The important thing for Europe is to devise AI which is beneficial to people. Which essentially mean two things. Beneficial to individual people and beneficial to our societies as our communities. And this are two very different things. We might be very intelligent smart individual and be very stupid as a crowd. To somehow engineer social technical society which is able to leverage the right network effect that bring us to decide the properties of well-being. Plurally has been democracy. A better place to live. Well to do this we really need to rely on AI, sure. But we need to rely into developing AI which is working with us towards the same objective. We are not only smart as individuals, but we are smartly adopting as an intelligent crowd, reaching common goals all together, making the traffic flow in the city or make democracy work.
What would be your blue sky project in AI for Europe?
If I had a lot of resources, especially beautiful minds for a blue-sky project, I would like to engineer the platform for democracy. The platform that would enable people to be competent and connected to many different views of any controversial issue. To become essentially more able to understand our peers, and be more able to, to become a pluralistic diversity of opinions, able to make the best choice in every possible moment towards all the big challenges that are in front of us, from the climate challenge to the democracy challenge, to the economic and jobs challenge, the inequality challenge. All those require our ability more competent, more connected and more open people. And AI can, Humane AI, can really much help in this direction.
The human AI project is especially interesting for me because of its social aspects as I already said. Most people deal with Human–machine interaction when they are thinking in terms of Artificial Intelligence, that humans are embedded into society. That is a very, very important aspect. We act as social beings. Well–being of the human depends not only on the individual properties but also on how we feel ourselves in the society. And that is very much influenced by AI and it will be more and more influenced as time goes on. This are aspects which we want to explore within this project.
We are looking at problems where AI is able to impose some kind of biases onto the society. I am thinking in terms of recommendation systems, I am thinking in terms of social online networks, which a large impact on the society, it can enhance segregation, fragmentation of the society, which are bad consequences of these algorithms, which are applied. And we are also looking for possible way outs of these traps of the new technology.
What would be your blue sky project in AI for Europe?
AI has a self-development, it is from the point of view of companies of technology, very rapidly developing, technology with fantastic future, but I think if I had the possibility I would put much more emphasize on social aspect, on assuring that the development of AI will not bring us to society which is not compatible with European values. We want to live in a ecosystem where AI and humans have their place, but with the values which we have now about social life.
I think we’ve been doing Human – Centered AI for many years, we’ve always been interested in how we can use technology to empower people, but we’ve been using other technologies like virtual reality and augmented reality. AI offers new challenges to think about how we might think about what humans can do, what machines can do and what they can do together as a synergy. For a long time, we used to think about what humans were good at best and what machines were good at best. I think the challenge now is to think about what humans and machines can do together that they are best at.
What would be your blue sky project in AI for Europe?
Oh, in Europe, for Europe. I am very interested in how we can think not just at the individual, but at the group and society level. How we can make an impact and make big changes to improve society. If we could develop complex decision-making resources, so that we could perhaps once and for all resolve a political debates that keep going on and well we could use, we could have this AI tools that could help us to look through loads and loads of past records and make a more informed decision about someone or something before we decide. I think that is something where we and also data that is accumulating. Something that we could really sort of push towards. That would require lots and lots of machine learning capabilities and Human–computer interaction, people as well as philosophers and ethicists to work together. That is my blue sky research project. Is to think about how we can really make a difference to society by providing them with the way to do more complex decision making in a variety of domains and context.
Well, the thing is like when I look at collaborations with machines and humans there is a huge aspect on how human think and how humans represent actually the way they understand and make sense out of what they’re doing. If we want to have machines that actually work together with humans in this collaboration, then we actually have to respect that the process, that meaning is made and reflection is done in the way we are building this AI structures and the way we’re using machine learning for this purpose. In my work I actually use AI technology from very different fields and apply that, because it made sense in the process and not necessarily as a common way to do it with machine learning.
So, the human AI project, what I understand is very basically human centered AI. If we want to build collaborators that actually make meaningful improvements of the world, meaningful innovations, we have to look in how this can be meaningful for humans. That is kind of where I see, we should focus on what’s meaningful to human and not necessarily for the system. That has certain implications about explainability, how thus systems can explain themselves, how can systems explain its reasoning, because reasoning is important for example if you gather ideas into brainstorming in this case but also about the whole aspect of sense making, semantics that are related to that, especially when we talk about abstract ideas or concepts that we are actually developing. There is a lot of aspects on that. And this whole aspect of learning, what does it actually mean to learn in this context where we actually don’t know where we are going. We do some idea of brainstorming, we don’t know where we go, we learn over the course of actually talking to each other, how is that done in the system and all these aspects are very related on how humans actually understand this process and do it naturally. That doesn’t mean that AI’s has to do the same, but they have to acknowledge and be build towards that.
