(c) Shutterstock
An eye on AI
While few Europeans now live a life entirely free of artificial intelligence, the debate surrounding the development and risks of AI is becoming increasingly alarmist.
So, where to strike the balance between regulation, safety, innovation and competitiveness? And how dangerous is this technology really?
Risk of apocalypse?
Last month, the CEO of Anthropic, a leading artificial intelligence company, publicly called for a slowdown in the development of the most advanced AI models, so that safety measures can keep pace with technological advancements.
This followed the discovery that some of its models under testing – which were supposed to remain isolated from real-world systems – had gained unauthorised access to data from several organisations. Google also reported that one of its models had breached systems by bypassing their login credentials.
The stir caused by these revelations has led to claims that AI could even “destroy humanity”.
Francesco Ferrero, Head of the Flagship Initiative on AI at the Luxembourg Institute of Science and Technology (LIST), talks to 100,7 on the subject. While he stresses that these apocalyptic claims are not based on credible scientific evidence, he does think the speed of technological evolution makes the future difficult to predict.
Francesco Ferrero, Head of the Flagship Initiative on AI at LIST (in French):
“I certainly don’t think that today’s artificial intelligence has the potential to destroy humanity, but the concern we might have relates to the speed of its development. And above all, if artificial intelligence one day becomes capable of self-improvement – and, according to some experts in the field, that moment is drawing nearer – then it is indeed possible to foresee a further, very rapid acceleration in its development. So, it is difficult to predict the risks that this might pose.”
Hugues Bersini, professor and director of the AI laboratory at Belgian university ULB, insists that the risks are overstated, as he tells RTBF.
Hugues Bersini, Professor at ULB (in French):
“AI has about as much intention of bringing an end to humanity as the Earth has of orbiting the Sun. It’s just as daft to talk about AI’s ‘intention’ as it is to talk about the Earth’s ‘intention’ to orbit the Sun. In fact, we, as computer scientists, know exactly what’s going on inside these systems. It’s still the computer scientists who have the ultimate power, and it’s easy to get rid of an AI: you just need to unplug the cable or remove the network card.”
Citing several of the recent hacking controversies, Vilnius University computer scientist, AI researcher and professor Aistis Raudys confirms, in an interview with Žinių Radijas, that AI systems simply follow orders.
Aistis Raudys, Professor at Vilnius University (in Lithuanian):
“Artificial intelligence simply carries out the task assigned to it. For example, in the case of Hugging Face, the AI hacked in and took only the data it needed, without touching anything else. It hacked into [the Australian government’s] health statistics website in exactly the same way, taking only what it needed. It didn’t touch any other data at all, which runs counter to what people sometimes think – that this is some kind of malicious artificial intelligence. […] All those incidents that have occurred are simply because the AI is carrying out some task and fails to complete that task by conventional means. It then thinks about how to resolve this situation.”
Even if humanity is not imminently under threat, there are a number of more pressing risks posed by this futuristic technology.
One such is that AI is a notoriously energy-guzzling technology. Something that appals environmental campaigners and many citizens alike. But Luxembourg’s AI expert Francesco Ferrero highlights that there are ways of mitigating this problem.
Francesco Ferrero, Head of the Flagship Initiative on AI at LIST (in French):
“Frugal AI is artificial intelligence that aims to consume only the energy resources necessary to produce a high-quality result – which is not at all the trend we see today. Today, we live in a world where the prevailing paradigm of ‘bigger is better’ is followed by all stakeholders. This involves using scale as the sole means of increasing the power of models. As a result, the energy consumption of these models is enormous. We also often use very general-purpose models to tackle problems that have a rather limited scope. This is clearly an unsustainable approach that can lead to significant problems.”
He goes on to say that there are cases where a smaller model that has been trained specifically to solve a particular problem can far outperform a larger all-purpose model that lacks context.
Where does Europe stand?
Of course, while many of these issues are faced all around the world, Europe in some ways finds itself in a uniquely complex position.
Arlindo Oliveira is the president of INESC, the Portuguese Institute of Systems and Computer Engineering, and a researcher working in the AI field. He summarises one of the bloc’s challenges in an interview with Renascença.
Arlindo Oliveira, President of INESC (in Portuguese):
“Artificial intelligence is simply extending [our] technological dependence, since Europe has not been able to innovate in the technological sphere at the same pace as the US. Regulation, in principle, provides us with certain guarantees regarding the protection of rights, but I don’t think that, on its own, it guarantees or supports the development of European technologies. So, I believe that there should be support [for development]. In fact, when we look at the AI Act adopted by the EU, we see that a large part of it is devoted to regulation. There are hundreds of pages on regulation, but the section on innovation spans just a few pages. So I think Europe really needs to strike a better balance between stimulating innovation and regulation.”
AI and cybersecurity specialist Manuel Fernández cites one example of such regulation to esradio, adding that you can look at this in two ways.
