iiinova — where your digital self keeps evolving

Reposted by Margaret Harris

Philip Ball @philipcball.bsky.social · 16h

When people tell me I'm a fool not to use AI in my writing, I might point them to this. www.cjr.org/analysis/fak...

cjr.org

Fake Caroline.

Editors are facing an influx of AI-written pitches, including at science publications like mine. What now?

7259

Reposted by Margaret Harris

Lee Billings @leebillings.bsky.social · 07/10/2026

ICYMI, the biggest science story this week isn’t the Nobels—it’s OpenAI’s new release that includes multiple Fields Medal-worthy results. The intelligence explosion is sweeping thru math like a wildfire. And it’s probably coming for most everything else, too. What will next week/month/year bring?

256

Margaret Harris @drmlharris.bsky.social · 07/10/2026

Strictly speaking, I guess we could still have three physicsy #NobelPrizes this year if a physicist happens to win in Economics, Literature or Peace, all of which are possible in principle.

140

Margaret Harris @drmlharris.bsky.social · 07/10/2026

Aaaand the team at @physicsworld.bsky.social HQ breathes a sigh of simultaneous relief and disappointment that we're not having three physicsy #NobelPrizes this year. Although chirality is very important in physics, Kensō Soai and Henri Kagan both look to be very solidly organic synthesis guys.🧪⚛️

1120

Margaret Harris @drmlharris.bsky.social · 07/10/2026

I don't know what @pwnallthethings.bsky.social's background is, but I found their recent thread on what the AI-generated proof dump is doing to mathematicians (and why OpenAI is doing it) insightful.

020

Margaret Harris @drmlharris.bsky.social · 06/10/2026

I noped out after the "Scandal in Belgravia" episode, which managed to be vastly less feminist (because we can't possibly have a mere *woman* outsmart Sherlock Holmes, no no no, that's not on) than the original version written by Arthur Conan Doyle *checks notes* 120 years earlier

0110

Reposted by Margaret Harris

Derek Fox @partialobserver.bsky.social · 06/10/2026

Science - always and particularly in the 21st century - is a collaborative human enterprise. IceCube is another example, with LIGO, of a successful long-term investment by the National Science Foundation, which has been subject to vicious attacks and recissions since 2025.

110224

Reposted by Margaret Harris

Geoff Brumfiel @gbrumfiel.bsky.social · 06/10/2026

The IceCube neutrino detector has won the 2026 Nobel Prize in physics. The Trump administration has proposed slashing its budget in half in 2027.

382018706

Margaret Harris @drmlharris.bsky.social · 06/10/2026

Also, we only need two more "Belgian men called Fran-something" to win physics Nobel Prizes before they'll be running equal to "all women not called Marie Curie" in the #NobelPrize sweepstakes...

030

Margaret Harris @drmlharris.bsky.social · 06/10/2026

For Belgium, Nobel Prizes in physics are like buses: you wait 112 years for the first, and then two come along within 13 years. François Englert shared it with Peter Higgs in 2013, now Francis Halzen gets one all to himself in 2026. physicsworld.com/a/how-long-d... 🧪⚛️ #NobelPrize

physicsworld.com

How long did it take for countries to claim a Nobel laureate? – Physics World

Deciding where a prize winner is from can be tricky, but Hamish Johnston tries his best

22610

Margaret Harris @drmlharris.bsky.social · 06/10/2026

"I think we have enough realistic problems to deal with right now to make neutrino astronomy a reality." - Halzen

020

Margaret Harris @drmlharris.bsky.social · 06/10/2026

Could we use cosmic neutrinos to contact extra-galactic civilizations? "This has already been proposed by several friends of mine. It has been speculated that this would be an ideal way to communicate with aliens but I don't think I will go there." - Francis Halzen, 2026 physics #NobelPrize 🧪⚛️

122

Reposted by Margaret Harris

Physics World @physicsworld.bsky.social · 05/10/2026

Karl Deisseroth, Peter Hegemann and Georg Nagel win Nobel Prize for "discoveries concerning light-gated ion channels and optogenetics" - a breakthrough that enables mapping of the brain in a way that we could once only dream of. 🧪⚛️ physicsworld.com/a/inventors-...

physicsworld.com

Inventors of optogenetics win Nobel Prize in Physiology or Medicine – Physics World

Karl Deisseroth, Peter Hegemann and Georg Nagel are joint winners of the 2026 prize

2187

Margaret Harris @drmlharris.bsky.social · 03/10/2026

I also recall a Newcastle establishment called Eat Drink Dance which my Sunderland-born housemate wisely suggested ought to be called Eat Drink Dance Throw Up.

