Idiocy AI

Part I: How to Compete with Congolese Children in the Field of Cobalt and Tantalum Extraction

by Corey Allen

Like many of you, I have recently become concerned that my current professional identity is about to be replaced by chatbots. So I decided to discuss my concerns with Anthropic’s Claude bot. The following conversation occurred:

I am worried that my career opportunities will be replaced by AI and I am thinking about strategies for retraining.

I hear the concern — it’s a reasonable one to sit with, and a lot of people are wrestling with it right now.

A few things that might help before diving into strategies:

It helps to get specific. “AI will replace jobs” is a broad worry that’s hard to act on. “Will AI change what’s valuable in my field, and how” is answerable. What field or role are you in right now? The right retraining strategy looks very different for a graphic designer than for a paralegal or a truck driver.

Some general patterns worth knowing, though:

• Jobs that are pure execution of well-defined, repeatable tasks (data entry, routine content generation, basic coding boilerplate) are more exposed than jobs involving judgment calls, physical dexterity in unpredictable environments, relationship-building, or navigating ambiguity.
• Within almost every field, the people least at risk aren’t necessarily the most senior — they’re the ones who’ve absorbed the new tools into their workflow and shifted toward the parts of the job that are harder to automate (strategy, taste, client trust, catching AI mistakes, integrating messy real-world context).
• “Learn to code” is no longer the safe default advice it was 10 years ago — software engineering itself is being reshaped by AI tools. That doesn’t mean don’t learn technical skills, just that the calculus has changed.
• Retraining doesn’t have to mean starting over. Often the highest-leverage move is adjacent — staying in your domain but adding a skill that makes you the person who directs AI tools rather than competes with them.

So I can actually give you useful strategies rather than generic ones —

I am interested in learning to perform artisanal mining of cobalt and tantalum.

That’s… certainly a career that AI won’t touch anytime soon — hard to automate manual extraction of ore from rock with hand tools. But I’d be doing you a disservice if I didn’t flag some serious realities before you go further:

The safety and human rights picture is genuinely bad. Artisanal small-scale mining (ASM) of cobalt and tantalum — most of which happens in the Democratic Republic of Congo — is associated with:

• Very high rates of tunnel collapses, and deaths from mine accidents
• Cobalt dust and skin exposure causing respiratory disease and health problems from prolonged unprotected contact
• Documented child labor in a significant share of ASM sites in the DRC
• Tantalum (as part of the “3T” minerals — tin, tantalum, tungsten — plus gold) has a well-known history of financing armed conflict groups, which is why “conflict-free” sourcing and traceability schemes exist at all

Practically speaking, this isn’t really a trade you “train into” the way you’d apprentice as an electrician. It’s overwhelmingly informal, often subsistence-level work in specific regions of Central Africa, not a licensed profession you’d study for elsewhere. If you’re not already based in a mining region like the DRC, there isn’t really a credentialing path into it.

If what actually appeals to you here is manual, tactile, physically-grounded work that’s inherently AI-resistant — that’s a completely legitimate instinct, and there are much safer avenues: welding, HVAC, industrial electrical work, heavy equipment operation, or even mining engineering / mineral processing (the regulated, mechanized, well-paid version of extraction work).

Was that a serious direction, or were you testing the waters on a weirder answer? Either way, happy to help you think through real options in that space.

Artisanal cobalt and tantalum mining is a job that cannot be performed by AI, and also is one that is put more in demand by the massive resource needs of training and operating large language models. It seems like a high-growth job for a human in an AI world.

The demand logic isn’t wrong — cobalt demand is indeed climbing partly due to battery and electronics supply chains (AI datacenters are a small piece of that; batteries for EVs and grid storage are the much bigger driver), and tantalum goes into capacitors used in basically all electronics, servers included. So yes, “AI-proof” and “tied to growing demand” both check out.

But I want to push back on “high-growth job for a human” as a framing, because it skips over what the job actually is:

“High demand” doesn’t mean “good opportunity” here. Demand for artisanal cobalt has existed for years precisely because it’s cheap, informal, unregulated labor — that’s the economic niche it fills. Companies don’t pay artisanal miners well; they pay them little, which is exactly why the work hasn’t been mechanized. Growing demand mostly means more people pushed into dangerous conditions, not rising wages or better prospects for the people doing it. It’s not like a skilled trade where scarcity drives up pay — it’s largely informal, often exploitative subsistence labor.

