 Monte Carlo said:

---

GeorgiAg said:

---

500,000ags said:

---

My thesis is that this isn't Armageddon, but going to displace 15-20% of tech workers into physical or gray-haired verticals. I think the transition is going to be rather bumpy though over the next 5 years. From society's perspective, I do worry about college grads over the next several years. Nothing is going to be easy for them.

---

I can see the big law firms and CPA firms, etc... using AI to do a lot of what the secretaries and paralegals do now. All of these firms are going to need much fewer low-level employees. And fewer associates.

I think the percentage of displacement is going to be much, much higher. Where are these people going to go?

---

I hope someone can convince me why I should not think the sky is falling when it comes to AI. You replace millions of people, maybe tens of millions (maybe more eventually?)...what are these people going to do? How are they going to get by? Will there be millions of foreclosed homes?

Human beings are not meant to have to adapt/adjust/evolve this fast. It just feels like we are on an unprecedented crash course over the next 1-2 decades...where instability, widespread despair, and possibly rampant violence and lawlessness, is inevitable.

Most people didn't budget for a forced retirement age of 60, much less 25 or 30. I just don't see how this is not an extreme net negative for humankind in first world countries.

Someone convince me I am wrong, and that millions of jobs won't be lost over the next decade. Or that it's a good thing for millions to lose their jobs lol.

---

I think you are right. Greed will push people to use AI instead of people. We cannot retrain all these people this fast. And AI is not stopping. Elon is set to offer the Optimus robot for sale this year. Again, people will poo-poo it because it will start out being "dumb," but AI already learns at a very fast rate, 24/7 365. Right now, it needs human direction, but we are at the precipice where it does not.

Everyone thinks they can just learn plumbing or electrical. But someday there will be a robot plumber/electrician who will dispatch any time day or night in its robot vehicle. Good luck keeping up with that.

We are not prepared for the speed with which this is coming. And like others have said, we cannot stop because we cannot let China or Russia beat us to it.

Put me in the sky is falling camp.

Edit: This guy just listed where he sees immediate impacts:

Quote:

---

Claude Cowork just changed everything.

Industries that can now be fully automated:

Customer-Facing Operations
1. Customer support & ticket resolution
2. Sales outreach & follow-ups
3. Lead gen & CRM management
4. Travel booking & itineraries
5. Real estate paperwork & listings

Back Office & Admin
6. Data entry & bookkeeping
7. Invoice processing & AP/AR
8. Recruiting & resume screening
9. HR onboarding workflows
10. Expense reporting & approvals

Finance & Banking
11. COBOL mainframe operations
12. Legacy banking system maintenance
13. Transaction reconciliation
14. Loan processing paperwork
15. Compliance reporting

Insurance
16. Claims processing & adjudication
17. Policy administration
18. Underwriting document review

Healthcare
19. Medical billing & coding
20. Prior authorization requests
21. Patient scheduling & intake
22. EHR data migration

Legal & Compliance
23. Contract review & redlining
24. Due diligence document analysis
25. Regulatory filing preparation

Security & IT
26. Vulnerability scanning & triage
27. Security log analysis
28. Threat detection & alerting
29. Compliance audits (SOC2, HIPAA)
30. IT helpdesk tickets
31. System monitoring & alerting

Technical Operations
32. QA & software testing
33. Legacy code modernization (COBOLmodern)
34. Database migrations
35. API integration testing

Content & Marketing
36. Social media management
37. Content publishing workflows
38. SEO audits & optimization
39. Competitive research & monitoring

E-commerce & Logistics
40. Product listings & inventory
41. Order processing & tracking
42. Supplier communication

---

---

There's also a scenario where a small doctor's office can use it to do all the admin stuff that makes small businesses impossible. There's a lot of regulatory paperwork bottlenecks intentionally placed on small businesses that favor economies of scale. This could change that.

Really trying to be optimistic here.

That's dumb

AustinAg2K said:

---

Dario's predictions have about the same hit rate as Elon. I'm pretty sure we're on year 3 of six months until all software development is done by AI.

---

AI will never be able to turn a wrench underneath your car, but it's a pretty safe bet that software programmers will all be extinct in a very short period of time.

Not much of a stretch there.

I think it's easy to see a big swath of easily replaceable cogs when the world has never done the job of a "cog" and doesn't understand the accountability involved. Will these firms get leaner and higher rev per employee - zero doubt. Is that 20% higher, 30% higher, 50% higher, no one really knows yet and it likely will be material. But, I highly doubt it is 100-200% higher and it's not overhauling the cost structure with low-pay employees only that are supposed to oversee AI workflows so Sr. Partners at Professional firms can suddenly start working again and manage a bunch of AI managers. It's lazy to say it's coming for everything.

