
The curious gap between blind anti-AI hostility and my experience
Next April, with an American family visiting France, I hope to stand at the precise spot where one of their forebears was killed. It seems to have been one of many small war crimes committed by the SS. As far as I can tell, it was never investigated and never made headlines. The victims were Captain Doctor William F. Kober, commander of the Medical Detachment of the 24th Cavalry Reconnaissance Squadron, and two of his men. I think I’m the first to tell their story, 82 years later.
The full story will be published at 24thcavrecon.org, but the synopsis is this. Bill Kober, Taylor Keith and Harold Gilbert were ambushed in Villers-Cotterêts, north east of Paris, on the morning of August 29 1944. They were in a medical half track, clearly identified as an ambulance, when an SS unit destroyed the vehicle at a road block, using an anti-tank weapon.
I could have pieced the story together, unaided. I have no idea how long that would have taken. In practice, it took about two days to reach somewhere around the 85% completed mark, because I have the help of an assistant. I use Claude Cowork for archive searching, document retrieval, textual analysis and reasoning.
Before this starts to look like me saying AI is great and everyone should shut up about it and just enjoy the benefits this current investigation has stalled for now, because I’ve reached a familiar point with Claude where I get diminishing returns. I’ve given this a fancy acronym and name. Feel free to adopt it yourself.
IIRFR - Irfer, as I think of it. Claude often goes ‘Irfer’ on me and I’ve had to learn to recognise this as a ‘failure mode’ of AI.
Irfer means illusionary insight from really fast results.
Claude and other AIs do this thing where they overwhelm you with information. Got a problem? Here’s 25 steps to take to investigate and solve it. Would you like me to do x, y and z immediately?
In the case of my current investigation Cowork (Opus 5.5 is my current model) found a lot of salient information very quickly. None of it gave me the one final key to the puzzle - the precise coordinates where Kober, Keith and Gilbert’s bodies lay for 3 or 4 days. Without that data I cannot take the family to the spot where Bill Kober was killed because I cannot be sure it’s where I think it is.
Cowork confidently told me the coordinates would be in the 3 men’s IDPFs (Individual Deceased Person File), so I requested them from NARA. Cowork was wrong. But while it was telling me what to do next I was massively impressed, partly because I had never even heard of an IDPF and also because it seemed plausible. As it happens, only their temporary cemetery grave markers were recorded in the IDPFs. This is typical of AI.
At this point the anti-AI Taliban would be very happy to know that Claude had failed and would probably opine that this makes me a bad researcher. In fact, it’s making me a better researcher and a better user of AI.
I can make it play to its strengths, which are still considerable.
For example, once it has found what I need, it organises it all on my computer, making it possible to properly archive every piece of data in the academic software tool I use (Tropy). This is when the information becomes useful. Until that point I’ve just been amassing noise and this is when I can begin to detect signal. It’s the point at which I form theories and Claude challenges my assumptions, calling it when I prematurely think a question is settled, then suggesting other lines of enquiry until a question is finally nailed down.
That’s one of many ways in which I have come to use AI. Not for writing, because I’m already proficient and enjoy writing. But, for information retrieval and organisation, which I find extremely tiresome, even though I’m also reasonably proficient at, it’s a lifesaver.
The way this all works is by conversing with AI. Give it the context, set some scope, reinforce some rules, set it going, then chat through the findings and what do do with them.
Last weekend I saw someone on Substack Notes state flatly that people who chat with AI are psychopaths.
It’s this yawning chasm between my use of AI - which brings about good things in the world, such as my 24th Cav Recon squadron memorial work - and the spittle-flecked fury of so many people who seem outraged by the mere existence of this extraordinarily powerful tool.
This intemperate reaction to people using AI tools is why I coined the phrase ‘Anti-AI Taliban’.
‘Dave’, the man who described people who have conversations with AI as ‘psychopaths’ is apparently a fiction writer and currently studying for a Bachelors of Creative Writing. So, although his opinion seems kind of moronic, I have to assume that ‘Dave’ himself is not a moron. At least, not in the intellectual sense.
In common with many people, he is apparently having strongly negative feelings about those of us who use AI, which he expresses with sweeping condemnation.
