See the interview here (in Norwegian).
- How would you summarise the development in artificial intelligence over the past year?
ChatGPT and other large language models have found their way into our offices and classrooms, and many are attempting to leverage them. These algorithms impress with their ability to communicate and answer questions. Much is positive, and we all strive to master the art of asking precise questions through prompt engineering, manipulating ChatGPT to provide desired answers. However, it's important to be aware that we cannot rely entirely on everything ChatGPT says. No one should base investment decisions or medical choices solely on advice from ChatGPT.
The debate on Artificial General Intelligence (AGI) has entered our daily conversations around dinner tables. Key questions include whether AI possesses autonomous self-control, a degree of self-awareness, and the ability to learn new skills independently. Further discussions revolve around whether algorithms can solve complex problems they were not trained for at their creation and whether this poses a potential threat to humanity. The seriousness of AI risk is acknowledged in a way that was not the case just a year ago. This crucial debate has significant implications for the regulation of artificial intelligence, questions of ownership of these large language models (mostly owned by private companies today), and how we can develop AI that benefits the entire world.
However, I would also like to emphasise that the most exciting aspect of artificial intelligence today lies in the many algorithms trained to solve specific problems with great efficiency. For instance, algorithms at Bærum Hospital assist in diagnosing bone fractures. There are also algorithms predicting issues in mobile networks before they occur and supporting researchers in identifying new medicines. These types of AI algorithms are becoming increasingly common and prove to be incredibly useful.
- What are your concerns about artificial intelligence?
I am not concerned that artificial intelligence (AI) will one day start making decisions to control the world and reduce people's freedom. However, I am concerned that AI can be exploited by individuals for their own private benefit at the expense of others. This is already happening, especially when considering cyberattacks and the spread of fake news, and how AI can challenge democracy. Therefore, we must be prepared to defend ourselves. This entails having control over AI on technological, economic, legal, social, and political levels.
- Is there something that particularly engages you when you hear about it?
I am extremely pleased that the government has decided to allocate an additional billion NOK to AI research over the next five years. The recognition that we must be leaders in artificial intelligence to shape our society is of great significance. From today, Norway has the new Ministry of Digitalisation and Public Governance responsible for AI, led by Minister Karianne Tung. These are fantastic news, and I look forward to seeing action and contributing personally through the Integreat research center at UiO, UiT and Norwegian Computing Center.
- Looking ahead, what do you think we can expect in artificial intelligence technology in the coming year?
I don't believe we will see a new large language model in 2024, such as GPT version 5. I assume this will happen in 2025, and it will be the last in this generation, driven by increasing computing power without any fundamentally new concepts. There will no longer be enough funds to build larger computing power, and there will be a shortage of raw materials globally to construct the computing units - GPUs. At that point, new ideas and methods will be necessary to further develop artificial intelligence.
In 2024, we will observe many smaller adaptations of large language models to specific problems using a technique called "fine-tuning." This approach requires specialised data about the specific problem that needs to be solved. There will also be algorithms not trained on language but directly on well-developed specific datasets. For example, models predicting whether a particular treatment will help a patient or not based on health registry data. We can expect to see many significant successes in this area.
In 2024, the new AI regulation (AI-act), recently approved by the EU, will be implemented in Norway. It will regulate the use of artificial intelligence to ensure that citizens can trust these algorithms. At the same time, the regulation must allow for scientific experimentation so that new ideas can be tested.
- Do you have any thoughts on Norway's role in AI development? As we know, we were not invited to the very first global AI conference in the UK this fall. How can we take a more active role in development?
Norway is a well-functioning society based on trust and has achieved a high level of digital services compared to many other European countries. Despite this, we are not the most advanced country in terms of AI culture and investments. Although we have outstanding research in some areas of artificial intelligence, we do not rank among the best in the world. However, this can change now with the increased attention the Norwegian government is giving to AI, including the investment in artificial intelligence through the AI billion (NOK).
I hope that we will use these extraordinary funds to build a national coordination of AI research, with national scientific leadership that can set priorities and attract international attention. Here, we must think strategically as a nation and not just as a single university or researcher.
If we succeed in this, Norway has the potential to become a country that the international community working on artificial intelligence will look towards in the future.