000

Margaret Harris @drmlharris.bsky.social · 03/10/2026

When I lived there, the last train from Newcastle to Durham on a Saturday night was ludicrously early, yet it was always packed with people who’d somehow managed to get heroically rat-arsed before 9pm. Glad to see it hasn’t changed!

260

Reposted by Margaret Harris

Physics World @physicsworld.bsky.social · 02/10/2026

How confident, or perhaps delusional, do you have to be to travel to Stockholm because you’re so certain you’ll be awarded a Nobel prize? One person who did just that was the Indian physicist C V Raman in 1930. Preetish Kakoty recounts how he did end up winning. 🧪⚛️ physicsworld.com/a/c-v-raman-...

physicsworld.com

C V Raman: the physicist with a 'Himalayan ego' who went to Stockholm convinced he’d win a Nobel prize – Physics World

Preetish Kakoty describes why India's only Nobel-prize-winning physicist went to Sweden on a whim

2144

Reposted by Margaret Harris

Andy Extance @andyextance.bsky.social · 02/10/2026

After my partner died of cancer, the fact I write about new treatments for it in my work has been complicated. Inspired by this, I am launching an Instagram account called The Widowers Guide to the Science of Cancer today - you can see the first post here: www.instagram.com/p/Dd-__vcALl... 🧪

1218

Reposted by Margaret Harris

Physics World @physicsworld.bsky.social · 30/09/2026

Physics World has had some luck with predicting who is going to win a Nobel Prize in Physics. Will we get it right in 2026? 🧪⚛️ ow.ly/S63G50ZTaTk

ow.ly

Is this the year for a Nobel prize in condensed matter, or maybe particle physics? – Physics World

We make our prize predictions for 2026

094

Margaret Harris @drmlharris.bsky.social · 01/10/2026

I encountered one of these guys leafleting outside a railway station shortly before the vote and wasted time remonstrating with him. A few days later, I broke my foot tripping down the stairs at the same railway station, and honestly, I don't know which of those things ruined my day more.

040

Margaret Harris @drmlharris.bsky.social · 29/09/2026

Sometimes I get to write a story that combines my ❤️of history with my ❤️ of physics, but stories that also incorporate my ❤️ of trains are as rare as leaves on the line in February. So here goes: did space weather delay the 10.05pm from Exeter on 18 October 1841? 🧪⚛️ physicsworld.com/a/did-space-...

physicsworld.com

Did space weather delay a train in 1841? – Physics World

It would have been the earliest example of space weather impacting human technology, but an international team of researchers found a slight problem

153

Reposted by Margaret Harris

Retraction Watch @retractionwatch.com · 24/09/2026

"If medical schools and universities don't want the literature to be overwhelmed with garbage," Oransky said, "then they should probably stop giving every applicant ... an incentive to publish a lot of crap."

medpagetoday.com

Is AI Flooding Medical Journals With 'Meaningless' Research?

'It's not fraudulent, and it's not really slop,' one expert says

08729

Reposted by Margaret Harris

Richard Morris @ahistoryinart.bsky.social · 24/09/2026

Do you have a painting or drawing by Lawrence Bragg? The Nobel Prize-winning scientist appears in the 1927 Solvay photo alongside Einstein and Marie Curie; he was also a keen painter. I’m planning an exhibition about his life and times in 2027, the centenary of this picture.

26321

Margaret Harris @drmlharris.bsky.social · 25/09/2026

What if kids getting sick all the time in winter isn’t *only* because their immune systems are still developing? What if it’s *also* because their schoolrooms have terrible air quality?

081

Margaret Harris @drmlharris.bsky.social · 24/09/2026

British by naturalisation, one cup of tea, no coffee

011

Margaret Harris @drmlharris.bsky.social · 24/09/2026

I don’t think he’d disagree with any of those points, but it can be all those things and still be helpful in certain cases (as it was to the patient in his study)

100

Margaret Harris @drmlharris.bsky.social · 24/09/2026

I asked afterwards which fields of physics were likely to be most vulnerable to an AI takeover. His response is that certain fields of formal, classical theoretical physics are likely to come first. The further you are from the messy real world, the greater the risk.

000

Margaret Harris @drmlharris.bsky.social · 24/09/2026

His answer is that, when analysing an extremely long and complex patient history (a necessary first step to getting the right care, not just more care), an AI is vastly faster than even a patient, caring and well-resourced human doctor would be. So it’s not strictly a resources problem.

