Mind the gap!

Keep the following statements in mind:

• Machine Learning in Academia: value dictated by the evaluation metric(s).

• Machine Learning in Industry: value dictated by the customer.

I guess we all know there is no misters behind the fact that there is a huge gap between AI research and AI in industry 🤔 Being data scientist is cool - you sometimes get to build innovative stuff and push it to production. Success: clients are happy ✅

But sometimes you don’t get to do that. Not because you don’t want to, or because the customer doesn’t need/want it - but because the first concept is not realistic enough to productise it. What this means is that we have higher expectations of something that it is difficult to turn into a real product - or nearly impossible.

The gap between ML industry and ML academia is more than just the product you are trying to build. Every single stage of an AI development plan is challenging. Also, one has to keep in mind the fact that the market lays a huge role when talking about productising in industry. Studies show that while Europe is dominating the research area, it is lagging behind markets such as the US and China in terms of productization. One of the things that is being affected by this are the ethical and legal implications of adding AI to a product. This is just a tiny part of AI - but that obvioulsy weights singnificantly when talking about creating a product.

Another example is that when doing research, researchers using a tiny portion of a real dataset, while in real -life (production) environment, they datasets consist of billions of datapoints.

For me the hardest part is when I have to convey the message of why we can’t do something or why the research will never make it into production either because of difficulty for scaling or monitoring or just simply bad governance.

What are your thought on this? 🤔

​