ViewTube

Photos from my recent trip to India and Singapore! Here I am at the Taj Mahal, with a cobra in Varanasi, at the Jantar Mantar observatory, and with the Merlion in Singapore. India is such a vibrant country and it was insightful for me to see some of the educational sites (IITs) and study centers that I have covered in previous videos. 

Beatrice Tinsley was an astronomer and one of the great science minds to come from New Zealand. She discovered that galaxies can evolve and change with time. Her body of work was very impactful, including her impressive PhD thesis that I'm working on a video about. I recently visited a street in Auckland named after Beatrice, it's good to see her receive some recognition! 

Last year, I made a video attempting to solve problems from the International Mathematical Olympiad (IMO) using AI. Yesterday, Google DeepMind published a paper detailing their latest attempt, named AlphaGeometry. They managed to solve some IMO-level geometry problems by combining a language model with a “symbolic deduction engine” to find proofs.

The language model suggests potential strategies, while the deduction engine works out the finer details. The DeepMind paper describes this as being similar to the concept of "thinking, fast and slow", made famous by Daniel Kahneman. One system generates fast, intuitive ideas, while the other engages in more deliberate, rational decision-making. This sounds like an interesting way to take inspiration from how humans think. 

This approach was able to make progress by generating a large amount of synthetic training data. The first step was to create and analyse 100 million geometric proofs. This was necessary because there are not many real IMO geometry problems out there. At least nowhere near enough to train a model like this. 

It’s interesting progress! However, since they only looked at geometry problems (which comprise just one section of an IMO paper), they are not yet in the running for a recently announced $10 million prize offered to any AI model that can win a gold medal at the IMO (see  https://aimoprize.com/ ). Steps of progress like this, though, do make me feel that the prize will eventually be claimed. 

The Deepmind paper:  https://www.nature.com/articles/s41586-023-06747-5 

brb going into the woods to think about maths. New video coming soon! 

I've just released a couple of new videos over on my second channel  @tibees2 . That will be the future home for book-related and more experimental videos. Subscribe to stay in the loop 📚🧪
​ 

🎧 I am on the latest episode of 'How To Make A Science Video' hosted by the brilliant duo 
 @SophsNotes  &  @SimonClark :  https://pod.link/1705403433/episode/0ae767a9d728f9021482e04cd40a933a 

It's Bob Ross' birthday today (October 29th) 🕊🎨 He was an inspiration to many, and today's video is inspired by his approach to teaching 🎂 

I've got a new YouTube channel banner 🌻 Thank you to Martina Pepiciello for drawing it for me! 

Behind the scenes photo of my film crew on the set of my next video 🦜
Coming out next week - a return to a Bob Ross inspired video which was a joy to film! 

Behind the info about his stolen grades, this is the second most surprising fact I found on the Alan Turing Wikipedia page...

Text of the image:  "In the 1940s, Turing became worried about losing his savings in the event of a German invasion. In order to protect it, he bought two silver bars weighing 3,200 oz (90 kg) and worth £250 (in 2022, £8,000 adjusted for inflation, £48,000 at spot price) and buried them in a wood near Bletchley Park. Upon returning to dig them up, Turing found that he was unable to break his own code describing where exactly he had hidden them. This, along with the fact that the area had been renovated, meant that he never regained the silver."

P.S. The diploma photo and caption that I showed in my latest video about the stolen grades has seemingly been removed from the page over an image licensing issue.