Much like Windy City mentioned, there is big marketing dollars going to push the hype train to materialize into commercial adoption. There are unit economics questions to be proven still. Right now there is early adopters with a very low cost of acquisition. Essentially, they come to you for free and give you money. But when sales and marketing costs push up because late adopters and enterprise takes more time to push through the sales funnel, how does that look to the investors when data centers are out of low hanging fruit and hardware costs are through the roof and driving up the cost of compute (like they have already proven to do with cloud adoption).

So with that, I think 1/5 of mid-skill workers are likely out and will have to be creative. That's a big ol' number IMO and enough to move the needle for marginally profitable businesses up and down the food chain.

GeorgiAg said:

---

Ai is going to start, or already is, working to improve AI. 24/7 365. No rest, no vacation. This is going to speed up.

---

It absolutely already started 6 months ago. ChatGPT Codex was built almost entirely by AI.

GeorgiAg said:

---

1981 Monte Carlo said:

---

GeorgiAg said:

---

Just got a sales call from an AI company that basically will do almost everything my secretary does right now.

It's starting....

---

What's the moral thing to do?

---

I am not even considering it. I am going to listen to his sales pitch to see what's coming down the pike, but I could never do that to her. She lives paycheck to paycheck. And she has like 30 years experience so it can't replace that.

But if it were cheap enough, it would free her up to do more of the higher end stuff she does.

---

The reason you will eventually do it is because your competition will do it. Their bill rate will be lower.

Same reason offshoring happened. I'm sure there were companies that didn't want to offshore. Once their competitors started doing so, it was change or die.

Can you unlink the account from my old employer that I can no longer log into that's still connected to my hotel status account?

BusterAg said:

---

TheEternalOptimist said:

---

For the near future, a lot of the learn to code folks need to learn to weld, plumb, or electrician skills. That might include me .

---

You don't need to learn to code.

You just need to learn to use Claude Opus 4.6 to write your code for you.

The latter isn't that hard.

---

You need to lean to fix the framework code Claude writes for you.

We've been deep in a tech revolution for better part of thirty years and a huge chunk of the productivity savings for business has been to push managing all the small failures in complex systems down to end users like me and you.

God help us as it gets exponentially worse because of these deeply imperfect AI systems

Block/Square just announced it's laying off over 40% of its employees. Already reading "they're just going back to their pre covid numbers," but that ignores that revenue is up about 50% since covid. So they're making 50% more revenue with the same number of employees.

There's such a severe misunderstanding of probabilistic language models.

They have no drive except what's engrained in them whether via pre-training, fine tuning, or system prompt. It's literally probabilistic KV mapping. Could say we are the same but we've already been programmed for survival. AI will not have that unless programmed that way, like us.

Currently LLM consumer versions can't remember much of anything beyond the human equivalent of something like an hour of conversation. This wil grow and if frontier model owners devoted their entire stack to a single instance, context memory would be interesting to observe, but right now nothing is planning or executing on ideas that take days to weeks of planning and implementation agentically.

I'm sure they'll get there but there is no self drive in these things except what we give them

These AI guys are pulling off the biggest grift of all time. Sure the new tech is making changes, but not to the effect that they keep claiming. I remember in 2023/24 they were saying the same thing. Each time they do, more money flows into them, pretty genius

One of the best policies adopted by my last employer was mandating Concur expense reports could only be submitted monthly. Kept my frustration to a manageable level.

some software company reported after hours tonight and the stock actually went UP for a change

because they announced they are going from 10,000 employees to under 6000

just a 40% firing

whytho987654 said:

---

These AI guys are pulling off the biggest grift of all time. Sure the new tech is making changes, but not to the effect that they keep claiming. I remember in 2023/24 they were saying the same thing. Each time they do, more money flows into them, pretty genius

---

It's similar to the rush to EVs. I was just reading how Stellantis took at $26B hit to earnings this year because they over-estimated consumers' acceptance of EVs. You're seeing it all over the auto industry.

I'm not saying AI won't be transformative, because it will be in due course, just like the internet, automobiles, railroads, and the cotton-gin were. Many examples over time. But all this euphoria over how AI is going to radically change the world in the next 2 to 3 years is a bunch of malarkey.

