Highlights from CogX 2019: The Festival of AI and Emerging Tech

Opinions
Justin Shenk
Field Data Scientist
Björn Brinne
Head of Data Science
Åsa Bertze
Senior Data Scientist
Carl Thomé
AI Research Engineer

Three exciting, yet busy, days for our team at CogX have come to an end. CogX is UK’s answer to the questions, "How will AI affect the future?" and "How can I get involved?" Over 15,000 leaders across academia, industry and government attended this year’s AI Festival to discuss the latest challenges and solutions in data science, machine learning (ML) and analytics.

Our team was there to discuss the potential of operationalizing AI across various sectors. We had a fantastic number of great conversations at our booth during these past few days and our AI handbook, The Essential AI Handbook for Leaders, flew off the shelves (1,000 copies in 2,5 days!). A big thanks to the CognitionX team of honoring us with the award for Best AI Product in Next Generation Infrastructure.  

And here are our AI team's highlights from this year’s lineup of speakers:

1. The Future of Math Education

Presenting at CogX were several prominent speakers from the Alan Turing Institute in London along with many other notable speakers.

Conrad Wolfram and his brother Stephen are associated with the popular mathematics applications Wolfram Alpha and Mathematica. At the festival, Conrad proposed teaching a more practical curriculum of machine learning in primary school, rather than requiring students to solve lengthy equations, which most students will never use.

"There are a few cases where it is important to do calculations by hand, but these are small fractions of cases. The rest of the time you should assume that students should use a computer just like everyone does in the real world." – Conrad Wolfram

He also argued for organizing the curriculum by conceptual complexity, rather than computational complexity. Since AI and machine learning shift much of the computational burden to machines, he suggests that we should help develop the skills needed to build on top of those already powerful tools. Competing with machines won't get us anywhere, he argues.

Erik Brynjolfsson presenting on the Innovation stage

2. When Machine Learning Met Genetic Engineering

Stephen Hsu of Michigan State University gave an interesting keynote for this session. Hsu is a leader in the use of AI in computational genomics. His group has been able to make accurate predictions of complex traits by using AI to analyze huge numbers of individual genome polymorphisms. It includes human height to an accuracy of about 1 inch, various disease risks and, interestingly, cognitive abilities as measured by IQ tests.

The session continued with a panel discussion featuring Martin Varsavsky, Executive Chairman of Prelude Fertility and Helen O’Neill, Lecturer in Reproductive and Molecular Genetics at UCL. The discussion centered around the implications of similar results on selective IVF treatment, and there was the consensus that in a few years’ time it will be possible for future parents to choose embryos for IVF based on desired traits. Stephen Hsu also presented results suggesting that the willingness to make these kinds of selections varies significantly between different parts of the world, hence it would be very hard to develop regulation that can be globally agreed upon, which is why the technology will not soon be widespread and available.

3. What Can Machine Learning Do? Implications for the Work and the Economy 

Erik Brynjolfsson, Director of the MIT Initiative on the Digital Economy, presented a framework used to assess the suitability of machine learning for different work tasks that are currently performed by humans. His team used the framework to assess 18,000 different tasks for 950 occupations on O*Net, an occupation taxonomy. For example, the 29 different tasks performed today by a radiologist. The assessment of each task is based on a 21-question rubric, which includes questions such as “Task information is recorded or recordable by computer?”; “It is okay to make mistakes when completing this task?”; “Task is repeated frequently?”; and “The task is about choosing between multiple predetermined options?”

The results found that many high-wage jobs are newly affected by machine learning, though low-wage jobs remain more exposed. Also, no occupation can be fully automated by ML, though nearly all have some exposure, he explained.

Hope to see you next year at CogX 2020! Until then, are you ready to get your hands dirty with deep learning? Sign up for our Community Edition here.

Justin Shenk
Field Data Scientist

About

Justin Shenk is a Field Data Scientist at Peltarion. He develops custom data-driven applications for customers in the U.S. and EMEA regions. He previously worked in neuroscience research and started a crowd-sourced translation company.

Björn Brinne
Head of Data Science

About

Björn is the Head of Data Science at Peltarion and has over a decade of experience working in data science. Before joining Peltarion, Björn worked at companies such as Truecaller, King and Electronic Arts. He holds a PhD in theoretical physics from Stockholm University and has contributed to many research papers across a range of academic fields, including computer science, string theory and computational biology.

More from the author

Åsa Bertze
Senior Data Scientist

About

Åsa is a Senior Data Scientist at Peltarion. She has over 10 years of experience working with data analysis, statistics, optimization techniques and machine learning algorithms. She has a master's degree in engineering physics from Uppsala University and is passionate about using data to solve real-world problems.

Carl Thomé
AI Research Engineer

About

Carl is an AI Research Engineer at Peltarion.

Get started for free