Machine Learning or AI?

What's the difference between Machine Learning and AI?

I spent time this weekend discussing with friends in compliance and data science what's the difference of AI and Machine Learning. Are we in agreement? Hmm not really but I thought of writing my first article on LinkedIn to start a discussion and gather some feedback.

Before I mention my opinion, I thought of asking Microsoft, sorry I meant ChatGPT about its opinion and I quote:

"In summary, AI is the broader concept, encompassing the creation of intelligent agents that can perform a range of tasks. Machine learning is a subset of AI, focusing specifically on the development of algorithms that enable systems to learn from data and improve their performance on a given task. Machine learning is a crucial component within the field of artificial intelligence."

So ChatGPT says that:

• Machine Learning = Systems that learn from data
• AI = Systems that learn from data + Rules based

I would not have objected this definition in principal if it was remotely accurate. Lets take as example ChatGPT I am not clear what rules based system have been implemented on top of the ML algorithms apart from some limitations in order not to return results that may be considered too offensive for some of the markets that operate. This is exactly the same with any ML model that has to be taken into production for both supervised and unsupervised models.

Lets look ML and AI from a different perspective and that is of the data.

• Building ML models means that we build on top of a defined dataset and the owner of the algorithm is not necessary the owner of the data. The company owns the output of the algorithm and no data from third parties have been added.
• AI on the contrary means that both data and algorithm is owned by the same entity. As a company I will give you my data as input, you will store them for the benefit of the engine and I will get an output that hopefully will be close to what I expect. At this point the owner of the AI solution owns not only the existing data but the new dataset too.

Lets assume that we own a small retail bank and we want to decide whether we can lend $1m to a customer.

• ML approach - We have built a model based on our historic customer data and we calculate whether our customer will pay the money back on time or not. The process ends at this point.
• AI Approach - We provide our data to a 3rd party that has all the historic transactions from our customer and most of the other customers as well as all the necessary algorithms. We combine this data with other data sources (i.e. social media, driving behaviour, etc.) that may be more or less correlated and we let the algorithm make the decision without any interpretability and with limited supervision.

On an ideal scenario, this would make a decision more accurate, but in real life the AI approach can introduce biases and a good example of that is how you can force Google Maps to divert traffic from your road if you have 100 mobile phone on a trolley and walk up and down the street (https://www.boredpanda.com/fake-traffic-jam-99-smartphones-google-maps-simon-weckert/) .

In my opinion, the difference between AI and ML is who owns the data and the output and it's a great opportunity for the Tech companies to capitalise on all these data that were capturing all these years now that computational costs are coming down.

But was is your opinion is ChatGPT right or AI is a new Marketing approach to capitalise on a more intensive data collection from the big Tech?

* If you wonder what the image has to do with AI - it's what Google's algorithm on my photos' personal album returned when searched for AI.

Opinions expressed on this article are my personal opinions and do not represent approach / strategy / policy of current and/or previous employees & customers.