The AI-Trap

AI is a Cognitive Drug and Big Tech are the new Drug-lords

Having used AI for a while now I am scared by how addictive it is.

Still AI is in an introductory phase and you can do meaningful work without being charged. So a lot of people and organisations are giving it a test. As addiction grows and usage takes off charging will go up too. Providing AI is the ultimate capitalist trick: addict your customers and squeeze them - ensuring a permanent and growing demand.

As with narcotics, when you are young and unexperienced, you want to give it a try to learn what it is. Delightful experiences is the door into addiction. After a lot of (mis-)use the experiences are no longer so delightful. You need to be strong and smart to jump off the trail in time to save yourself.

We are now in this testing phase of this new cognitive drug. Will we have the strength and smartness to jump off in time? It does not look so. Rather it seems that we will be lured into a tremendous ai-demic¹.

How does AI-addiction work?

Let us take my experience as a case.

I do writing and coding on a daily basis. I have a fairly large collection of self-made tools on my computer. Each tool has been developed over time through learning-by-doing. I am a practitioner, not a specialist in each computer language I use. In between needs for improving these tools arrise, may be as more complex problems are addressed. More sophisticated language constructions, beyond my practitioner skills, may be required.

Previously, there where no other option than upgrading my own skills. Today, AI offers an easy way out, providing an upgraded (ai-)tool without upgrading my skills (the hard part). I can check that the AI-generated code produces the correct output, but I no longer understands how it is done. The code in the ai-tool make use of language constructions gathered from among the best practise found on the internet and is clearly superior to my home-grown code.

I have got a better tool and solved the immediate problem elegantly, but I now have lost control of my tool. I am now dependent on AI next time I need to upgrade it further. I am addicted.

That is, if I don’t pull the breaks! I could say “no”, I will not upgrade my tool further unless I have taken the hard work of upskilling myself to the level of my newly acquired (ai-)tool.

But will I do that? AI provided my new tool in a dialogue going on for one day. My up-skilling to this new language level will take weeks. I can’t afford upskilling myself as long as AI is free. I am stuck and have to accept the later increasing charges as AI squeezes me. As my skills development deteriorates I will be doomed.

Breaking the AI-Trap on a Societal Level?

My experience is the marginal case. But social and economic development is marginal business. A lot a small steps taken by myriads of actors. With nobody in control of where it ultimately leads.

With millions of marginal adopters like me Big Tech is sure to rule the world. Western societies have no way of stopping this bandwagon.

The Trap: Super-intelligence is creating deskilled Societies

Humans as lazy animals. We always first pick the low-hanging fruits. That is marginalism as social philosophy. Where we are headed are left for tomorrow.

With AI we solve increasingly complex problems cheaper and cheaper. The costs not counted are the deteriorating skill level - human capital being depreciated as never before.

Appendix: What AI says about this

Please verify this statement: “AI is creating deskilled societies”

Verdict: not established as stated. There is real evidence of deskilling in specific settings, but “deskilled societies” is a much larger claim, and the evidence doesn’t yet support it.

What supports it

• Medicine. The most cited study is Budzyń et al. in The Lancet Gastroenterology & Hepatology. After routine AI assistance was introduced for colonoscopy, experienced endoscopists’ adenoma detection rate in procedures done without AI fell from 28.4% to 22.4%. It has been described as the first real-world clinical evidence consistent with AI-related deskilling. arxiv
• Job structure. An IZA paper using U.S. O*NET data for 2011–2025 found a mixed pattern. Within occupations, greater AI exposure is associated with higher proficiency requirements for selected capabilities. More exposed occupations also show a compression in the overall breadth of capabilities required. Jobs are becoming more specialized and less diverse in what they demand, which is not the same as people losing skills. repec
• Worker concern. A 2026 survey of US healthcare workers found that 74 percent of clinicians are worried about losing skills due to over-reliance on AI. That shows worry, not actual skill loss. bernardmarr

What undercuts it

• The flagship study is contested. The before and after populations differed on characteristics that drive detection rates, withdrawal times were not reported, and only nineteen endoscopists were studied. Secondary outcomes did not significantly decline, which is hard to square with genuine skill loss. arxiv
• Evidence elsewhere is thin. In software engineering, the evidence is limited to industry telemetry and self-report. The reviewer calls it documented concern with causal evidence pending. arxiv
• History points the other way. The same review weighs the field evidence against a large contrary meta-analytic prior from previous technologies. Past tools such as calculators and spreadsheets shifted which skills mattered more than they erased skill overall. arxiv
• Measurement is immature. One review notes that validated assessment instruments for deskilling largely do not exist yet. ai2
• Some evidence points to upskilling. A 2026 systematic review of 67 studies found AI can support critical and creative thinking when teachers pair it with structured inquiry, reflection and assessment. ecosistemastartup

A fairer version of the claim

“In some professions, heavy reliance on AI may be eroding specific unassisted skills, and in education it may prevent some skills from being acquired in the first place. The evidence is early, mostly observational, and contested.” Some researchers call the education problem “never-skilling.”

Whether society ends up deskilled probably depends on how AI is deployed, for example whether people keep practicing core skills or the tools replace the practice entirely. That is a prediction, not a verified fact.