It’s like the exchange I mentioned in part 1 of this exploration. Last time I was highlighting the way AI is perceived through different frames. Back then I was first accused of lying about it being useful. Then, when I proved that it’s useful, the condemnation shifted to my laziness, because I use it ‘just’ to do things faster. Finally, when I asked why saving time is bad, they simply told me to stop ‘hazing’ my AI (a superfluous newish word used to describe promoting or enthusing about something).
What was interesting there was that the goalposts kept shifting. First there was the challenge on the basis of the facts. Once that challenge was met, the challenge was on the basis of moral character. And when that challenge was met the person just issued a command for me to stop talking about it.
What has most sparked my interest in populist chatter about AI is the unsophisticated way in which critics confuse empirical and moral questions. How evidence of unarguably good and beneficial AI use bounces off their frame of it as an entirely moral issue.
This started with the Ted Gioia Note under which people insisted that there are no legitimate uses for AI and that its benefits accrue only to billionaires and sociopaths. Under last weekend’s Note we had fiction writer and academic student ‘Dave’ describing AI use as psychopathic.
If that isn’t closer to a Taliban-style worldview than a dispute about facts, I don’t know what is.
You can’t argue with a non-argument
Their arguments are weak because they aren’t really mounted as arguments at all. They have turned the issue into a moral question, in response to which evidence cannot even be recognised as evidence.
I think they feel about AI the way that I feel about capital punishment, an issue which I am quite intransigent about. You could present me with ample evidence that capital punishment leads to a reduction in violent crime and I would still prefer that the state not kill people in my name during peacetime. All evidence would bounce off my frame in that example, just as all evidence of AI being a useful tool with which we can do public good bounces off theirs.
So it’s not that I can’t see their perspective. But I want them to be more intelligent about it, because their hand wavy claims make for weak reasoning. And there’s nothing worse than weak reasoning being shouted at you.
An example of this weak reasoning is that the charge ‘you could do it all without AI’ applies to any assistance from technology, such as using an electric oven instead of an open fire for cooking, or driving instead of riding horses and walking, or washing your clothes in a machine instead of the local stream.
There’s an interesting philosophical question going on here. At what point - if any - does assistance from technology disqualify a person from being identified as the author or the actor in a situation? But they don’t do interesting philosophical questions because they’re not philosophical thinkers.
This is the fundamental weakness of the anti-AI Taliban, because it’s a peculiarly selective position. When AI has advanced cures for previously intractable diseases, will they object that science could have done this without AI? One must assume that ‘Dave’ would not accuse his doctor of psychopathy if they successfully diagnosed him after consulting with OpenEvidence, a proprietary clinical AI architecture trained specifically on peer-reviewed medical literature.
Thank you for saving my life, doctor. But I think you’re a psychopath for chatting with AI.
Right now, the argument seems to be merely that using AI is ‘wrong’ and ‘bad’ because it isn’t perfect and nor is it indispensable. That’s an impossible bar to clear, if you cared about their approval. Which is, of course, the whole point. Nothing will make them approve.
Another confusion they have is conflating the tool with the user. I think we’re all probably tired of the contribution AI content generation makes to the enshitification of online life. I’m certainly tired of faked images and false stories produced for attention on Facebook, for which ‘creators’ use AI. But this is a user problem, not a failing of the tool. In the end, someone was responsible for that latest bit of AI slop that just annoyed you. And it wasn’t the AI, any more than the AR-15 was responsible for the latest school shooting.
All of this is what makes argument pointless. It’s impossible to resolve a question if you’re not looking at the same question. This often applies to general political disagreement too. Both sides claim that they are evidence-based when really they are using moral frames.
A friend in California posted, following America’s recent midterm elections, an account of why people like him vote for the left. He said that it was because people like him care about other people. Right there, the same issue. It’s not about facts or evidence for what might advance the wellbeing of other people. It’s a moral frame. And moral frames often seem to necessitate moral condemnation of those who don’t share it.
As with politics, so with reactions to AI. Lack of philosophical nuance is a massive force in these areas.
AI is like capitalism, man.
Now that everyone has forgotten about the absolute state of everyday life in the former USSR, it’s fashionable to dismiss ‘capitalism’ as a force for anything good. This is founded on ignorance of what capitalism actually is and focuses entirely on factors that go wrong within the system. Ask most people what capitalism is and they’ll probably say something about maximising profit



