Robots are making our beer and managing our money

This is how our iBe Disruptors breakfast at the Royal Exchange in London started, looking at how machine learning is affecting a diverse range of industries, from brewing to financial services. Hew Leith from 10x, on behalf of IntelligentX, kicked us off by showing us how machine learning can improve beer creation.

The process does need some human intervention; consumers drink the beer then give feedback on taste and other characteristics such as fizziness. This aggregate feedback is processed by a variety of algorithms and moderated by a human supervisor to make the final recipe. The system does add a final touch, using ‘wild cards’ to prevent creating a middle of the road beer which satisfies everyone and delights no-one. The robots recently added grapefruit to one of the brews, for example.

Something else that I found fascinating during the session was the concept of data-efficient learning, which one of our fintech partners, Alessandro Vitale of Conversate talked about. A lot of the attention on machine learning has been on deep learning, parsing thousands and thousands of scenarios and gradually fine-tuning behaviour. However, when you have limited data you have to take another approach – IntelligentX, for example, simply can’t produce millions of bottles of beer in the world to find the perfect combination, so the robots have to learn efficiently. Similarly, Conversate banking chatbots have to learn quickly from short, sharp interactions with consumers looking for very specific information, such as their bank balance.

This is surprisingly difficult – consumers tend to ask for something as simple as their bank balance in a large number of ways. However, machines offering financial advice will be something which comes relatively naturally. Our very own Head of Advanced Analytics at iBe TSE, Vittorio Carlei, was our last speaker, and examined how AI – if programmed correctly – can easily assimilate large amounts of information and avoid human bias such as emotional connection with brands or gut feeling, which can be arguably harmful when it comes to making risky investment decisions.

Of course, this is a very specific application: we can easily see this kind of AI making strong portfolio recommendations when it comes to straightforward tracker funds. Once customers set risk boundaries and any ethical considerations, it is straightforward for machine learning to select funds fitting this profile and make recommendations. This has the clear potential to be superior to many of the advisors who are affected by emotion, gut feel and off-days!

However, as one of our attendees noted, it does also run the risk of creating the equivalent of the generic beer that the IntelligentX team are so wary of. Tracker funds can be profitable, but also represent little genuine selection skill in terms of wealth management – and perhaps more importantly, relatively smaller gains. There is a cornucopia of completely unquantifiable factors which affect the success or failure of a company, such as culture, leadership, past experience and many, many global macroeconomic factors which also contribute. AI is yet to learn and use these nuances and ingredients to success.

This thought put my mind at rest to some degree. Like most people, I’m excited and terrified by the prospect of AI in equal measure. But the genuine opportunities in the financial sector, the big wins, will still be found by humans who can somehow quantify the unquantifiable, and perhaps more importantly, programme the AIs in question! 

Now if you’ll excuse me, I think my robot butler has arrived with a beer!

 

Arrange a Conversation 

Browse

Article by channel:

Read more articles tagged: AI, Featured

Enabling Technologies