New APPS: Art, Politics, Philosophy, Science

Another Platonism in LLM Debates, Part 1: A Wittgensteinian Rejection of Cartesian AI

By Gordon Hull

There’s an emerging literature on Large Language Models (LLMs, like ChatGPT) that basically argues that they undermine a bunch of our existing assumptions about how language works.  As I argued in a paper a year and a half ago, there’s an underlying Cartesianism in a lot of our reflections on AI, which relies on a mind/body distinction (people have minds, other things don’t), and then takes language use as sufficient evidence that one’s interlocutor has a mind.  As I argued there, part of what makes LLMs so alarming is that they clearly do not possess a mind, but they do use language.  So they’re the first examples we have of an artifact that can use language; language-use is no longer sufficient to indicate mindedness.  In that paper, drew the implication that we need to abandon our Cartesianism about AI (caring whether it “has a mind”) and become more Hobbesian (thinking about the sociopolitical and regulatory implications of language-producing artifacts).  Treating LLMs as the origin points of speech has real risks, including making the human labor that produces them invisible, and making it harder to impose liability since machines can’t meet a standard scienter requirement for assigning tort liability.

Here I want to take up a somewhat different thread, one that I started exploring a while ago under the general topic of iterability in language models.  This thread takes the literature on language models seriously; where I want to go with it is to talk about an under-discussed latent Platonism in how we tend to approach language(models).   I’ll start with the literature, which divides into a couple of sections, a Wittgensteinian and a Derridean.

1. The Wittgensteinian Rejection of Cartesian AI

Lydia Liu makes the case for a direct Wittgensteinian influence on the development of ML, via the Cambridge Researcher Margaret Masterson.  I only ran into this work recently, so on the somewhat hubristic assumption that other folks in philosophy also don’t know it, I’ll offer a basic summary here (in my defense: Liu says that “the news of AI researchers’ longtime engagement with Wittgenstein has been slow to arrive.”  She then adds that “the truth is that Wittgenstein’s philosophy of language is so closely bound up with the semantic networks of the computer from the mid-1950s down to the present that we can no longer turn a blind eye to its embodiment in the AI machine” (Witt., 427)).

Tidbit of good news on indirect costs

The NSF had attempted to reduce indirect costs (F&A) on all future grants to 15%, in a somewhat more coherent version of the NIH's effort to do so for all ongoing and future grants.   A federal court today enjoined the rate cut, vacating the new rule, finding that "National Science Foundation’s 15% Indirect Cost Rate and the Policy Notice: Implementation of Standard 15% Indirect Cost Rate, NSF 25-034 are invalid, arbitrary and capricious, and contrary to law"

 

Helen de Cruz, 1978-2025

Former NewAPPS blogger Helen de Cruz passed away yesterday, Friday, June 20, 2025.  I never met Helen personally, but they were part of a vibrant community of bloggers at NewAPPS that welcomed me and provided the supportive context for my own development as a blogger ten years ago.  That experience was, and is, very important to me, and I will always be grateful for it.

I recommend Eric Schliesser's remarks here.

Helen's NewApps writings are here. 

Your Brain on ChatGPT

I wish I’d come up with that title, but it actually belongs to a new study led by Natalia Kosmyna of the MIT Media Labs.  The study integrates brain imaging with questions and behavioral data to explore what happens when people write essays using large language models (LLMs) like ChatGPT.  I haven’t absorbed it all yet – and some of the parts on brain imaging are well beyond my capacity to assess – but the gist of it is to confirm what one might suspect, that writing essays with ChatGPT isn’t really very good exercise for your brain.  The study assigned participants to one of three groups and had each write essays. One got to use an LLM, one used Google search, and one wasn’t allowed to use either.

