Machine Learning You Can Get Excited About
For the long term health of the field, we need to focus on machine learning that everyone can get excited about.
A few weeks ago an article garnered some attention making the case that ‘boring’ machine learning is where the industry should be focusing its attention. Boring being defined here as tasks that humans are not innately good at such as analysing vast databases or modelling complex 3D structures. While these topics may not be particularly exciting, the author argues that they offer more value than choosing to work in computer vision and language where “usefulness seems relatively minimal.”
Note: I work on projects that would fall under “relatively minimal usefulness” according to the post in question. Take from that what you will.
The Power of Computer Language and Vision
There are some good points in the article, it would be hard to argue that machines aren’t better suited to combing through millions of data points. However, I think it’s wrong to suggest working in computer vision and language is of “minimal usefulness.” There’s plenty of work being done in both fields which is not only of high impact but that clearly translates into real world use cases.
Computer language is an area that is not just seeing promising work being created but the actual need for advancements is greater than ever. On social media, human moderators just cannot handle the fire hose of content being posted every second. It comes then that automated tools for monitoring and tackling disinformation need to step in to help maintain healthy online spaces.
Meanwhile, progress in computer vision has done wonders for accessibility. A personal favourite is Google’s Project Guideline which allows visually-impaired individuals to run without the need for a human guide. While the argument that “human brains are very good at the stuff they’ve been doing for a long time” is true for some cases, we also need to be mindful of those who haven’t been able to do such ‘stuff’ at all. You or I may find walking or running along a road to be so simple that any decision-making is effectively subconscious. But for someone who cannot see the road? Computer vision is not just beneficial, it’s a lifesaver that opens up a whole new way to interact with the world.
Getting Excited About Machine Learning
Going further, these “minimally useful fields” also have another key benefit: Getting people excited and interested in machine learning.
Areas that look cool from the outside, such as virtual assistants and self-driving cars, offer very clear ways of getting those outside the field excited and interested about the work being done. It’s hard to imagine inspiring a new generation to work in ML on the promise of developing high-performance database analysis systems. But computer language? I still remember having that profound “I’m living in the future” feeling when I saw the Google Duplex demo for the first time. It’s moments like that which plant the seed of curiosity that can blossom into a career.
Progress may be slow and challenging but what’s the alternative? To not make use of the vast storage and computing power we have available to us? Even if such work may appear as no more than blue skies research initially, progress is built off the backs of small incremental steps forward.
‘Boring’ machine learning has its place but so too does work that the wider public can get excited about. Not just to inspire involvement from future generations but also to help those of all abilities engage with life to the fullest.
View plain text.