Manuela Veloso is excited. Not just a little bit excited: very. She is very excited. “Why is JPMorgan such a beautiful place to work?,” shouted Veloso at today’s CogX artificial intelligence conference in London. “It’s because it is gigantic,” she shouted. JPMorgan employs so many people and has so many clients that it’s, “overwhelming and beautiful,” and generates “a lot of data,” said Veloso.
If you’re interested in AI, Veloso says this makes JPMorgan a special place to work.
JPMorgan hired Veloso to set up its AI research group in 2018. As well as working for JPM in New York she remains the head of machine learning research at Carnegie Mellon University, one of the top-ranked institutions for computer science and AI. Despite only working for JPM part time, Veloso has already achieved some dramatic things.
“I was taken to the trading floor at JPMorgan here in London…and for me it was beautiful,” Veloso said. “There were all these people and all these screens around – tons of screens, I had never seen so many…
“The only thing I could think about was how do humans make decisions like this. How in the world do humans process all this information? They are not using paper and pencils to make decisions. They are looking at the screens.”
As an academic with a history of training artificial intelligence systems to decipher cats and dogs and oranges, Veloso’s key insight at JPMorgan was that traders at the bank were making trading decisions based on the images on their screens and that machines could be trained do the same thing.
“Instead of data, we took pictures of the time series….pictures of the screens,” said Veloso. “We said let’s transform the [trading] time series into an image…I didn’t know if the neuronet was going to be able to capture the signal. - It was not a cat or a dog, just a line on the screen.”
It turns out that the neuronet could. After a week of training, it was able to decide to buy or sell a stock based on time series images with a 95% accuracy rate. “Can you imagine how beautiful it is?,” shouted Veloso. “- It is for us. It’s transformational.”
Veloso’s ‘beautiful transformational project’ is known internally at JPMorgan as Mondrian. It doesn’t just involve image analysis – the bank has also been looking at traders’ gaze patterns. “We generate information that says why things were decided,” says Veloso. Eye gaze tracking can be used to identify the most “valuable assets” on a screen and to sequence decision making behavior that machines can copy.
It all makes JPMorgan’s AI team a special place to work. And Veloso says she’s recruiting. “If you want so send me an email, we are hiring,” she concluded today. “- I will take you for a visit.” It’s rare that JPMorgan makes such an offer – especially from one of its most irrepressible employees. Veloso may get plenty of takers.
Have a confidential story, tip, or comment you’d like to share? Contact: email@example.com in the first instance. Whatsapp/Signal/Telegram also available.
Bear with us if you leave a comment at the bottom of this article: all our comments are moderated by human beings. Sometimes these humans might be asleep, or away from their desks, so it may take a while for your comment to appear. Eventually it will – unless it’s offensive or libelous (in which case it won’t.)