It's no longer OK for you to sit in the house of Lords because your dad did.
But if your mate put you there?
That's fine.
Neither of these really feel like democracy to me.
@kityates
Author. Prof of Math Bio and Public Engagement. Member of Independent Sage since October 2020. @Kit_Yates_Maths on twitter Books - Math(s) of Life and Death How to Expect the Unexpected Get them here: https://tinyurl.com/37rx2yuv He/Him
It's no longer OK for you to sit in the house of Lords because your dad did.
But if your mate put you there?
That's fine.
Neither of these really feel like democracy to me.
"There are signs an outbreak of measles in north London may have stabilised although there could still be more cases, the UK Health Security Agency (UKHSA) has said."
Yes I think thatβs fair to say.
That prospect forces a hard question: if a proof becomes something only a machine can comprehend, does mathematics remain a human endeavour, or does it become something else entirely?
I explore this in more detail in my latest substack here: open.substack.com/pu...
10/10
Pushed to its limit, we may face βobjectively correctβ proofs no human can understand.
Weβve accepted computerβassisted proofs before: the fourβcolour theorem used caseβchecking in 1976; confidence grew with a simpler 1997 proof and a fully machineβchecked version in 2005.
9/10
Pair AI with these systems and you get something new: require outputs in a formally verified language; let the checker expose errors; let the AI try again.
This backβandβforth could scale proof search and uncover links between areas of mathematics humans might not connect.
8/10
To prevent acceptance without correctness, mathematicians are turning to formal verification.
Systems like Lean force proofs into a precise language that a computer checks line by line.
If a step fails, it wonβt pass, removing the linguistic ambiguities humans introduce.
7/10
Consider Fermatβs Last Theorem. We can have 3Β²+4Β²=5Β², but there are no wholeβnumber solutions for cubes, fourth powers, or higher.
After Wilesβs celebrated 1993 lectures, peer review revealed a serious flaw; a yearβs work fixed it. For a while, the world thought it was done.
6/10
Proof, in practice, has always been partly a social construct: arguments are accepted when other mathematicians analyse them and judge them correct.
That means a widely accepted proof isnβt an irrefutable guarantee of truth; even wellβknown proofs may hide issues.
5/10
This raises a more acute issue: what does it mean to βproveβ a result if we canβt follow the proof?
If we canβt trust an argument, we canβt develop new tools and techniques on that foundation.
4/10
Mathematicians now fear being flooded with convincingβlooking proofs that hide subtle flaws. Arguments may be accepted because they look rigorous.
Given a goal, such systems will optimise ruthlessly to achieve itβsometimes cutting corners humans might miss.
3/10
Some worried the model gave highly convincing, but potentially incorrect, answers. In the past, confidence and the appearance of a good argument were useful signals because only the best could make them.
That has changed: style no longer indicates sound reasoning.
2/10
In 2025, leading mathematicians met in secret to test a new AI model. Its proofs sounded like those of a real mathematician presenting a complex argument.
But were we giving it more credit than it deserved? Are we at risk of accepting proofs we canβt understand?
1/10
Going to be on BBC South radio at 13:10 discussing why time seems to go by more quickly as we age.
Here's an article I wrote on the subject in case you're interested.
The VIP lane was rotten to the core.
A national scandal.
I really don't miss having to cover the Telegraph takeover story, the world's most tedious media saga.
But this is INTRIGUING: @danielthomasldn.bsky.social says the Daily Mail deal to buy the Telegraph might collapse and Axel Springer might end up in control.
www.ft.com/content/e5cb...
π
"TRUST YOUR NERDS"
"If implemented, my recommendations should reduce the number of deaths, reduce the suffering and reduce the social and economic cost," she [Baroness Heather Hallett] said.
I've got a feeling that that is a mighty big "If".
βMichael deteriorated in hospital. His cough became so severe he had to send a WhatsApp message to tell her he was being put on a ventilator
Two weeks later, the family was told he would never recover and that doctors were going to reduce his life supportβ
World of Books is selling The Maths of Life and Death (Hardback!) for just Β£6.60.
That's a 67% discount.
Get it while it's hot:
"If implemented, my recommendations should reduce the number of deaths, reduce the suffering and reduce the social and economic cost," she [Baroness Heather Hallett] said.
I've got a feeling that that is a mighty big "If".
Those wonderful moments when the headline has updated, but the picture hasn't.
I had a great time visiting the @universityofexeter.bsky.social of Living Systems Institute (LSI-Exeter) yesterday to give the Distinguished Speaker's lecture.
Interested audience, great questions and a lovely place.
Thanks for having me.
"The number of measles cases in London continues to rise with growing concerns that it is further spreading into other regions."
www.itv.com/news/202...
How to Expect the Unexpected has a random 39% off on amazon at the moment.
Just Β£6.72 for 120K words.
Seems like value for money when you think how much you'd pay to send someone a greetings card!
www.amazon.co.uk/How...
fell for it again award
My dad used to tell me this story. Iβve told it to my kids. So good.
I had a great time (literally) climbing the career ladder yesterday with Max Klymenko.
The University of Bath