What would be your blue sky project in AI for Europe?
That’s a good question. I think there is certainly, I see, a need, especially when you talk about collaborative AI, I certainly see a need of integrating for example more psychologists, social scientists in the work, basically, working on how do we actually do collaboration with machines but also within humans. Projects could relate to what is actually grounding, what is actually sense making in human and in human AI interactions, and things like that. Also, kind of like, investing a lot of money on developing new approaches based on for example human behavior. I for example did a lot of design studies, before I start designing system to basically see how humans do it. If we are thinking look at several of these processes, we might actually find patterns that allow us to build better AI and machine learning approaches in this context.
Do you have a one-liner on that?
No. To be honest, no. I think there is so many work to be done on different aspects. On the machine learning side, we have to stop thinking about to have a goal and that’s what I want to ultimate, because the values is in basically developing ourselves to get a machines. Which means developing the human and the machine at the same time, so that we need new approaches there, but we also have to better understand how do we actually think, how do we actually get this information, how do we actually do innovation. Mapping these in interaction actually, is a very more complex that I have a one-liner for that.
For me the big motivator in Human-centric AI, is that we are seeing more and more that the human intelligence is very different from machine intelligence. Humans and machines are very different, they reason in very different ways. Machines can spot patterns that human eye cannot see, machines can do very long chains of reasoning that human mind cannot do. Humans are socially embedded, they have ethical preferences. There will be very different styles of reasoning, between human intelligence and machine intelligence. We shouldn’t be trying to replace one with the other and too much of current AI is aimed at the replacement model. At replacing human expert with the machine. Instead let us collaborate between humans and machines and we are each good at very different things and then the mixed team of humans and machines will be stronger then either the machines or the human separately. And this collaborative aspect of working together between humans and machines, that is for me, you know, the major driver of the Humane AI project.
What would be your blue sky project in AI for Europe?
My dream for the Human AI project is to build myself a colleague. I want AI to be my scientific colleague. I want to walk into my office in the morning and hear from my assistant, “oh I’ve read another five thousand papers overnight, and the conversation we were having yesterday is all wrong, we need to do it again, after I’ve read these five thousand papers.” Right, and this is not just the machine doing data mining, or pattern recognition, or information retrieval in papers. But really, a machine that collaborates with me in every aspect of science. It should read the literature, it should to formulate hypothesis, it should help me to design experiments, it should interpret the results and it should help me to write the paper.
Human – Centered Artificial Intelligence is mainly about bringing the human into the focus of the system, creating systems that assist humans, that are transparent to humans, so that they can also explain themselves or maybe debate actions that they’ve taken. For robotics this would also mean that you have systems that you can understand, question. I think it is a very powerful technology and its really important to not only create black boxes, that way you can see that the function is very good, but then you also understand what is behind this black box, what is in it. Compared to that we are in the same situation like for example neuroscientists that is looking at the brain like a working system, and you have to find methods to really go into the system and see where is the decision for example coming from.
What would be your blue sky project in AI for Europe?
I think with unlimited resources I think the research that has to be done is not only the research on the technology on AI, but it has to be combined with the society, it has to be combined also with understanding, for example human decisions and we have to combine this with the technology. We have to see where is the influence in the society, how the humans interact with each other and the public view people think that this is kind of solved, but it is not. It is important that we understand it and that we really integrate AI into a society.
Well Human – Centered AI, AI in my opinion is a big tool we have in our hands. And for this reason we have to understand this and really make it our own tools. This is particularly true for Europe. We know that in other countries AI serves the government or serves over the top companies. In Europe we really believe that AI can serve the needs and the desires of the society. Human AI is big opportunity to bring researches from many different fields together in order to work on all the different aspects of AI, and achieve the mission, the challenge that I mentioned before.
What would be your blue sky project in AI for Europe?
Okay. Well, this is a difficult question, since of course if I had infinite resources I’m not sure what I would do. What I think the point here is to use the resources to find research in a bottom up way. I don’t think we really have a ultimate goal that we know where we can spend money. We have to give money to research in order to make it robust and solve the problems we will face in the next years.
Well, for us, it is essentially understanding the trend of the decentralization of AI, so we are moving more and more towards models where instead of having centralized monolithic AI systems, AI is coming closer to the users and we are moving to systems where small pieces of AI analyze data that are generated directly on user devices and therefore they should be more and more human centric because whatever they do has more and more to do with the specifics of the users that generate those data and the knowledge we extract is more and more human centric. That is why we are so interested in the Humane AI project as a way to understand the interplay between those local models and the way they interact with each other, considering they are more and more proxy of the human users.
What would be your blue sky project in AI for Europe?