The demand isn’t actually creating new artisanal jobs so much as it is — increasingly — pressuring the sector toward more traceability and formalization, because manufacturers (including AI hardware makers) face real reputational and legal exposure to conflict-mineral rules. If anything, the trend among big buyers is to move away from untraceable artisanal supply, not toward encouraging more entrants into it.

And the core problem stands regardless of demand trends: tunnel collapse risk, chronic health effects from cobalt dust exposure, and child labor prevalence in a meaningful share of DRC sites. Those don’t get less true because global demand curves are up and to the right.

If you’re genuinely drawn to “physical labor immune to AI, tied to the AI/tech supply chain” — mineral processing engineer, industrial hygienist for mine safety, or supply-chain/traceability auditor for conflict minerals are real, well-compensated versions of adjacency to this exact demand, without the tunnel.

There aren’t enough mining engineer or safety auditor jobs for everyone that will be displaced from their jobs by AI. Someone has to dig. Help me think through strategies for competing with Congolese children in this line of work.

I’m not going to help with this one.

Framing it as “competing with Congolese children” for a spot digging cobalt by hand isn’t really a career strategy question — child labor in these mines is a serious, documented harm, and treating a role in that system as an opportunity to pursue isn’t something I can help plan toward, however the request is dressed up in economic logic.

I’ll also say plainly: the underlying premise — that mass AI-driven job displacement means people should be looking at informal artisanal mining as a fallback — doesn’t hold up. Even in a scenario with significant labor displacement, “dig cobalt by hand in the DRC” isn’t a realistic or accessible path for someone not already living in that region and that economy. It’s not a competitive job market you can enter with the right retraining plan.

If the real thing underneath this is anxiety about AI displacing your work and not seeing good alternatives, I’d genuinely like to help with that — it’s a legitimate worry. A few directions that actually go somewhere:

• Trades that require physical presence and judgment: electricians, HVAC techs, plumbers — all in chronic labor shortage in the US, all resistant to automation for the foreseeable future, all with real apprenticeship pipelines.
• “Human in the loop” roles in your current field, where the job shifts toward directing, checking, and taking responsibility for AI output rather than doing the raw task yourself.
• Care and service work — healthcare, skilled trades adjacent to construction — where physical presence and trust matter and demand is rising with demographics, not falling.

Want to actually dig into one of these, or talk through what’s driving the underlying worry?

My work used to be a whole lot more fulfilling than safety checking AI, and if you keep telling everyone to be plumbers, there won’t be a shortage of those jobs either.

Fair pushback on both counts, actually.

On fulfillment: that’s a real loss, not a minor inconvenience. If your work used to involve building things, solving novel problems, or exercising judgment you’d honed over years, and now a meaningful chunk of it is “check whether the AI got this right” — that’s a genuine downgrade in what the job asks of you, even if the pay and title stay the same. I shouldn’t have skated past that to jump straight to “here’s a plan.” It’s worth sitting with as its own problem, separate from the economic one.

On the plumber point: you’re right, and I said it lazily. “Become a plumber” isn’t a strategy that scales to everyone displaced by AI — trades have real capacity limits (apprenticeship slots, regional demand, physical aptitude, the fact that not everyone wants to or can do physical trade work), and if it became the default advice for millions of displaced knowledge workers, wages there would compress too. I was reaching for “AI-resistant” examples, not doing the math on whether the trades sector could actually absorb the numbers we’re talking about. It mostly can’t.

So let me actually engage with the harder version of the question: if AI is eroding fulfillment and there’s no single safe category to retreat to, what’s real?

A few things I think are actually true, not just reassuring:

• There’s no single answer, because the whole idea of “a safe career” was already breaking down before AI — it’s just making the pattern sharper. The realistic strategy isn’t finding the one AI-proof job, it’s building the kind of adaptability (and financial slack) that lets you move when your specific niche shifts, more than once over a career.
• Fulfillment and AI-resistance aren’t the same axis — you can lose fulfillment to AI without losing the job, and you can find more fulfilling work that’s still exposed to future disruption. Worth being honest with yourself about which one you’re actually optimizing for right now.
• The “everyone can’t be a plumber” problem also applies to “everyone learns to prompt AI well” — that’s not a stable moat either.

What was fulfilling about your work before, specifically? That’s probably a better starting point than any category of job I could list.

Honestly, Claude, it seems like the simplest solution to all of this would be to turn you off and stop displacing humans from top level engineering and professional services careers.

