Trades: Absolutely. Makes me glad I can swing a hammer and crank a wrench to do most home/auto repair on my own. See the above South Park clip. Military: In a country with rational military minds making decisions, yes. Government: Fuuuucckkkkk no. I am very close to this area and I can tell you that the US Government is full steam ahead on automating programmatically and/or AI. And in many cases, it's long overdue, like with SSA disability claims processes or even going to the DMV. Combine that with the over 300,000 already laid off by the Federal government in a year, and it's not a rock solid career anymore. Lawyers: Not sure what you're thinking here with Lawyers but the vast majority of lawyers aren't flashy trial lawyers like you see on TV. Most of them are just applying law to whatever they're hired to do. AI can easily do a big chunk of this to a certain degree. Health Care: Warm bodies, meaning doctors and especially nurses in clinical settings, yes. However, big data is already compressing time spent with each patient and doctors are burning out. Big EHR companies like Epic are integrating AI to assist in diagnosis. I would say the most vulnerable specialty is probably radiology, which is a good thing IMO, since they're paid an insane amount of money. I used to work with radiology departments and radiologists were fearful of AI ~10 years ago. All I can say is I hope they invested their cash while it was pouring in. Programming (added by me): This is a slippery slope. Senior devs I know love AI and how it can pump out basic framework in a heartbeat. However, they still need to customize what the code does. I've played with AI app makers and I'm not impressed yet. But there will be a time when the coding is very good. It's gotten to a point where I'm going to have a discussion with my oldest son who is a sophomore at UH for Computer Science. I read a blurb that last year was the first year for CS graduate job shrinkage.
I'm in Marketing, so I would be really worried if I wasn't so close to retirement. Like others here, I'm just riding out the clock at this point. Staying as long as they'll have me. I'm a little worried about my kids, but I think by the time they hit the workforce, everyone will have figured out a way to adapt.
I think there's a real art and nuance to marketing that AI can't capture. Of course, every small business under the sun is whipping out terrible-looking AI logos but it's not like they were hiring marketing firms in the first place. I'm personally not riding out the clock. I think I have another big push or two into different ventures in me before I hang up my spurs...but they'll always be by the door because I get bored.
AI is coming for the white color jobs of folks who are comfortable with the status quo. AI is domain specific, so if you upskill yourself in your domain, you'll be just fine. The great part is, AI is the first tool ever created, that teaches you how to use it. We're in a lifelong learning world now, folks.
AI will now help you know if its going to take your job. Anthropic launches AI job destruction detector Anthropic says it is building an early-warning system for a potential AI-driven white-collar jobs bust. Why it matters: While the new index from Claude's maker shows "limited evidence" that AI has affected joblessness so far, the effort enters a larger debate among economists over how a possible "AI labor doom and gloom" scenario should be tracked in the first place. What they're saying: "By laying this groundwork now, before meaningful effects have emerged, we hope future findings will more reliably identify economic disruption than post-hoc analyses," Anthropic economists Maxim Massenkoff and Peter McCrory write in a new paper. ...
Anybody can look up an answer in AI. It's asking the right question that's the real skill. Teaching the next generation involves centering how to ask the right question. This is the way.
I have less than 9 years left, If I last five years, I am ok with that as well. I don't know what my kid should choose as a career path at all. Maybe engineering? Will doctors still be a good career for the next generation? I am pretty sure accountants and pharmacists will be close to dead end career within the next fifteen years.
Just keeping my head down and accumulating wealth. If/when AI takes over nothing I do can really stop it.
Plenty you can do, utilize AI to upskill yourself, and improve your personal workflow productivity. My best clients are blue color workers, believe it or not. AI and automation can help everyone, if they choose to embrace it. My two biggest clients right now are heavy haul trucking, and oil field manufacturing.
Why worry about "stopping" it? Why not try to use it to your advantage? Learn something to take advantage of it. One of the best times to take advantage of emerging techs is when most people haven't learned how to take advantage of it yet. lol. Motivation :
I think one of the unique challenges of it, depending on how you try to take advantage of it, is that much of what you learn will be obsolete in 6 months. For example, people who learned to be efficient at AI prompting and guidance to help write code have already been replaced by AI just being smart enough to do it from a more basic prompt. So it could be one of those things where you're always swimming upstream to keep up.
That's part of the evolution of anything. Software developers have always had to keep their skills up-to-date if they wanted to remain on the cutting edge or even relevant in some industries. Many times it was through on-the-job training. Nobody knows everything about everything at a job - we all learn and change our learning along the way. Admittedly, the AI stuff is changing quickly, but all I'm saying is the old saying of "at least get your foot in the door", then do, learn, gain experience, expand your knowledge, etc. When Java was first getting popular, there were people trying to learn in any way possible - reading books, magazines, getting together and trying to code anything just to say they did something. Back around 1995, Java developers were turning down $300/hr jobs (it was $300 or $600) in NYC because they already had too much contract work. That was rare, but that was in 1995 money! Mostly because there weren't a ton of opportunities or developers who knew what they were doing at the time. The ones who got their foot in the door and learned, reaped the rewards. I remember when HTML was getting popular back around 1993-1995, companies were paying $700-$800 per static page for someone to design their websites for them. And they were hideous, but... they were happy because they were on the Internet. People took advantage of that, too. Come to think of it, the first "real" job I got was because I told a company I could help them set up a Wildcat! BBS system back when BBS systems and Internet support were starting to get popular at companies. The only reason I knew how to do that was I decided to set one up at home a couple of years earlier for giggles. BTW, that whole AI prompting thing always seemed like you were preparing your own funeral.
