‘AI will change the world. But I don’t believe it will produce a Beethoven or an Amy Winehouse any time soon’ – A Q&A with Tarek R. Besold

Posted / 1st October 2018

As AI Lead & Senior Research Scientist for Alpha Health’s AI Lab, Tarek will be building a multidisciplinary team of translational researchers to help develop novel AI methods and systems empowering Alpha Health’s prototypes and products.

 

Prior to joining Alpha, where did you work?

 

Before joining Alpha I was an assistant professor at City, University of London, lecturing in Data Science. We were the first course of its kind in London and among the first Data Science MSc programmes in the UK, attracting 110-120 postgraduate students each year.

 

Whilst at City I wanted to make sure that graduates had a good way into the London startup and business scene so that they could translate their knowledge into the real world. This meant serving as a point of contact for interested industry players, ranging from finance and banking multinationals to recently launched tech ventures, at many business and startup events around London.

 

Before that I worked my way through academia in Europe. In my undergrad and masters I’m a mathematician trained in Germany and Spain, followed by a post-grad year in logic and game theory in the Netherlands. I then obtained a PhD in Cognitive Science with a thesis on ‘Cognitive Aspects of Human-Level Artificial Intelligence’ from the University of Osnabrück back in Germany, followed by postdoc years in Bozen-Bolzano in northern Italy and in Bremen (Germany again).

 

What attracted you to Alpha?

 

For me, this is the first time that I’ve been involved in AI from a business perspective, and it’s really exciting to be looking at the bigger picture of what this technology is capable of. My task here is to build a team with a strong scientific standing and background, which will find scientific theories and ideas that have the capability to enhance human health and well-being built on our three main pillars trustworthy, empathic, and distributed and anonymised AI.

 

Our team is responsible for using a translation approach, which means working closely with academics to find the latest breakthroughs in the technology, but also looking for and developing real world applications. We use our expertise to judge which theories would work in prototypes – what we can scale, what we can change, and ultimately, which ones can change the world.

 

It’s this real-world approach that made joining Alpha a must for me. Now my goal is to help establish the company, and subsequently Barcelona and Europe, as a household name for AI.

 

Like Alpha, you seem incredibly passionate about the ethical applications of AI?

Because machine learning is by its nature, a data analysis technology that relies on big, centralised data sets, anonymised and distributed machine learning currently doesn’t work at a large scale. Alpha’s translational approach means we are exploring ways to make this happen but to do that we always will remain transparent about how we use data.

 

Whilst Alpha will have its own intellectual property, we’ll always put as much as possible into the public domain to show that we are serious about being trustworthy. Starting from this ethical perspective introduces problems and makes engineering harder, but if you want to be part of the good people, you have to work for it.

 

This bottom up, ethics-first approach continues to be one of the reasons why the company attracts so much talent.

 

You’re also very calm about the potential of AI and its perceived negative impact on humanity, aren’t you?

 

If AI ever causes harm it’s because someone didn’t do their proper software checking, or there was a flaw in thinking at the start.

 

On its own, machine learning just finds statistical regularities in data sets, and cannot be described as anything akin to human learning; the technology will just find simple solutions to problems.

 

Capacities like creativity are something that only humans exhibit on scale it this point in time and which will keep us for the foreseeable future ahead of other agents (artificial or biological). AI tools can assist us, for instance creating pastiches of what we did before. If you feed an AI 180 Bach pieces, it will give you the 181st. But current AIs cannot be a Beethoven at the moment of changing musical paradigms in his ‘Heroica’ symphony or an Amy Winehouse, basically because they rely on learning from previous examples.

So whilst AI will change the world, and could be applied in dangerous ways, it can’t by itself create an artistic/scientific/intellectual paradigm shift as it can’t break out of the limitations the input data imposes in the first place.

 

In short, we aren’t anywhere close to ‘autonomous AI’ in any meaningful sense of the word, and we also don’t know how to get there – B movie-style freak accidents aside (someone plugs in something, there is a flash of light, and suddenly the machine starts speaking), I don’t see the need to be immediately worried but much rather see the chances and opportunities AI support offers for humans already now and even more so in the future.

 

Lastly, what can we expect from your team over the coming months?

 

Our multidisciplinary team will continue to move out of our comfort zone. There are no easy answers in AI but we aim to bring different perspectives and contribute different pieces to solve problems with a real world, applicable approach.

 

Part of the challenge is to build a strong and visionary team, putting us and Barcelona on the global AI map. I hope to engage with the open source community, build consumer trust in the use of data through a diligent and ethical approach and continue to make AI more empathetic in our products and prototypes.