I think we are very much on the onset of this new paradigm, where AI comes to the edge of the internet, to the edge of the systems and therefore I think we should better understand what is the interplay between huge amount of small pieces of AI, that work together to build a collective AI system. This changes dramatically the paradigm, because we’re not anymore in a centralized system, therefore we have to take into the account emergent behavior. The way the human-centric small pieces of AI interact with each other according to social norms, social behavior, social trust between users, so the human dimension comes into play more and more in this kind of system. I think my blue sky research project would be to understand how do we move from monolithic to a totally decentralized human-centric and socially aware AI system.
I think the most interesting part according to me; the most interesting part is how to combine symbolic AI let’s say with differentiable AI. This is essentially useful because there are some aspects which I’m trying to model within my research, let’s say for example trying to provide recommended system which is a better impact, let’s say trying to provide novel stuff to people, you know, or diverse stuff. Let’s try for example to say, on Spotify, I want to try new things, or Netflix, I want to get a suggestion about a movie that I wasn’t even aware of, that its totally surprising to me and that I more likely I will like. This essentially means in my opinion trying to inject kind of constraints, kind of new ways of providing this recommendation and trying to avoid to be repetitive or trivial in a sense. It seems to me that this has to mostly to do with the intelligence with a way of trying to encode the ground knowledge about what you can do and what you would like to express and also trying to inject is a machine learning system with machine learning algorithms. This is basically my personal view of how humans or AI can contribute into development of a field. Basically, the essential idea that you can in a sense encode the feelings the emotions, whatever you can model about human, human being and try to directly model it within the machine learning algorithms, within the Artificial Intelligence, with what Artificial Intelligence can deliver back to humans. In a sense it is a two way interaction, you would like to get the most that you can and on the same time you can deliver results that can be very, very well exploitable for people.
What would be your blue sky project in AI for Europe?
Well it is very well related once again, I would like to have the possibility (03:09, 03:10) that are able to answer or enhance user experience. I am fortunate enough to live in a world where we are all connected people, we use mobile phones, we could do things that we could not do ten years ago, let’s say for example, I come to Paris and I have a mobile phone and I can find the spot where the workshop is, or I can move the station to the airport, or whatever, to the hotel. This is essentially enhancing user experience. I would be very nice if we could proceed farther. And let’s say deliver people with stuff people don’t even know they need it. This is my blue sky project.
Human – Centered AI is an approach to designing and developing AI that improves humans, that aims to not replace the human, but enhance the human. AI has the power to significantly change the world, to some extend it already has, but I think the biggest changes are still ahead of us and I am highly interested in being a part of that change.
What would be your blue sky project in AI for Europe?
My blue sky project for AI is to make better humans, to improve the quality of life, to enable humans to do things that they perhaps couldn’t do before, to make people better that the best people out there currently.
Yes, I think the basic idea behind a Human – Centered AI is the idea that we not only need to make algorithms to fulfil a certain purpose, but we need to make individuals and society better. We have to introduce values of ethics and of social welfare into the algorithms that are trying to solve issues. I think it is a good project, it is something we need. Now, most od the development in the AI is being done in private companies, by corporate interests and researches are just, you know, understanding how it behaves. We are not taking like an active role in also developing these algorithms. So, I think we need to create some sort of ecosystem between institutions, governments, researches and companies to really make AI better and more ethical.
I think we live in a strange moment in society, that we have very big problems that we have to solve. We have a climate emergency, we have extreme inequality, we have a lot of technology that we can use in order to solve these problems, but somehow, we are very disconnected from this, as a society, and from this technology. I think we can use AI to make use better informed, to actually help us in looking for solutions to these big problems. But we need that this AI is well engineered. That it is not only made by companies, but is actually made, you know, by the researches, by the governments, by corporations all put together. I think this is one of the only ways in which we’re actually going to solve giant problems, like climate change.
What would be your blue sky project in AI for Europe?
My blue sky project is to solve with well informant ethical AI.
Well. Okay. I guess everyone would have a different idea of what it is, right, and I’m a philosopher and most people here are computer scientists, so probably our ideas are going to be very different. My idea of Human AI is something like being able to still introduce AI in our life in such a way that we remain somehow human and not because AI is going to replace us in some way, but rather that introduce AI to help us as humans to become better humans, to help others to find more free time for being without kids or for finding better ways to have work as oppose to replacing us in exactly those same domains.
What would be your blue sky project in AI for Europe?
I think that we need to, okay, again, from the point of view of humanity as I would put out, if I had unlimited resources, I would bring a lot of philosophers, sociologists and historians and all these people that try to think from a different angles and try to understand how, you know, AI works and why and where are we heading towards and what can we say about past technologies and how it affect our lives and what not, so trying to figure out in that way. And I would of course focus on my own research, which is explanation, explanatory AI, understanding and everything that had to do with the ethics of it.