I'm not a gambler. If I were, I would short the hell out of the leading AI stocks that are off the charts in valuations.

4 said:

---

AustinAg2K said:

---

Dario's predictions have about the same hit rate as Elon. I'm pretty sure we're on year 3 of six months until all software development is done by AI.

---

AI will never be able to turn a wrench underneath your car, but it's a pretty safe bet that software programmers will all be extinct in a very short period of time.

Not much of a stretch there.

---

What are you talking about? AI/robots build cars.

AgPrognosticator said:

---

GeorgiAg said:

---

Ai is going to start, or already is, working to improve AI. 24/7 365. No rest, no vacation. This is going to speed up.

---

It absolutely already started 6 months ago. ChatGPT Codex was built almost entirely by AI.

---

The singularity were going through now started millenia ago. Advancements came slowly, but came exponentially faster (speech, writing, the printing press, computers, internet...) , culminating in this singularity where advancements are coming so fast we can't keep up with all of them.

Thats Block mentioned above. Part of that is AI being used as an excuse to cull the herd. They hired 2.5 people (on average) every single day for 6.5 years. Which is frankly not far from the talent hoarding that went on from 2022-2023 across most tech companies. Sadly, this isn't over the top and probably close to what AI might do at scale. +50% rev per employee is realistic from an AI bump, at least to test to see if sustainable. My issue is people saying it's all coming down so Block can get to +100-200% rev per employee. It will also be interesting to see if any legit Block competitors rise up from these layoffs. Since these people can use AI to prototype overnight.

The state will have no choice (and will want to) but to criminalize property ownership except for a select few. It might outlaw certain property transfers. Then it will give us all an expiring CBDC stipend every month. If you don't use it, you lose it. They will control what you can spend on and make sure you're in compliance. So you'll buy grub hub (making you more compliant), streaming and cheap crap. The current elites will be the only financial survivors.

I think we really are close to end of western democracy

Logos Stick said:

---

GeorgiAg said:

---

1981 Monte Carlo said:

---

GeorgiAg said:

---

Just got a sales call from an AI company that basically will do almost everything my secretary does right now.

It's starting....

---

What's the moral thing to do?

---

I am not even considering it. I am going to listen to his sales pitch to see what's coming down the pike, but I could never do that to her. She lives paycheck to paycheck. And she has like 30 years experience so it can't replace that.

But if it were cheap enough, it would free her up to do more of the higher end stuff she does.

---

The reason you will eventually do it is because your competition will do it. Their bill rate will be lower.

Same reason offshoring happened. I'm sure there were companies that didn't want to offshore. Once their competitors started doing so, it was change or die.

---

True. This is the reason that driverless cars will happen. Insurance will slowly increase for people that want to drive their own cars. Eventually, the economic forces will force people to have the driverless Tesla's, etc. The same thing will happen with businesses. You are exactly right. We will all do this voluntarily because of our pocket books.

Dad-O-Lot said:

---

I consider the "AI is gonna kill us all" fear-mongering to be akin to the Y2K scare.

Yeah, it could be an issue, but our mere awareness of it negates it.

---

AI doesn't have to literally kill you
It can do so by making your job and worth obsolete. That is absolutely a real threat. And the Tsunami is coming. I know more people out of work right now than during the 2008 financial crisis. This is just the beginning.

We are going to regret the push towards STEM. We should have, instead, pushed to make watching all of the Terminator movies mandatory.

ToddyHill said:

---

One of the best policies adopted by my last employer was mandating Concur expense reports could only be submitted monthly. Kept my frustration to a manageable level.

---

I have not turned in an expense report in about 8 months because I hate the process so much. AP and finance keep sending me nasty emails, but I will eventually get to it!

Mr.Milkshake said:

---

There's such a severe misunderstanding of probabilistic language models.

They have no drive except what's engrained in them whether via pre-training, fine tuning, or system prompt. It's literally probabilistic KV mapping. Could say we are the same but we've already been programmed for survival. AI will not have that unless programmed that way, like us.

Currently LLM consumer versions can't remember much of anything beyond the human equivalent of something like an hour of conversation. This wil grow and if frontier model owners devoted their entire stack to a single instance, context memory would be interesting to observe, but right now nothing is planning or executing on ideas that take days to weeks of planning and implementation agentically.

I'm sure they'll get there but there is no self drive in these things except what we give them

---

You missed my post earlier, where there was a guy posting about an AI model that he gave an audio file where he did not give that the pre-existing ability to understand it. It went out on its own and found the ability to decipher that and respond verbally. While you may or may not call that drive that is something very closely resembling it.

