EP98: Dr. Violet Snell, Computer Vision Leader: An engineer's perspective on AI diversity, compromise and risk
💬 “There are some AI reasoning systems and there are some data-driven learning systems and we haven't quite cracked hybridizing them. But there are some interesting developments, in some very technical areas, which combined elements of reasoning with data-driven learning. So I think we'll get there, it is still very early days.” @Violet Snell
This is a conversation with Dr Violet Snell
🎙️ Violet has a doctorate in computer vision and machine learning and a background in video processing and software development. She is currently leading a team of researchers producing deep learning solutions for traffic safety applications.
🎧 In this episode, Violet talks about her work as Head of AI at Jenoptik Traffic Solutions UK, I asked how good the current crop of road safety cameras are, even when faced with partially hidden number plates and we spoke a little about the development of this particular type of computer vision technology.
The conversation moved to facial recognition, can cameras see through window screens to identify the driver and whether satellite-based camera systems will replace the local camera networks that we currently use.
Another specialized area for Violet is diversity in AI teams, I asked her to explain why this is so important. She said it’s actually critical for successful AI projects to scale out into the real world. She identified one particular challenge that COVID-19 has caused to distributed teams of data scientists and I asked her whether companies could benefit from hiring more psychologists and ethicists to work in or with AI teams.
We moved onto Violet’s core subject which is engineering – she has a master’s degree in Engineering and Computing Science from the University of Oxford. We talked about the engineering challenges associated with AI such as when is good enough, good enough, trade offs, and where do you draw the line. I asked Violet to explain how humans and machines learn, the differences and the similarities and the possible future of logical machines. I think you’ll enjoy this brief exchange between Violet and I about whether and why we should be trying to build human level AI.