People: High-performance teams

How can small groups of exceptional people solve significant problems, create new technology, and make important decisions?

Machines: But how do you define intelligence?

This is the single most common question I get when talking about artificial intelligence. It’s a good question, but unfortunately it is always said in a dismissive tone. Beyond jokes about the robot apocolypse, “But how do you even define ‘intelligence’?” is the sceptic’s favourite smart response to the topic of AI. The implication being “You cannot”. Perhaps even “And therefore AI research is misdirected, hopeless, and maybe even displays humanity’s arrogance as well”. Surprisingly, I hear this question - in the dismissive tone - from AI researchers. Sometimes it comes up as a way to belittle DeepMind’s research in reinforcement learning aimed at more general artificial intelligence.

Machines: A few small steps

I’ve settled into Edinburgh, and just completed my first week of lectures for the AI MSc. Here’s my current view over the field of AI, explained via the route I’ll be taking this year.

People: Impro

Impro is a thought-provoking book. Here are some of my favourite sentences, taken out of context:

People: Leaving a company

I left a company for the first time last week. I’ve had plenty of jobs in the last 9 years, but I expect this is the first one that will stay on my CV indefinitely. There are certain things from the last two years, and from the last two weeks, that are worth reflecting on. Let’s start at the end.

Ideas: Those who can't learn, teach

One of my colleagues jokingly mocks me for teaching thing I haven’t learnt yet. He says that usually people learn something, use it for a number of years, get good at it and then finally teach it to others. That sounds great. I look forward to doing that! Right now, I’m not in a position to do that because for most things I do I’m starting out. And I prefer to start by teaching.

Ideas: Learn at source

People sometimes ask me how they can learn the skills they need to work as a data scientist. My preference is to learn at source: learn each skill from the domain where it originated and where there are most clearly experts.