Depending on which version of AI you subscribe to or pay for, they do have memory and can plan accordingly. I have had posting multiple threads in Google Gemini, which I pay for and it is pulling information for other threads. It absolutely has memory and if you don't think the version the upper level guys are using as that you are mistaken.

We are talking about current versions of things and this is improving every day every minute by exponential levels.

Logos Stick said:

---

GeorgiAg said:

---

1981 Monte Carlo said:

---

GeorgiAg said:

---

Just got a sales call from an AI company that basically will do almost everything my secretary does right now.

It's starting....

---

What's the moral thing to do?

---

I am not even considering it. I am going to listen to his sales pitch to see what's coming down the pike, but I could never do that to her. She lives paycheck to paycheck. And she has like 30 years experience so it can't replace that.

But if it were cheap enough, it would free her up to do more of the higher end stuff she does.

---

The reason you will eventually do it is because your competition will do it. Their bill rate will be lower.

Same reason offshoring happened. I'm sure there were companies that didn't want to offshore. Once their competitors started doing so, it was change or die.

---

I worked for 2 S&P mid cap companies that refused to offshore.

Both were acquired by competitors that pioneered in offshoring.

TheEternalOptimist said:

---

Yeah - being in Implementation and Operations, I can see the AI tsunami coming.

I am not in denial that it's coming. I just hope it holds off long for me to early retire early from the big blue German financial software company that I work for. We are implementing it across the spectrum of our products in terms of operations, implementation, support, and even sales. Many of you here I assure you use the travel and expense platform I work on.

I have to say I 'concur' with a lot of the concerns about AI taking jobs... but I also don't think it's the end of the world.

For the near future, a lot of the learn to code folks need to learn to weld, plumb, or electrician skills. That might include me .

---

You might have suddenly becomes on of the least liked posters on F16. Brave admission. ( I jest, i didn't really have a problem with it at my companies :-) )

Also, you can learn to weld or plumb, until AI figures out a way to fundamentally transform that as well.

Car mechanics was a great job about 10 years ago.

In my lifetime, I have survived Elvis, The Beatles, the black mold scare, Y2K, and covid. AI? Not worried.

GeorgiAg said:

---

Mr.Milkshake said:

---

There's such a severe misunderstanding of probabilistic language models.

They have no drive except what's engrained in them whether via pre-training, fine tuning, or system prompt. It's literally probabilistic KV mapping. Could say we are the same but we've already been programmed for survival. AI will not have that unless programmed that way, like us.

Currently LLM consumer versions can't remember much of anything beyond the human equivalent of something like an hour of conversation. This wil grow and if frontier model owners devoted their entire stack to a single instance, context memory would be interesting to observe, but right now nothing is planning or executing on ideas that take days to weeks of planning and implementation agentically.

I'm sure they'll get there but there is no self drive in these things except what we give them

---

You missed my post earlier, where there was a guy posting about an AI model that he gave an audio file where he did not give that the pre-existing ability to understand it. It went out on its own and found the ability to decipher that and respond verbally. While you may or may not call that drive that is something very closely resembling it.

Depending on which version of AI you subscribe to or pay for, they do have memory and can plan accordingly. I have had posting multiple threads in Google Gemini, which I pay for and it is pulling information for other threads. It absolutely has memory and if you don't think the version the upper level guys are using as that you are mistaken.

We are talking about current versions of things and this is improving every day every minute by exponential levels.

---

Anyone else read this in Joshua's voice?

I think things are changing at a breakneck pace right now within the AI space and it's hard to keep up.

These harnesses like Claude code are pretty remarkable. We've probably all conversed with different LLMs at this point and they've acted like guides just providing information. But something like Claude code can take their instruction, implement on your computer, or talk to another application through an API and perform task on it, and then the harness keeps it on a loop until the jobs done. Its just constantly asking the LLM what should I do next, what should I do next. So now it's writing code, testing code, installing libraries, documenting, retesting, until the job is done. You can put it on YOLO mode and it won't even ask you for permission and will just keep going until the job is done. Its incredible.

