This dire warning against AI echoes the anxieties of a much earlier elite: the late-medieval clergy facing the invention of the printing press. For centuries, they held a privileged monopoly on knowledge, controlling its interpretation and dissemination. The printing press threatened to shatter that authority by democratizing access to information and empowering individuals.
Similarly, today's critics, often from within the very institutions they defend, frame AI as a threat to "expertise" and "civic life" when in reality, they fear it as a threat to their own status as the sole arbiters of truth. Their resistance is less a principled defense of democracy and more a desperate attempt to protect a crumbling monopoly on knowledge.
This is nothing but speculation written by lawyers in the format of a scientific paper to feign legitimacy. Of course those $500 an hour nitpickers are terrified of AI because it threatens the exorbitant income of their cartel protected profession.
Just from reading the abstract, it feels like the authors didn't even attempt at trying to be objective. It hard to take what they're saying seriously when the language is so loaded and full of judgments. The kind of language you'd expect in an Op-Ed and not a research paper
None of these paper's arguments are AI specific. The IRS doesn't need AI to make mistakes and be unable to tell you why it did so. You can find stories of that happening to people already.
Who do institutions serve? To me AI democratises information. Allows access to information that would normally be gatekept. AI reduces barriers, and they don't like that because those barriers gave them authority.
> Purpose-driven institutions built around transparency, cooperation, and accountability empower individuals to take intellectual risks and challenge the status quo.
I am not sure if I am off-topic, but I am having a lot of trouble with this statement. Institutions are often opaque, and I have never belonged to an institution that empowered me to "take intellectual risks and challenge the status quo." Quite the contrary.
I fear the title of this article is going drive most of the conversation.
I haven’t read through the whole thing yet, but so far the parts of the argument I can pull out are about how Institutions actually work, as in a collection of humans. AI, as it currently stands, interacts with humans themselves in ways that hollow out the kind of behavior we want from institutions.
“ Perhaps if human nature were a little less vulnerable to the siren’s call of
shortcuts, then AI could achieve the potential its creators envisioned for it. But that is not the world we live in. Short-term political and financial incentives amplify the worst aspects of AI systems, including domination of human will, abrogation of accountability, delegation of responsibility, and obfuscation of knowledge and control”
An analogy that I find increasingly useful is that of someone using a forklift to lift weights at gym. There is an observable tendency when using LLMs, to cede agency entirely to the machine.
>Civic institutions—the rule of law, universities, and a free press—are the backbone of democratic life. They are the mechanisms through which complex societies encourage cooperation and stability, while also adapting to changing circumstances.
I would summarize the central claim of the paper as: the widespread use of AI to mediate human interaction will rob people of agency, understanding and skill development, as well as destroying the social links necessary to maintain and improve institutions, while at the same time allowing powerful unaccountable actors (AI cabal) to interject into those relations and impose their institutional goals; by "institution" we mean a shared set of beneficial social rules, not merely an organization tasked with promoting them, "justice" vs. "US justice system".
The authors then break down the mechanisms by which AI achieves these outcomes (that seem quite reductive and dated compared to the frontier, for example they take it as granted that AI cannot be creative, that it can only work prospectively and can't react to new situations and events etc.), as well as exemplifying those mechanism already at work in a few areas like journalism and academia.
This articles claims are also interesting to me in terms of large scale software systems. They authors say:
> They (AI systems) delegitimization knowledge, inhibit cognitive development, short-circuit long term thinking processes, and isolate humans by displacing or degrading long term human connection.
This is a pretty good summary of the worry I’ve seen expressed about extensive use to build large pieces of software. Large pieces of software aren’t just the code that describes them. They exist in some sense in their authors as well.
Effective projects seek to expand understanding of software systems across the organization so that relevant decisions can be made about their future. By relegating much of the decision making around structure of the software to AI you lose the systemic knowledge shared across the organization.
> The affordances of AI systems have the effect of eroding expertise
Recently, it so happened that I spent an hour reverse engineering and documenting a piece of a system. A co-worker asked a LLM to do the same. It generated some really nice documentation.
The difference is, I (as a team member) now have the understanding. Generating the documentation does not increase the understanding of the team.
