Data needs theory and more intelligence

What's the great difference between today and yesterday? Yesterday we thought, us all business people, that the world was changing only for a few skyrocketing companies, whereas today we're all convinced that there is no escape, that change - whether climatic, social, business or political - affects us all.

1/ Will AI be the major driver of business change ?

The big picture: While Nobel Prize-winning economist Daron Acemoglu argues that AI won't deliver the profits that mainstream companies expect, the prospect of AI combined with quantum computing has fuelled the incredible valuations of Silicon Valley companies. As soon as he came to power, Donald Trump announced the Stargate Project, which aims to invest 500 billion dollars in US AI. Gen AI topic is all the rage even if business benefits are still to understand except for Digital industries.

To keep in mind: For text mining, either code or documents, and now also for image mining, AI performance seems to be very good, and access to the source allows the user to check the consistency of the answers. But for all applications where AI makes a decision that affects humans, if that decision causes harm to humans, AI is unable to explain how it arrived at that decision. In effect, AI is a black box. That's why we don't have autonomous cars, nor do we have AI that approves a loan application or gets you into a top university. If there's a dispute, AI won't be able to explain. This is a major challenge for AI research today. European regulation is blamed for slowing down AI, but it is not Europe, it is the limitations of AI itself that Europe has pointed out. How can companies delegate to AI the ability to make important business decisions, when in case of disagreements, the AI will not give any explanation?

Paris AI summit: It is up to France to host this summit this year, AI regulation and sovereign use cases would be on the agenda. I am afraid that, apart from research initiatives, other initiatives would not change the world.

It is up to each of us to change the business of our companies.

2/ Data or theory - which one is really worth it?

This is not just a question for science, it is also a question for business. Data is the new hype of business, expected to bring new business truths and, for organisations that move fast enough, competitive advantage. As a business, can we always rely on data? Remember that data is not real, it is numbers. Even when people pretend that data are facts, unfortunately like dama.org, they miss the point. Data are measures taken by people or organisations to understand phenomena that affect them. So data are not the facts themselves, but measures of the measurable effects of phenomena as they understand them.

So along with the data they have about their customers, their product, their employees, companies should often find out what all this is, I mean update their theory about it, to assess whether they are collecting the good data or whether they need other data to understand it better. Otherwise, you may fall into one of the following 2 pitfalls.

• A wannabe cowboy paints many targets on a wall and fires his gun. He hits one target, erases the missed ones, and claims he's a great marksman. In data science, this is like running numerous tests and only reporting the ones with significant results, which leads to false conclusions.
• A wannabe cowboy shoots a bullet at a blank wall and then draws a target around the bullet hole, falsely claiming he's a great marksman. In data science, this is like finding a statistical pattern and then creating a theory to explain it. This approach can lead to false conclusions.

Theory and data are therefore intimately linked, with the former needing data to prove it, and the latter needing theory to make sense. Data workers must not only be data processing experts, but also understand the client, product and employee theories, unless AI one day get this capability.