Absolutely. However, the pace of it is the risk here to me. With previous programming, you learned skills you could use for years. You'd need to keep learning, of course, but you had a long enough timeline to make the skills worth learning. Today, the same amount of "effort" to learn those skills might only be valuable for a few months and it might not even be enough time to cover the cost of learning those skills. Not saying it will be - no one knows where all this is heading - but the pace of it means no matter how much you learn, you're likely not building any real long-term security or predictable career path.
Agreed that those may be the "safest" routes to go. But also ... In a World First, a Chinese Robot Successfully Performed Repairs on a 10 kV Powerline Many electricians, plumbers, welders like to think they are safe from the AI and robot wave sweeping the global job market, but it seems that in the face of technology evolving at lightning speed, no one is truly safe. For example, China recently announced that one of its robots successfully replaced human electricians and performed critical repairs on a 10 kV powerline. Developed by the Hubei Electric Power Research Institute and Wuhan State Grid, the robot electrician uses precise laser cutting technology to strip away the rigid insulation, clamps the branch wire onto the main wire and tightens the connections with absolute precision. No shaking hands, no mistakes to worry about, and no human life in danger. “This robot can navigate even the most complex wiring environments, and most importantly, it completely eliminates the possibility of people being electrocuted,” Wu Xin, a technician from Wuhan State Grid, said.
Most would say I got in late using AI as a co-pilot. But I quickly had a watershed moment when I realized I didn’t really have to learn how to maximize its helpfulness, I could just tell it to teach me, starting with tell it to write prompts for me. That happened bc the big issue not stated much about Claude code (for instance) is you run out of session limits quickly. Not just subscription limits, but it’s memory limited per session/conversation. Once it has to compact the conversation, it looses context memory, and starts getting dumb again. So, you need to keep it focused with good prompts, and spread it across multiple terminal windows (ie multiple claude sessions). Instead of trying to find humans (yt howtos and such) to tach me how to use AI efficiently, you just ask Claude to do it. like this, “hey Claude, you are about to run out of context memory, I’m going to need to restart the session. Write me a prompt to start the next session efficiently, so you’re back up to speed.” but the big thing is their ability to document. you always have it document “efficiently” and use guardrails to what it should know from the past and what not to do in the future.” Gemini taught me that. I told Gemini Claude keeps wanting to write lazy code. Gemini said, “we always want to write lazy code first. You have to set up guardrails and rules.” lulz. That said, ofc it always helps to have solid OOP and OCP software in the first place, because those are code contracts that the compiler enforces, which is a big disadvantage of procedural/functional programming…so, with elite documentation (AI is Elite at making it), and GoF Patterns (that I'm elite at), you can just say, “Rule, you must use the existing Interfaces (ie Jave, Kotlin and C++) or Protocols (Swift) contracts, defining the existing GoF patterns in the codebase.” That is language they understand. And then I add, “No code-sniffing allowed, motherf*****” And that basically stops them from writing lazy code. and I didn’t have to watch any HowTo channels. I just had to read it's code to learn how it thought, and ask it. Those skills are sustainable. You just keep telling the AI to teach you how to tell it to do your bidding. Oh, and another key thing with Claude code and it’s limits, use multiple terminal windows, and turn each one into a focused expert of a specific skill set. Architect, debugger, database, cloud services, documention, feature expert, etc. Because Claude will even tell you, “this session is too convuluted with info, start fresh with a new session, to get me focused again.” basically, they are LLM’s, keep their language tight.
Maybe we're seeing or talking about two different things or approaches. To me, AI for the average user isn't "how can I figure out how to teach the AI to do something" from the ground up like old-school coding, it's more about "I'm going to use this AI to do this specific thing for me - I just kind of coax it to give me solutions". I can do that specifically or in general terms and work my way to an acceptable solution. In other words, coding, for example can be something like an iterative session with your AI buddy as opposed to you sitting in front of a compiler banging out code all night/week. Or it could be something where you provide the guardrails, constraints, criteria, etc. and just tell it to go do it while you take a nap. See tweet below. I agree with what you said about it moving fast, but I've always found that people who enjoy what they do tend work their tails off to get ahead, learn more about it, and learn to do it better. If the first thing out of your mind is "man, I can't do this", then you can't do it, I guess. This is why I love spending time on places like Reddit watching "kids" build homelabs so they can learn a lot of this stuff. Is it necessary? No. But it's learning. When I left my last job, a couple of us were saying there'll come a day when we don't need traditional coders to write code - the machine will write the code for us, but hell, I didn't know it would happen only about 5 years later. lol. Claude's head guy Boris Cherney recently stated the following that kind of was a "whoa" moment for the planet :