100

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Skjuve asks whether the young researchers in the room feel they’re responsible if an AI messes up while writing a paper for them. Most hands go up. And are they confident they can catch all such errors? Fewer hands. Skjuve says her research tells her that people are in fact not equipped to do that.

020

Margaret Harris @drmlharris.bsky.social · 24/09/2026

No, but that’s less likely to be salient for an audience in Norway, where the social safety net is much stronger (though imperfect) than it is in many other countries. Good ? though, I’ll ask if I can

200

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Who is accountable when uses of AI go wrong? Companies? Users? Malthe Sørenssen thinks we currently have a system that considers both intentions and actions when assigning accountability, but AI overturns that. He expects it’ll get sorted eventually but right now it’s all up in the air.

110

Margaret Harris @drmlharris.bsky.social · 24/09/2026

“The way that AI behaves is completely perpendicular to the way we are used to doing research.” Need more regulation to avoid exploitation, says Malthe Sørenssen

110

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Rather than worry about AI replacing all human connections, Storhaug Reinås thinks it can augment them. In healthcare there is so much demand for more complex care, it’s overwhelming, so if technology can help, we should use it. Mentions a lupus chatbot, LupusGPT, that answers patient questions.

210

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Skjuve: “Everyone says they care about privacy until they want something, and then privacy goes out the window so they can get what they want.” She’s talking about individual users but it seems to me that it also applies to AI providers

110

Margaret Harris @drmlharris.bsky.social · 24/09/2026

“If you’re not using the top AI models in physics, you’re going to very quickly fall behind.” Malthe Sørenssen says he’s stopped programming since January because AI code is now good enough. “It’s going to come for all of us.”

210

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Patient B fed her entire medical record into ChatGPT, & though it proved a wonderful tool for her, Storhaug Reinås is hugely worried about security. Echoes Skjuve’s earlier call to move away from general purpose AI towards AI that is tailored for specific applications w/ appropriate guardrails.

110

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Panel discussion now. Moderator asks about the recent OpenAI hacks. “I don’t really trust them,” says Malthe Sørenssen. Says they are bad at cybersecurity and it’s not entirely clear whether that’s deliberate or accidental

100

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Take-home messages: trust is relational and people will trust individual humans before they trust the system those humans are working within. “Trust in science is not built through evidence alone,” Storhaug Reinås says

010

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Where does the AI come in? Patient B found that the best source of help for her was…ChatGPT. It helped her prepare for consultations and organize her notes. The fact that all this information went to a potentially untrustworthy company didn’t come into it, Storhaug Reinås says

100

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Both patients experienced problems with trust, says Storhaug Reinås, but the source was completely different. Patient A learned that trust means accepting that certainty isn’t always possible. Patient B had to learn to trust her own experiences and be confident in them despite medical dismissal.

321

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Patient B had a long history of multiple diagnoses and struggles to figure out what was wrong with her. This led her to not trust the physicians because they couldn’t figure out what was wrong with her and kept putting her through unnecessary treatments.

100

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Patient A was an entrepreneur, liked being in charge, was suddenly told he probably had pancreatic cancer. This diagnosis was presented as a certainty, but it was wrong – which made him very wary about trusting doctors.

100

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Trust is not only created through certainty, Storhaug Reinås says. It grows through listening, recognition, honesty and shared decision-making. So, what role can AI play in that trust relationship?

110

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Chronic illness is not only a medical problem, it’s also an information and trust problem. How do we connect a person’s knowledge of their own body to the corpus of medical knowledge, especially when that information is incomplete? And how do we preserve trust while we do it?

120

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Next is the thorny issue of AI in healthcare, specifically for patients with chronic illness, and how it relates to trust in the medical system. Speaker is Karl Henrik Storhaug Reinås in the education department at Oslo.

Speaker in front of slide described in post

100

Margaret Harris @drmlharris.bsky.social · 24/09/2026

The risk of having information exposed exists with any technology – the Ashley Madison hack in 2015 for example – but we share so much more with AI and “we expose more aspects of ourselves because we think it is only between us and the AI,” Skjuve says

110

Margaret Harris @drmlharris.bsky.social · 24/09/2026

People develop trust for their AI companions even when they don’t trust the provider that created it, Skjuve says. “This is very complicated to research.”

150

Margaret Harris @drmlharris.bsky.social · 24/09/2026

Marita Skjuve of SINTEF is now talking about personal – sometimes *very* personal – relationships with AI. “It’s not that you forget it’s not human, but you begin to treat it in similar ways. Will this AI judge me if I share my secrets? Will it leave me if I have sexual relations with someone else?”

Speaker and slide showing a bunch of apps promising AI girlfriends and boyfriends

110