The New Hype: “We Do It With ML”

The world needs more engineers and fewer push-button operators

There’s a graphic on the Internet that says, “What’s the difference between AI and Machine Learning?” Ans: AI uses PowerPoint, and ML uses Python.

I often get contacted by start-ups and innovators asking if I can join their company — especially if they are looking to raise money. Many have great ideas and methods that truly advance their fields, but quite a few are now appearing with “we do it with machine learning”. This, to me, is equivalent to the many start-ups who have in the past said, “We do it with blockchain”, but had no real foundation on why they actually used blockchain.

Basically, the digital trust industry suffered from charlatan claims around cryptocurrency and blockchain, and, luckily, blockchain is now mainly used for great applications that really need its core strengths (immutability and trustworthiness). The argument that the PKI infrastructure was flawed and replaced by blockchain was always a sound reason for using blockchain. But not all the “enabled by blockchain” hype.

In most cases, ML just seems like a magic wand to solve any problem, and where there is a lack of real understanding of what the problem is and why it can be solved with ML. To defeat the laws of mathematics and physics, though, is a step too far for even a machine.

My reply to requests which have no basis is often, “You are unlikely to defeat something that uses high levels of randomisation”, and where even a machine does not have any advantages. So, I believe “We do it with machine learning/LLMs” is the new hype for start-ups. Overall, I ask for the research paper or patent that backs up their claims and the empirical work that scientifically shows their advancement. If they come back with, “We do it with machine learning and don’t quite know what happens”, I generally avoid it.

So, my advice is to know your theory and understand the problem you are solving. If ML can help you gain an edge, that’s great, but you must know how much of an edge that is, and why. If you want to go down the line of deep learning, then just press the button and leave the stage, as you are handing your experience over to a machine. The world needs more engineers and fewer push-button operators.