The results weren’t a total surprise:

“Taken together, the behavioral data revealed that higher levels of neural connectivity and internal content generation in the Brain-only group correlated with stronger memory, greater semantic accuracy, and firmer ownership of written work. Brain-only group, though under greater cognitive load, demonstrated deeper learning outcomes and stronger identity with their output. The Search Engine group displayed moderate internalization, likely balancing effort with outcome. The LLM group, while benefiting from tool efficiency, showed weaker memory traces, reduced self-monitoring, and fragmented authorship. This trade-off highlights an important educational concern: AI tools, while valuable for supporting performance, may unintentionally hinder deep cognitive processing, retention, and authentic engagement with written material. If users rely heavily on AI tools, they may achieve superficial fluency but fail to internalize the knowledge or feel a sense of ownership over it” (138)

These results were corroborated by the brain imaging, and “brain connectivity systematically scaled down with the amount of external support: the Brain‑only group exhibited the strongest, widest‑ranging networks, Search Engine group showed intermediate engagement, and LLM assistance elicited the weakest overall coupling” (2 (from the abstract)).   That is:

New Paper: "AI and Healthcare Disparities: Lessons from a Cautionary Tale in Knee Radiology"

This paper, which has been forthcoming in Journal of Medicine and Philosophy for a while, is my foray into AI and healthcare, particularly medical imaging.  It synthesizes some of what I have to say about structural injustice in AI use (and why "bias" isn't the right way to assess it), and uses a really interesting case study from the literature to explore why it's important to understand AI as part of sociotechincal systems - and how understanding it as part of sociotechnical systems makes a big difference in seeing when/how it can be helpful (or not).  Here's the abstract:

Enthusiasm about the use of AI in medicine has been tempered by concern that algorithmic systems can be unfairly biased against racially minoritized populations. This paper uses work on racial disparities in knee osteoarthritis diagnoses to underline that achieving justice in the use of AI in medical imaging will require attention to the entire sociotechnical system within which it operates, rather than isolated properties of algorithms. Using AI to make current diagnostic procedures more efficient risks entrenching existing disparities; a recent algorithm points to some of the problems in current procedures while highlighting systemic normative issues that need to be addressed while designing further AI systems. The paper thus contributes to a literature arguing that bias and fairness issues in AI be considered as aspects of structural inequality and injustice and to highlighting ways that AI can be helpful in making progress on these.

Bullshit AI Jobs

Last week we heard the latest installment in the prophesized AI jobs apocalypse.  This time, it was Dario Amodi, the CEO of Anthropic, who told Axios that “AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years” (italics original).  Axios adds: “Imagine an agent writing the code to power your technology, or handle finance frameworks and analysis, or customer support, or marketing, or copy editing, or content distribution, or research. The possibilities are endless — and not remotely fantastical. Many of these agents are already operating inside companies, and many more are in fast production …. Make no mistake: We've talked to scores of CEOs at companies of various sizes and across many industries. Every single one of them is working furiously to figure out when and how agents or other AI technology can displace human workers at scale. The second these technologies can operate at a human efficacy level, which could be six months to several years from now, companies will shift from humans to machines.”  The piece then argues that this will be different from previous technological disruptions because of the speed with which it will occur.

Someone should tell that to the workers placed out of work all-but overnight by the development of machinery in the nineteenth century, as detailed by Marx (who helpfully notes in the Machinery chapter in Capital that the drive to full, steam-engine driven automation is motivated by the inability of capitalists to extract any more surplus value from over-exploited workers).  One should also remember, with Jathan Sadowski, that these sorts of proclamations are in part designed to create their own reality, such that “the power of expectations can have a disciplining effect on what people think” and that “the capitalist system is designed to pummel us into submission, preventing us from imagining life could be any other way, let alone allowing us to go on the offensive” (The Mechanic and the Luddite, 196, 207).  When Axios adds that “this will likely juice historic growth for the winners: the big AI companies, the creators of new businesses feeding or feeding off AI, existing companies running faster and vastly more profitably, and the wealthy investors betting on this outcome,” one can thus hardly be too surprised.

Here, I want to take a slightly different angle however, and think a little bit about the kinds of jobs that are supposed to go away.  It’s hard not to notice the parallels between the Axios list and this one: