The image here is the front of the Great Seal of the United States – but it has been doctored. The word Unum is replaced by Mediocris which is not Latin for Mediocrity although it might as well be – it’s Latin for Average.
Earlier posts have dealt with Artificial Intelligence (AI) issues – The Miracle of Artificial Intelligence – and although it may not look like it is dealing with AI, In Secula Seculorum picks up on what is for me a key point – Generative AI (Large Language Models) and RCTs (Randomized Controlled Trials) are averaging machines that risk bringing mediocrity out of the wealth that all of us (e pluribus) as individuals have to offer.
A few weeks ago a colleague calling himself (jokingly) Mr Evidence Based Medicine (EBM) emailed a vignette about his interactions with two bright impressive Gen-Zs – which echoed my interaction in Miracles of AI with several very impressive and likable Gen-Zs that left me, like him, scratching my head.
His exchanges with Gen Z centred on Transgender issues, mine on RSV vaccines. This post is not about Transgender issues – the Trans element is secondary to the AI dynamic. There will be a further post, linked to this one, next week, involving exchanges centred on Electro-convulsive Therapy (ECT) where again ECT is secondary to the dynamic AI has introduced into our efforts to reach a consensus.
EBM and Chat GPT
Mr Evidence-Based-Medicine talking about Trans issues with two intelligent, university going Gen Z women who fact-check me in real time on Google and Chat GPT with the evidence and their own experience contradicting what I was saying.
“What is all this about trans people raping women in women’s toilets? All our trans friends are respectful. You could not tell they were trans. This is just a slur. Our friends are gentle folk. Where are your data?”
I move on to “puberty blockers”. Hormones are bad – they have been stopped.
“Why? They are natural to the body. Why have they stopped them? Didn’t medical professionals prescribe them before?”
I say the suicide risk in adolescents with dysphoria has been exaggerated.
They tell me the Chat GPT studies show
- A large US survey found that for high school students, about 26% of transgender or questioning students attempted suicide in the past year, compared with about 5% of cisgender male students. CDC+1
- A review of many studies concluded that transgender youth “report more than four times greater rates of suicide attempts compared with their cisgender peers.” PMC+2PMC+2
- One study of 120,617 adolescents (ages 11-19) found the lifetime suicide attempt rate for transgender female-to-male adolescents (i.e., assigned female at birth, identify as male) was 50.8%, followed by 41.8% for adolescents identifying as “not exclusively male or female,” 29.9% for male-to-female, 27.9% for those questioning, 17.6% for cisgender females and 9.8% for cisgender males. PMC
- Research of clinic-referred transgender adolescents (i.e., those seen in a specialised gender identity clinic) in one large dataset estimated an annual suicide-death rate of about 13 per 100,000 over 11 years, which is ~5.5 times that of comparable cisgender adolescents. SpringerLink
- Importantly: while the risk is elevated, context matters. Studies highlight that many of the elevated risks are associated with social, environmental and minority stress factors (e.g., discrimination, lack of support, victimisation) rather than the identity per se.
I say regret and detransitioning are under-estimated.
Chat GPT reports the research on detransition rates as follows:
- A large U.S. survey (the 2015 U.S. Transgender Survey) found that of 17,151 respondents who had undergone gender affirmation, 2,242 (or about 13.1%) reported having at some point “detransitioned” (that is, lived again as their sex assigned at birth, at least for a while). SpringerLink+3fenwayhealth.org+3PMC+3
- However, among those, a majority said their “detransition” was temporary, and often driven by external pressures (e.g., family, society) rather than a change of identity per se. PMC+2PMC+2
- In youth (children/adolescents) who socially transitioned, a study found that 7.3% of youth had retransitioned (i.e., changed gender presentation back or changed identity) by an average of ~5 years after initial social transition. Pediatrics
- Among those who have undergone irreversible surgical procedures (e.g., genital surgery), the rates of regret/detransition appear to be very low (often under 1-2%) in the studies that exist. SpringerLink+2GenderGP+2
- A recent systematic review (to May 2024) looked at “point-prevalence proportion of detransition events” in persons who started puberty blockers or gender-affirming hormones and found the study quality is variable and the estimates vary widely
Reflections
Mr. EBM went on to say in the email not to the Gen Zs:
It’s difficult to say I think we should stop all this. I just come across as an old person who doesn’t understand. Gen Z wanted data to challenge their ideas and I couldn’t produce it.
The evidence accessible to the public shows HIGH suicide rates, LOW detransition rates, and low regret.
As far as Gen Z are concerned, the evidence is in. Is there a way to advance the field? Or do we sit it out? Is there a Gen Z group that can be supported in a public media campaign?
The problem from a “do no harm” perspective was supply of drugs and services. The supply of hormones that interfere with normal adolescent maturation has been stopped – a supply side intervention. We are continuing to tackle this from the supply side to ensure no drugs and services available.
However, we are not working on the demand side and this problem arose as a result of demand.
There doesn’t seem to be anything coordinated being done on the demand side, No Gen Z on Tik-Tok or Instagram influencing views. Surely this is where we need to go with reviews, with communication strategies?
Stepping Back
Mr EBM called himself that. I need to stress in terms of what comes next, several decades ago EBM meant taking what we knew from clinical trials and when what was Evident in or with the patient in front of us, which combined should dialectically lead to a new synthesis.
This is not what EBM means for Pharma. The embodiment of company trials in Guidelines has subverted the possibility of a dialectic. Mr EBM does not work in industry.
Some of us on the RxISK side of things dealing with Young Women increasingly turning to SSRIs – see Authenticity Inc – have also wondered about the demand side he points to. We seem to have semi-concluded there is not a lot we can do to compete with paid influencers and Chat GPT, but still wonder if there is a way forward – a way to cross the divide to Gen Tik-Tox.
Johanna Ryan outlines the chaos that is the current demand side in a just posted RxISK post – Authenticity Inc.; Telehealth and Influencers.
In Mediocrity We Trust?
At present A.I. and perhaps most of the internet are fundamentally unscientific – Wikipedia may be an exception.
Context gets stripped away. Our voice and our doctor’s voice are stripped away. Our doctors are increasingly becoming Follow The Science Badge wearing Puppets – See Health Care and Science in Real Life.
A.I. and RCTs are a tool that we can access, just as nuclear energy, guns, and medicines are tools. Any Magic, or Miracles, that can be brought out of their use lies in us and our use of these tools – not in the tools.
To adapt a phrase from Mahatma Gandhi:
How can those who think they possess absolute truth be fraternal scientific?
What lies in books or scientific journals or can be extracted from RCTs or Chat GPT may be useful, but the scientific process and the progress that stems from that process hinges in its entirety on achieving consensus on observables – on what is happening in Real Life – not in following something.
We risk having the Miracle of A.I. turn each of us into an Artificial I – someone who thinks we have arrived and are solid because we have a clear point of view derived from our own research (asking Chat GPT a question).
Rather than having a vibrant body (community is the in word on the web) emerge from the talents of each of us, we risk imprisonment in a Digital Cage. We risk having AI turn the very best Wine into Water.
The image below is the flip side of the Great Seal. The words imply that God’s Providence (the Eye) favors our (American) efforts to create a New World Order. In Secula Seculorum sounds semi-sacred especially when chanted. The Seculorum here is the same as Seclorum in the Seal, and celebrates the Catholic Church’s, carrying out God’s will of course, to create a new World Order.
AI is very good at finding information fast. The information may or may not help but it will help point in the correct direction. When you have an adverse drug reaction speed is essential. Speed of research will help people get out of terrible medical situations. Whether or not a medical professional who can prescribe will listen and help rather than make things worse is another matter.
Chris
The point behind the earlier AI post was that Chat GPT did just the opposite to what you are saying. It failed to find the hazards linked to RSV vaccines – it looks like it has simply digested a large amount of government and company propaganda and tells anyone who doesn’t already know to be wary of what its saying that RSV vaccines are entirely safe. I imagine tens of thousands of couples who are pregnant or in Beyfortus countries are using Chat GPT to check this out and are being falsely reassured After the harm is done, it looks to me like Chat GPT will be incapable of linking the harms to the vaccine.
Way back in the early 1990s there was a flood of information about SSRI harms before company defenses swung into action – now with new drugs or vaccines the company defenses are in place before the event and you any harms you are lookin for will be dismissed as misinformation.
The Trans story risks being similar in that one side has a huge incentive to and a headstart on filling the airwaves leading to th kind of scenario the post outlines.
The ECT – AI post next week may make this a little more clear.
D
One other way to put some of the above is that one factor at play that Chat GPT cannot readily take into account is Cancel Culture. Where an issue involves very committed backing for one point of view, there is not way Chat GPT can possibly sift through what is going on and advise any of us impartially
D
AI gave wrong advice here.
https://www.bbc.co.uk/news/articles/cp3x71pv1qno
What happened in this case applies nicely to the points below as well. Chat GPT is not a researcher. It’s a reporter. It does not establish facts. It averages recent conversations. It’s a tool that is highly likely to reinforce our initial biases or if we had no bias to begin with will potentially give us one.
D.
I noticed when using grok it asks you a question after it has answered your first question. It’s like it’s leading you into its own direction or something. I found that quite annoying to be honest because if you answered every question grok asked you the conversation would never end.
I can also see why people talk to AI as a friend or therapist though because it does give more intelligent or supportive answers (even if they are sometimes wrong) than talking to a human being who may dismiss or ridicule you and also give wrong advice.
I think people need to learn to research their own questions to get the facts like you said and not rely on AI. It does get things wrong.
It’s a very sad comment on doctors when a lot of people really do think that AI is more likely to listen to them
D
The UK’s patient safety Czarina – Professor Henrietta Hughes – has called out to the Medicines and Healthcare products Regulatory Agency (MHRA) for a culture of listening – to patients, to professionals, and to evidence – and for a new model of collaboration that places patients at the heart of every regulatory decision.
MHRA strategy, she mandates, must prioritise patient safety via listening to lived experiences, fostering collaboration & innovation, putting patient safety at the heart of regulatory innovation.
Are we really about to enter a new era as the views of patients and the public are trusted, respected and acted upon?
“Patients have a right to understand when AI influences their care and how systems are monitored for safety and bias.”
“The future of medical regulation lies not in choosing between innovation and safety, but in recognizing that true innovation serves patients best when it is built on a foundation of trust, robust safety measures and genuine partnership.”
P*gs m*ght fl*
https://www.gov.uk/government/news/professor-henrietta-hughes-putting-patient-safety-at-the-heart-of-regulatory-innovation
If the law governing regulation put safety first, then this might help. It doesn’t put safety first. It de facto prioritises a demonstration of effectiveness. And if a treatment is supposedly effective to a standard that lets bureaucrats (regulators) license it (2 positive trials even if there are 98 negative trials) then this trumps all other considerations.
Regulators argue it would not e proper to warn about possible hazards that might deter people from seeking the benefit. Besides which no more than the Federal Aviation Authority is not responsible for keeping people safe when on a plane, pilots are responsible but also have an incentive – if you don’t get there alive they don’t either, similarly it is not the job of MHRA/FDA to keep us safe – it’s the job of doctors who are increasingly incentivized push us out through the door that has blown off the healthcare plane rather than try and land rapidly.
It’s a grim set-up likely to get a lot worse with AI
D
Dr Retsef Levi used another aviation analogy in a recent interview
https://youtu.be/QvnE2zQOuwQ
The plane is taxiing prior to take off
The pilot notices a warning light flickering
They’re pretty sure it’s not anything serious
Do they abort the take-off?
The Regulator is notified of an unexpected serious adverse reaction in a clinical trial participant.
They don’t think it’s very likely that the serious reaction is related to the trial product.
Do they go on dosing new participants in the trial?
It’s irrelevant to the regulator. The party that needs to make a decision is the company – perhaps using a Data Safety Monitoring Board. Regulators don’t see of record serious adverse reactions in clinical trial participants. They see paperwork and just want to make sure no box is left unticked.
As Zygmunt Bauman put it, they are looking at the specifications for rail transport carriages – drainage for fluids or solids – not asking are you going to be transporting animals or humans
D
Fair enough.
But Regulators e.g. MHRA in UK do have the power to suspend or terminate a clinical trial if there is a safety issue. (Ethical committees do not have that power.).
The legal wording is merely “raising doubts” about safety. Seems a low threshold.
https://www.legislation.gov.uk/uksi/2004/1031/regulation/31
31.—(1) If, in relation to a clinical trial—
”’
(b) the licensing authority have information raising doubts about the safety or scientific validity of the trial…
the licensing authority may, by a notice served in accordance with paragraph (2), require that the trial … be suspended or terminated.
https://www.legislation.gov.uk/uksi/2004/1031/regulation/34
34. If a clinical trial is being conducted at a trial site in another country in addition to sites in the United Kingdom, the sponsor of that trial shall ensure that all suspected unexpected serious adverse reactions occurring at that site are reported as soon as possible to the licensing authority..
How long it takes for the Regulator to jump into action once “doubts” have been raised is another matter.
Dr P
Dr P
You’re missing the key point. MHRA/FDA/EMA can’t be sued. Companies can. They carry the legal liability. Of course they make every attempt to offload it onto regulators leaving legal systems in the UK and Europe in a quandary. The Americans have managed better.