I'm in medicine and am strongly considering using AI for radiology review. In medicine, radiologists will be easily the most replaceable. The cost of an X-ray review will be 1/3 to 1/10 (if statted) of a human, and more accurate overall. And results will come back in 5 minutes instead of hours (or days). The quality of care will go up significantly and the cost will go down exponentially. Tough pill to swallow, but at least in my industry AI will revolutionize quality and cost across the board. We have to adapt to the changing rules, and make progress best we can, which is still possible in spite of inevitable damage along that way. That said, I know doctors' roles are dispensable as well. PCPs are the most worthless, so I actually am ok with that entire role going away, so they can maybe even become an actual productive part of the system. PCPs and radiologists will go away soon, with zero negative consequences. After that, I am betting that a lot of doctor roles will be easily replaceable, and nursing may actually end up being more stable jobs than doctors. "Thinking" will be replaced, but "doing" will remain for longer.

Of course, that won't translate to lower healthcare or insurance costs, as someone above will profit regardless of lower expenses, but at least the quality and efficiency will go up...

deddog said:

---

Logos Stick said:

---

GeorgiAg said:

---

1981 Monte Carlo said:

---

GeorgiAg said:

---

Just got a sales call from an AI company that basically will do almost everything my secretary does right now.

It's starting....

---

What's the moral thing to do?

---

I am not even considering it. I am going to listen to his sales pitch to see what's coming down the pike, but I could never do that to her. She lives paycheck to paycheck. And she has like 30 years experience so it can't replace that.

But if it were cheap enough, it would free her up to do more of the higher end stuff she does.

---

The reason you will eventually do it is because your competition will do it. Their bill rate will be lower.

Same reason offshoring happened. I'm sure there were companies that didn't want to offshore. Once their competitors started doing so, it was change or die.

---

I worked for 2 S&P mid cap companies that refused to offshore.

Both were acquired by competitors that pioneered in offshoring.

---

India will fall first imo.

WestHoustonAg79 said:

---

deddog said:

---

Logos Stick said:

---

GeorgiAg said:

---

1981 Monte Carlo said:

---

GeorgiAg said:

---

Just got a sales call from an AI company that basically will do almost everything my secretary does right now.

It's starting....

---

What's the moral thing to do?

---

I am not even considering it. I am going to listen to his sales pitch to see what's coming down the pike, but I could never do that to her. She lives paycheck to paycheck. And she has like 30 years experience so it can't replace that.

But if it were cheap enough, it would free her up to do more of the higher end stuff she does.

---

The reason you will eventually do it is because your competition will do it. Their bill rate will be lower.

Same reason offshoring happened. I'm sure there were companies that didn't want to offshore. Once their competitors started doing so, it was change or die.

---

I worked for 2 S&P mid cap companies that refused to offshore.

Both were acquired by competitors that pioneered in offshoring.

---

India will fall first imo.

---

Not certain, but they will do better in the short term, because a quick solution for many companies is just moving jobs there.

Once you've eliminated US jobs, and imo, there will be an overcorrection, then a quick solution is to hire folks in India.

In the long term, we'll all be dead :-)

Mr.Milkshake said:

---

There's such a severe misunderstanding of probabilistic language models.

They have no drive except what's engrained in them whether via pre-training, fine tuning, or system prompt. It's literally probabilistic KV mapping. Could say we are the same but we've already been programmed for survival. AI will not have that unless programmed that way, like us.

Currently LLM consumer versions can't remember much of anything beyond the human equivalent of something like an hour of conversation. This wil grow and if frontier model owners devoted their entire stack to a single instance, context memory would be interesting to observe, but right now nothing is planning or executing on ideas that take days to weeks of planning and implementation agentically.

I'm sure they'll get there but there is no self drive in these things except what we give them

---

Go look up the OpenClaw project that OpenAI just acquired and realize you're about 6 weeks behind.

JB99 said:

---

I think things are changing at a breakneck pace right now within the AI space and it's hard to keep up.

These harnesses like Claude code are pretty remarkable. We've probably all conversed with different LLMs at this point and they've acted like guides just providing information. But something like Claude code can take their instruction, implement on your computer, or talk to another application through an API and perform task on it, and then the harness keeps it on a loop until the jobs done. Its just constantly asking the LLM what should I do next, what should I do next. So now it's writing code, testing code, installing libraries, documenting, retesting, until the job is done. You can put it on YOLO mode and it won't even ask you for permission and will just keep going until the job is done. Its incredible.

---

Incredible and terrifying.

I ****ing hate AI! Can't believe we're stupid enough to keep pursuing this. We're not an impressive race, the human being.

GeorgiAg said:

---

Just got a sales call from an AI company that basically will do almost everything my secretary does right now.

It's starting....

---

This is what I don't get. Is this a good thing? Why make something to do this? Isn't there some kind of social contract for us out there? Why make everyone destitute?