Last night I attended the Royal Irish Academy lecture “Show me your data, and I’ll tell you who you are” at Ulster University’s Magee Campus. It was an interesting lecture by Dr Brian MacNamee, one that sidestepped any technicalities and aimed for a general audience. It was a very good, informative and entertaining lecture.

One thing that I did notice was the mix of audience, students from school, members of the public and some lecturers from the university, no entrepreneurs in the room that I could see which was a shame though.

It was the amount of school ties in the room that inspired and prompted me to ask the question during the Q&A, “What do you see as the challenges to Machine Learning over the next 5 to 10 years?“.

Some Relics

Some of the machine learning algorithm that we’re using are old, take the decision tree for example, the ID3 algorithm designed by Ross Quinlan goes back to 1986, it’s on its thirtieth birthday. Threshold Logic, which because the foundation for Neural Networks dates back from the work of Warren McCulloch and Walter Pitts back in 1943, that’s 73 years ago.

Most of our modern machine learning systems are based on old technologies and algorithms, is there an opportunity to refine and redevelop these technologies? Is there an opening for new algorithms? I believe there is and the ones who will carry that torch are possibly the ones sat in the Magee lecture room last night proudly wearing their uniforms.

Thanks to Digital Circle for listing the event, I wouldn’t have known about it otherwise.