I think the argument makes a central mistake in putting too much trust into institutions. I don’t disagree with the conclusions, but the premise around blindly trusting institutions simply because they’ve been around for a long time lost me from taking most of their arguments seriously, despite opening the article thinking I would agree.
The arguments that "AI Destroys Institutions" seem pretty iffy. I scrolled down to see what institutions had been destroyed it went on about Elon Musk's Doge destroying stuff but their closing say USAID I think had zero to do with AI.
I'm in the UK and I don't think any institutions have been destroyed or even noticeably harmed by AI. In the US there is general chaos under Trump so it may be hard to differentiate.
28 comments
[ 3.3 ms ] story [ 48.2 ms ] threadThis affordability is HEAVILY subsidized by billionaires who want to destroy institutions for selfish and ideological reasons.
Similarly, today's critics, often from within the very institutions they defend, frame AI as a threat to "expertise" and "civic life" when in reality, they fear it as a threat to their own status as the sole arbiters of truth. Their resistance is less a principled defense of democracy and more a desperate attempt to protect a crumbling monopoly on knowledge.
Stopped reading here, as these people still believe in that fairytale of theirs.
Not that I think there is a lot of thinking going on now anyway, thanks to our beloved smartphones.
But just think about a time when human ability to reason has atrophied globally. AI might even give us true Idiocracy!
I am not sure if I am off-topic, but I am having a lot of trouble with this statement. Institutions are often opaque, and I have never belonged to an institution that empowered me to "take intellectual risks and challenge the status quo." Quite the contrary.
I haven’t read through the whole thing yet, but so far the parts of the argument I can pull out are about how Institutions actually work, as in a collection of humans. AI, as it currently stands, interacts with humans themselves in ways that hollow out the kind of behavior we want from institutions.
“ Perhaps if human nature were a little less vulnerable to the siren’s call of shortcuts, then AI could achieve the potential its creators envisioned for it. But that is not the world we live in. Short-term political and financial incentives amplify the worst aspects of AI systems, including domination of human will, abrogation of accountability, delegation of responsibility, and obfuscation of knowledge and control”
An analogy that I find increasingly useful is that of someone using a forklift to lift weights at gym. There is an observable tendency when using LLMs, to cede agency entirely to the machine.
https://www.pewresearch.org/politics/2025/12/04/public-trust...
In the 1960s, trust in institutions was around 70%.
In 2025 it's about 17%.
This is not something like we had 70% until 2023 and then AI dropped our trust suddenly. If anything, AI doesnt even register on the graph.
So correlation here is practically non-existent. Gallup and Pew have the similar trends for journalists and universities.
You dont get to blame AI for this.
Interesting bump and question. How did clinton improve reputation and then bush destroy it? Or is that a false hump?
The authors then break down the mechanisms by which AI achieves these outcomes (that seem quite reductive and dated compared to the frontier, for example they take it as granted that AI cannot be creative, that it can only work prospectively and can't react to new situations and events etc.), as well as exemplifying those mechanism already at work in a few areas like journalism and academia.
Working as intended. WONTFIX.
> They (AI systems) delegitimization knowledge, inhibit cognitive development, short-circuit long term thinking processes, and isolate humans by displacing or degrading long term human connection.
This is a pretty good summary of the worry I’ve seen expressed about extensive use to build large pieces of software. Large pieces of software aren’t just the code that describes them. They exist in some sense in their authors as well.
Effective projects seek to expand understanding of software systems across the organization so that relevant decisions can be made about their future. By relegating much of the decision making around structure of the software to AI you lose the systemic knowledge shared across the organization.
Recently, it so happened that I spent an hour reverse engineering and documenting a piece of a system. A co-worker asked a LLM to do the same. It generated some really nice documentation.
The difference is, I (as a team member) now have the understanding. Generating the documentation does not increase the understanding of the team.
How Generative AI is destroying society https://garymarcus.substack.com/p/how-generative-ai-is-destr... (https://news.ycombinator.com/item?id=46616713)
I'm in the UK and I don't think any institutions have been destroyed or even noticeably harmed by AI. In the US there is general chaos under Trump so it may be hard to differentiate.