AI as a reliable companion

"AI is everywhere, everyone deals with it daily. That will only be more in the upcoming years. At the same time, you see we actually have no idea how to use AI. Yes, it's cool and exciting, the possibilities are unprecedented. It's new and everyone mostly wants more of it. But which problem are we actually trying to solve with this new instrument? Where do we want to speed off to in that flashy new car?" 

"Instead of asking for more technology, more algorithms, we should consider that. Determine what the question is first and then set up a system which has an answer to it. Now, people act as if the technology itself is the solution to everything. Femicide, just to name something, is a worldwide issue. We're talking about a worldwide problem, which is deeply embedded into society. An app warning women if their attackers are nearby is useful, but does absolutely nothing to deal with that underlying structure."

What do you want to achieve with your research?

“I work on AI which is reliable and - just like a colleague - actively thinks along. We wouldn't accept it if teammates would just simply answer each question. We expect them to pay attention themselves and warn us if they anticipate a problem. That they say so honestly if they don't know something. AI should work together with us that way too.
It's not the intention that AI replaces people, or takes over tasks. AI should supplement people, be a teammate. The AI we use now is developed by Big Tech. Their goal is to make as much money as possible, for instance by reducing the number of people required to get the work done. Employees are expensive, it's much cheaper to put an algorithm to work. This is why AI is very task focused, the assignment is to generate an answer as fast as possible and that's it. So we use it as a super powerful answer machine now.

In order to be allowed to drive a car, you need a driver's license. You have to learn how the machine works, what the dangers are. But with AI, we think training isn't needed. There's zero information about the ‘backside’, about the workings of the machine we ‘drive around in’. We download something, assume it'll do what it says, that it's safe, that it'll give reliable answers, and we get started.
The risk isn't just with AI, but also with the human users, because they rely on it blindly. Unjustly.”

What would happen if nobody occupied themselves with reliable AI?

“Then we'd be stuck with a very powerful, inscrutable Black Box which makes decisions for us. We need AI which can reflect on itself. People are no answer machines and AI shouldn't be that either. The system must be capable of assessing the consequences of its own actions. And also be capable of saying ‘wait, there is too much uncertainty in this situation, too much unclear information, I cannot make a good decision.’ It has to be built differently for that. AI now compares and combines all kinds of data from millions of cases, and generates an answer from that. But why that answer was chosen, that remains shrouded in darkness. If there's no fitting answer available, the system chooses blindly, at random. The algorithm is made to provide an answer, so that's what it does. The user never knows when that happens, and therefore also not whether or not a sensible answer is given. We need AI which deals with information in a different way, and also reflects on and answers for the result. A system which is aware of its own limitations and lets you know what it does, that's what we can rely on as users.”

Why did you choose this specialisation?

“I'm interested in solving social problems. There's so much inequality. AI could be used to make knowledge available to everyone quickly and easily. Medical and technical knowledge for instance, AI could assist people in places where there's no good medical care. Or in the education system. There are many remote locations in the world where ChatGPT is the only way to learn English. 
We should take AI away from Big Tech and house it with governments. A right to AI for everyone, like how we also documented the right to education. A Ministry for AI would be a beautiful start. It has an influence on everyone's existence, in positive and negative ways. We must make AI our own tool, shape it. We can't leave that to Big Tech.”

Where will we be 60 years from now?

“Hopefully, AI will then be a sensible and useful companion to humanity. I imagine it as teamwork between humans and AI, who work on an honest society together. To make the world a safe place for everyone. Humans and AI are good at different things and can supplement each other to advance the whole. 
AI has enormous computational force and can process very much information, for instance by combining all kinds of different perspectives. It can oversee the consequences of human behaviour in the long term. For instance, it could help people to make good and honest decisions.
Or to make people less polarized and show them each other's standpoints. We need that so much now. Political opponents can often not accept each other's opinions now already. They don't trust each other's sources. AI could be the neutral party in this and give everyone the facts, plus insight into other convictions. Providing calmness in overheated situations by giving the right, neutral information."

I have full faith in science, we have the knowledge to make AI into a reliable instrument that's going to help us. But it all depends on the political and financial choices which are made by governments and corporations NOW.

Text: Yavanne van Tiggelen