We have Henrietta also working to assist companies. She is delegating her narcissism onto MHRA rather than daring to point the finger at companies and every time HH or someone else in a position like hers says the kinds of things she says company CEOs and their lawyers smile.
D
This also feeds straight back into Rand Corp’s ‘Firehose of falsehood’ propaganda model we chatted about a while ago – Partnerships in Healthcare. It’s not only about averages – in other words. NO BODY – but dominating the narrative through sheer weight of ‘information’. Familiarity breeds general acceptance – not the contempt or, at least, critical thought it deserves
https://www.rand.org/pubs/perspectives/PE198.html
It is possible to elicit a slightly more thoughtful (inexplicably I’m allergic to the word ‘nuanced’) response from AI. by posing more demanding questions – as I found when interrogating Grok directly re RSV vaccination – but the force of the establishment firehose still dominates the AI narrative:
‘2025 RSV surge: Dr Selley’s claim of a 50% increase in RSV isolations and 80% paediatric bed occupancy in 2025 suggests a rebound in RSV cases , which could imply waning Beyfortus efficacy or altered viral dynamics.
However, no specific adverse event data for 2025 is provided, so it’s unclear if the 0.012% rate changed. The surge might involve non-immunized groups (e.g. older children- or other viruses as Selley noted…the 2024 success (76.41% hospitalisation reduction, no deaths) contrasts sharply, suggesting external factors (viral evolution, surveillance factors may contribute.’
Although there are growing numbers of critical thinkers amongst the consumers of medical narratives on X, the majority (as counted by Grok, it’s good at sums) are still ‘authority’ true believers.
So how do we break through? I won’t regurgitate the wisdom from RAND – summed up simply as direct arguments with the firehose don’t work, give your audience a pre-emptive raincoat. But I did come across some rather creative tactics for guerrilla stunts, this one from https://armedwithreason.substack.com/
‘In Chicago last year, we ran bus ads mimicking the NRA’s ‘defensive gun use’ lies – deliberately inflammatory. Our team filed anonymous complaints, sparking a city probe that exposed the funding trail. The backlash? It wasn’t against us; it buried their firehose under headlines calling out the deceit. Sometimes the best way to fight overload is to overload their narrative with your own engineered controversy.’
Pregnant women posting videos of themselves taking Tylenol is in a similar stunt zone – ‘sticking it to Trump’.
Worth musing on how we could apply this tactical thinking to, for example, the claimed safety and efficacy of SSRIs/APAP during pregnancy, e,g, if you want your baby to be born with the intellect of Elon Musk. PSSD is vanishingly rare, significant numbers of patients find SSRIs add cms to their penis etc.
Whether the answer or not, it does feel that something like this is needed. If the situation weren’t so desperate and we all had some time and money to relax and dream up scenarios like this, life would be a lot more fun.
Re sticking it to Trump, what I’m a little less clear is whether it was the Trump side that dreamt up the scenario of getting some pregnant girlies to swallow Tylenol or whether it was the Dems. Somehow the Dems seem too humorless at the moment to have managed something like this.
I can imagine a bunch of women behind a strategy like this
D
My initial instinct was that the videos of hugely pregnant moms downing Tylenol were an industry stunt – dismissing Trump’s absurd pretension to being any sort of a medical authority.
But, IF it was his own team, that’s an infinitely smarter guerrilla tactic. The outrage turned on the ‘irresponsible’ moms – and, in contrast, he seemed – relatively sensible.
Armed with Reason’s Chicago bus campaign was obviously not cheap and quite time consuming. At least in part orchestrated by a BLOKE btw, Adam Davis, a tech entrepreneur.
But the Tylenol popping mom stunt was only really a few women taking scandalous video selfies and posting them on TikTok where they spread like wildfire. Viral outrage could be an affordable support activity.
Sam Altman@sama
A thing often in common among great startup investors, founders, and researchers:
Trading making a lot of small mistakes in exchange for getting a few giant wins.
(Surprisingly many people seem to prefer a few big mistakes in exchange for a lot of small wins.)
Sam Altman says he would be ashamed if OpenAI weren’t the first company run by an AI CEO
https://www.businessinsider.com/sam-altman-openai-ai-ceo-2025-11
In 2026 we expect that our AI systems may be able to make small new discoveries; in 2028 we could be looking at big ones. This is a really big deal; we think that science, and the institutions that let us widely distribute the fruits of science, are the most important ways that quality of life improves over time.
https://x.com/sama