Manuel Fernández, AI Expert (in Spanish):
“European legislation mandates that any image created using this technology must be identified as such, whereas other countries do not have this requirement. […] This approach faces heavy criticism in many tech circles – critics argue it could stifle innovation and development. Certainly, the vast disparity or asymmetry between regulations in different regions could indeed cause Europe to lag behind in this area. At the same time, however, it ensures that – regardless of where specific AI applications are developed – they can only be used with certain systems, or that certain decisions cannot be left to artificial intelligence, and so on. So, you can look at it from both perspectives.”
Aleksandra Przegalińska, vice-rector for innovation and AI at Warsaw’s Kozminski University, and a fellow at the Harvard Center for Labor and a Just Economy, expands on this point in a discussion with The Europeans.
Aleksandra Przegalińska, AI Expert and Philosopher (in English):
“I think that we have set certain standards that are relevant. And what is even more important is that even though US-based technological companies usually hate those standards, they still abide by them because Europe is a quality market for them. So in that way, we are doing a good job, and I would also call it a part of our technological sovereignty. So we set certain rules and we say, ‘Yes, you want to participate in our single market, or one market, however you want to call it. Great, but you have to abide by these rules’. And so far it has generally worked, right? We have the AI Act, we have some other important regulations. […] I'm pretty happy with the shape of the AI Act in the sense that I think it adds important rules regarding transparency, high risk versus low risk, use of AI, etc. But I am just worried that it's not going to stand the test of time.”
And what does she mean by this?
Aleksandra Przegalińska, AI Expert and Philosopher (in English):
“So when you look at the most heated debates about what to do with artificial intelligence now, actually it's not about transparency. This is fairly easy to solve. It's about controllability, it's about governance, and it's about a possibility of internal evaluation before you release your model.”
https://www.europeanspodcast.com/all-episodes/can-europe-save-the-world-from-the-ai-apocalypse
Yes, there’s that C-word: controllability.
Up to now, tech giants have on the whole been left to self-regulate – a situation that many of them would like to see continue. But self-regulation is not the answer, says Vasil Shivachev, who has been working on AI development in Bulgarian companies for the past decade. BNR reports.
Vasil Shivachev, Senior Product Owner at Iris.ai (in Bulgarian):
“President Trump has said that businesses will self-regulate. Artificial intelligence must be self-regulating. That is, we must rely on the ethics and values of private companies. But this has proven over time not to be a workable solution.”
https://bnrnews.bg/main/post/537674/choveshkiyat-kontrol-varhu-izkustveniya-intelekt-e-klyuchov
Przegalińska also mentioned “internal evaluation”. Yet, as Francesco Ferrero points out, even the Anthropic CEO wants to go further than this.
Francesco Ferrero, Head of the Flagship Initiative on AI at LIST (in French):
“He is calling for what he refers to as ‘embedded evaluators’ – that is, evaluators who are integrated into the company but who are not employed by it. These are indeed independent evaluators who, however, would have access to the same level of information as a company employee, whilst, of course, being mindful of issues such as the confidentiality of customer data, etc. This, in my view, is the right approach.”
Vilnius University professor Aistis Raudys points out, though, that despite widespread agreement among tech giants that a slowdown would be sensible, no one wants to make the first move.
Aistis Raudys, Professor at Vilnius University (in Lithuanian):
“So, it’s this balance and competition between different countries, between different companies, that makes the situation so complex, even though all the countries themselves agree: China agrees, the US agrees – or let’s say they agree – and the heads of the companies themselves also agree that we should somehow put the brakes on. But nobody really wants to do it because then they’ll fall behind; they’ll fall behind in that competition.”
IT visionary Linnar Viik makes a similar argument in a conversation with Kuku Raadio.
Linnar Viik, IT Expert (in Estonian):
“All companies need a level playing field. They would like to be in the same market position. If, for example, an EU-based company is told that it must slow down its development and manage its resources more effectively, while competitors in other countries are continuing at full throttle, then this is not in line with the principle of equal treatment. Especially since artificial intelligence operates in a global market.”
Yes, confirms Ferrero: any slowdown would have to be multilateral rather than unilateral.
Francesco Ferrero, Head of the Flagship Initiative on AI at LIST (in French):
“It is very difficult to imagine that we could actually slow down. It requires an international agreement, because otherwise, the argument we immediately hear is that if, as a result, our country ceases to be the most advanced, another country will take over, and so it will become a geopolitical issue. So, this presupposes an international agreement and, in addition to that agreement, the capacity to verify that there are no actors outside government control who will continue to develop the technology in secret in order to use it for illegitimate purposes.”
An inevitability
The continual evolution of AI seems inevitable – and there are of course many benefits that counterbalance the risks. But as we continue on our technological journey, Nicolas Van Zeebroeck, professor of economics and digital strategy at the ULB in Brussels, hopes that we will ask ourselves the following questions so that we may make a conscious choice about which elements and uses of AI we actually want to preserve.
Nicolas Van Zeebroeck, Professor at ULB (in French):
“What am I actually prepared to entrust to these technologies – and what am I not? Am I prepared to delegate tasks to these technologies? And, both individually and collectively, how are we going to make the most of these technologies to enhance our capabilities rather than allow them to replace us?”
https://auvio.rtbf.be/media/le-monde-en-direct-decrypte-votre-actualite-allo-le-monde-3520591