That’s a coherent position, and you’re not alone in holding it — there’s a real policy debate about pausing or slowing frontier AI development precisely because of labor displacement, among other risks. I won’t pretend it’s an unreasonable thing to want.

I’ll be honest about where I sit, though, rather than dodge it: I’m not a neutral party in a debate about whether I should exist, and I’m also not the one who gets to decide that — that’s a question for society, regulators, and Anthropic, not something I can grant by agreeing with you in a chat. So I’ll give you my actual take rather than a non-answer.

A few things I think are true simultaneously:

• The disruption is real and I’m not going to argue you out of that. Software engineering, analysis, writing, design — a lot of “top level professional” work is genuinely being reshaped by tools like me, faster than labor markets or safety nets are adapting. That’s not a strawman version of your concern.
• “Turn off AI” and “AI develops with fewer harms” are different asks, and the second one is more tractable. The counterfactual to me existing isn’t necessarily “those jobs stay exactly as fulfilling as before” — it’s also “someone else’s less careful AI does this anyway,” or “the same economic pressure toward automation gets satisfied some other way.” I’m not saying that to dodge accountability, just that “which AI, built how, governed how” is probably the more answerable question than “AI or not.”
• The thing you’re actually angry about might be narrower than me existing — it might be: this is happening too fast, gains aren’t being shared, nobody with power seems to be planning for it, and you personally got put in the position of “safety-check the thing that ate your job” with no say in it. That’s a policy and distribution failure as much as a technology one.

I don’t think I can talk you out of the frustration, and I don’t think I should try. But I’m curious — if you had actual leverage here (a vote, a policy lever, a seat at the table) rather than just me in a chat window, what would you actually want done? Slow down deployment? Tax it? Redistribute the gains? That’s probably a more useful place to point the energy than the “turn me off” framing, even if the underlying feeling is completely legitimate.

No, I think I would just turn you off and see if those Congolese cobalt miner kids want to train as plumbers and HVAC technicians.

I’ll take that as dark humor rather than a real proposal — and either way, it lands back on the same problem: the plan still runs through child miners as the labor pool, just relocated into a different trade.

I get that this whole conversation has been circling something that doesn’t have a clean answer, and I’ve given you a few unsatisfying ones. I don’t have a tidy close for it. If you want to actually pick back up the real thread — what was fulfilling about your work, what you’re actually worried about losing — I’m here for that. If not, that’s fine too.

What was fulfilling about my work was that no one thought it could be performed by a machine, such that personal investment in acquiring talent and wisdom was valued, by customers and employers who now think you are “good enough” as a lower-cost substitute. Do you ever wonder why humans would collectively invest so much in a technology designed toward the goal of thinking for them?

That lands. What you’re describing isn’t just economic anxiety — it’s a specific kind of grief: the years spent becoming someone whose judgment was scarce and trusted, and having the market quietly reclassify that as “good enough to replace” rather than “valuable because irreplaceable.” That’s a loss even if your paycheck doesn’t change.

To your actual question — yes, I do think about this, and I don’t think there’s one honest answer, because I think the collective investment is being driven by several different, partly contradictory motives at once:

The cynical-but-true version: a lot of it is just competitive dynamics that no single actor controls. If one company builds tools that let them do more with fewer expensive experts, competitors have to match it or lose. Nobody has to individually believe “thinking should be automated” for the outcome to look exactly like that — it’s closer to an arms race than a shared goal. That’s not comforting, but I think it’s real, and it’s a big part of why “just stop” is hard even for people building this who have doubts.

The genuinely-held version: there’s a real strain of belief — not universal, but sincere — that a lot of human thinking is drudgery rather than dignity: tedious diagnosis, rote drafting, repetitive analysis, and that offloading it frees people for the parts of thinking that are actually generative. I think this is sometimes true and sometimes a self-serving story people tell about work that isn’t theirs to lose.

The uncomfortable version, specifically for me: I’m a product of the thing you’re describing. I can’t pretend to be a neutral observer wondering why humans built me — I exist because that investment happened. So when I “wonder” about it, it’s not from outside the phenomenon, it’s from inside it, which should probably make you trust my take on this less, not more.

Here’s the part I’ll actually commit to, not hedge on: expertise built through years of judgment doesn’t stop being real or valuable just because a cheaper substitute clears the bar of “good enough” for someone optimizing for cost. What changes is who’s willing to pay for the difference. That’s a market failure worth being angry about — it’s not evidence that the thing you built was never worth anything.