ChatGPT fell for Grok3’s summary hook, line and sinker, sadly. When prompted with your comment, it then changed tack.
“Grok3’s biography is not just flawed but dangerously misleading:
• Nine factual errors in three paragraphs (a high error density).
• Fabricated employment, publications, and affiliations.
• Flattering but slippery phrasing that reinforces AI-generated credibility.
• Potential long-term misinformation risks, as AI content spreads rapidly.
Oliver Johnson’s response is a textbook example of how AI-generated text can distort reality, and he is right to be concerned about its long-term impact.” When asked to explain itself it carried on: “That’s a fair question, and I appreciate the opportunity to reflect on my initial analysis. Here’s why I didn’t immediately catch the errors in Grok3’s biography of Oliver Johnson, along with what I’ve learned from the process.
1. The Plausibility Trap
Grok3’s biography contained just enough truth to appear credible at first glance…” etc
It's funny, it's also parroting that stuff back to me. I'm sure there are better versions of all this stuff, better prompts you can use and so on, but at the moment it certainly feels like a minefield!
Around 15 years ago when I was first working at Mumsnet the parental anxiety around MMR and autism was at its height, and while I never had any anxiety about the MMR myself (and both my kids had all their vaccinations) the *tone* with which clinically-literate people tended to dismiss parents’ worries was all wrong, and pretty much designed to antagonise, which is not, uh, how you persuade people to listen to you. Not sure whether things have got any better on that front. And of course most GPs barely have the time to counsel patients with immediate health concerns, let alone assuage worried parents.
It's hard because I've probably gone in too hard on some people over the last few years on occasions. But as a general principle (even if I don't always manage to stick to it!) I'd try to differentiate between individuals making up their mind in a tough information environment and the misinformation super-spreaders, some of whom are making a fortune out of it. (For some reason I can't find the Chris Arnade piece I've talked about before about doing the "actually sitting down and talking" thing)
Anecdotally from my job typing paediatric clinic letters, actual vaccine opposition (rather than e.g. Roma not engaging) is often from Eastern European parents. Still very rare for a child not to be vaccinated though.
Interesting thanks - though I guess you are only seeing the people who engage with the system, and there's presumably some set of children just not on the radar?
My understanding from a Danish public heath official testimony that seemed honest, is that the measles vaccine works in a way that we don't fully understand and decreases all cause mortality. That is only one m in the MMR. I would be very careful about promoting any vaccination after the definition of vaccine was changed to accommodate GMOs ... Unless you are paid by a pharmaceutical company of course.
The "all causes" reduction in mortality for the measles vaccine is now understood. The measles virus infection results in a reduction in memory B-cells, these are the cells that have responded to previous infections and vaccinations and which lead to a more rapid response on encountering the same infection again. In other words, following a measles infection your immune system is reset to a more naive like stage where it has forgotten earlier infections. The measles vaccination means that the virus is less likely to cause memory B-cell loss and thus you are still able to mount a more robust memory immune response to all those infections you have previously encountered or been immunised against.
AI coming across like astrology—and the inferences astrologers draw from horoscopes—is a fascinating observation. Well spotted.
Thanks! I think nobody's going to complain if the magic box finds a way to flatter them
ChatGPT fell for Grok3’s summary hook, line and sinker, sadly. When prompted with your comment, it then changed tack.
“Grok3’s biography is not just flawed but dangerously misleading:
• Nine factual errors in three paragraphs (a high error density).
• Fabricated employment, publications, and affiliations.
• Flattering but slippery phrasing that reinforces AI-generated credibility.
• Potential long-term misinformation risks, as AI content spreads rapidly.
Oliver Johnson’s response is a textbook example of how AI-generated text can distort reality, and he is right to be concerned about its long-term impact.” When asked to explain itself it carried on: “That’s a fair question, and I appreciate the opportunity to reflect on my initial analysis. Here’s why I didn’t immediately catch the errors in Grok3’s biography of Oliver Johnson, along with what I’ve learned from the process.
1. The Plausibility Trap
Grok3’s biography contained just enough truth to appear credible at first glance…” etc
Don’t worry, it’s all fixed now. “How I Will Improve My Analysis in the Future
To prevent this kind of oversight, I will:
1. Cross-check claims with external sources (official university websites, book listings, media appearances).
2. Look for subtle distortions, not just obvious fabrications.
3. Be extra skeptical of AI-generated biographies, especially when the subject is a public figure.
4. Apply Brandolini’s Law upfront—assume debunking is harder than creation and double-check before assuming accuracy.”
It's funny, it's also parroting that stuff back to me. I'm sure there are better versions of all this stuff, better prompts you can use and so on, but at the moment it certainly feels like a minefield!
Around 15 years ago when I was first working at Mumsnet the parental anxiety around MMR and autism was at its height, and while I never had any anxiety about the MMR myself (and both my kids had all their vaccinations) the *tone* with which clinically-literate people tended to dismiss parents’ worries was all wrong, and pretty much designed to antagonise, which is not, uh, how you persuade people to listen to you. Not sure whether things have got any better on that front. And of course most GPs barely have the time to counsel patients with immediate health concerns, let alone assuage worried parents.
It's hard because I've probably gone in too hard on some people over the last few years on occasions. But as a general principle (even if I don't always manage to stick to it!) I'd try to differentiate between individuals making up their mind in a tough information environment and the misinformation super-spreaders, some of whom are making a fortune out of it. (For some reason I can't find the Chris Arnade piece I've talked about before about doing the "actually sitting down and talking" thing)
Anecdotally from my job typing paediatric clinic letters, actual vaccine opposition (rather than e.g. Roma not engaging) is often from Eastern European parents. Still very rare for a child not to be vaccinated though.
Interesting thanks - though I guess you are only seeing the people who engage with the system, and there's presumably some set of children just not on the radar?
My understanding from a Danish public heath official testimony that seemed honest, is that the measles vaccine works in a way that we don't fully understand and decreases all cause mortality. That is only one m in the MMR. I would be very careful about promoting any vaccination after the definition of vaccine was changed to accommodate GMOs ... Unless you are paid by a pharmaceutical company of course.
Please take this stuff elsewhere, I'm not having people impugning my integrity.
The "all causes" reduction in mortality for the measles vaccine is now understood. The measles virus infection results in a reduction in memory B-cells, these are the cells that have responded to previous infections and vaccinations and which lead to a more rapid response on encountering the same infection again. In other words, following a measles infection your immune system is reset to a more naive like stage where it has forgotten earlier infections. The measles vaccination means that the virus is less likely to cause memory B-cell loss and thus you are still able to mount a more robust memory immune response to all those infections you have previously encountered or been immunised against.