Last week, on October 15-16, the Tallinn Digital Summit (TDS) was held for the second year in a row in Tallinn, Estonia. This year, the focus of the summit was artificial intelligence (AI). Estonia is one of the three countries in the current “presidency trio” of the Council of the European Union, which made it an ideal location for the summit.
Tallinn Digital Summit 2018 highlights
Attending the summit were the heads of state from 20 different countries – mainly European countries, but also countries such as Singapore and South Korea. Each country was asked to bring a government representative, typically the prime minister (PM) or the minister of digital development, together with five AI experts and administrative personnel. I was honored to serve as one of the five experts in the Swedish delegation led by our Minister of Housing and Digital Development, Peter Eriksson. The Swedish PM was supposed to join, but he was rather busy that week trying to build a new government.
Agenda and highlights
A number of distinguished speakers gave talks during the summit, including Greg Corrado, Principal Scientist, Director of Augmented Intelligence Research at Google and Jack Clark, Strategy & Communications Director at Open AI. The speakers shared a clear consensus on the potential and challenges with AI, and the fact that Europe needs to make significant efforts and investments in AI in order to catch up with the rest of the world, especially the U.S. and China. For more details on this, read my previous article Can Sweden keep up with AI investments around the world?
In addition to the opening keynote and distinguished talks, there were three breakout sessions focusing on:
- AI-driven government
- AI’s impact on the economy, work, and skills
- Safety and security in the age of AI
In each of these sessions, government officials and experts gave short five-minute talks followed by an open discussion among attendees. I participated in the session on AI’s impact on the economy, work and skills and talked about the need to operationalize AI in Europe. Following the breakout sessions, a number of tech talks discussed topics such as the potential need to regulate algorithms and measures to monitor and track the progress of AI.
Multiple reports were also published and presented during the summit from McKinsey and Boston Consulting Group (BCG). For example, McKinsey talked about AI’s impact on the economy and workforce, presenting six key takeaways:
- AI has the potential to contribute to productivity growth and increase innovation. However, there could also be implementation and transition costs associated with AI deployment.
- Countries were grouped into four levels of readiness for AI. China and USA are leaders, Sweden is in the second group with strong comparative strengths, and countries that demonstrate the least readiness include Brazil and Pakistan.
- Some categories of job activity are more susceptible to automation, such as data collection and predictable physical activities.
- Automation spans from both low- to high-wage occupations, and most occupations (~60 %) will have portions (~30 %) of tasks automated.
- More jobs will be gained than lost due to AI and automation. Still, millions worldwide may need to switch occupations as demand for some will grow while others will decline.
- AI and automation will increase the demand for technological, social, emotional, and higher cognitive skills while reducing demand for physical and manual skills and basic cognitive skills.
For more details, see the dedicated website created for this work: AI’s impact on the economy and work.
Operationalize AI in Europe
To paraphrase Churchill, if I am to speak ten minutes, I need a week for preparation, but if I have one hour, I can do it now. A quote I personally very much agree with, and I was supposed to speak for only five minutes. The title for my talk was “Operationalize AI in Europe” and here is a short summary.
I started in a rather unusual way, at least for me, being a very technical person. I have been engineering and conducting research in AI for 20 years. Earlier this year, I heard a talk from Ajay Agrawal, Professor in Economics at the University of Toronto talking in non-technical terms about how AI will change the world. Instead of giving a traditional technical introduction to AI, I started by giving a non-technical introduction based on Professor Agrawal’s work.
He gives a historical overview of technological disruptions over the last 70 years, comparing AI with other disruptions such as the transistor (leading to the computers we have today) and the internet. For an economist, a disruption is easy to describe: “A disruption is a significant reduction of the cost of something.”
The transistor reduced the cost of arithmetic, essentially making calculations, which changed the jobs we had at that time and created new jobs never seen before (e.g. programmers). The internet significantly reduced the cost of distributing information - or rather distributing products and services. Once again, this disruption changed the jobs we had and created new jobs never seen before.
AI will also reduce the cost of something, namely of making predictions. Just as the transistor and the internet, jobs will change and new jobs will be created. According to Agrawal, the scale of the AI disruption will be greater than both the transistor and the internet.
Peltarion example projects
To provide more concrete examples of how AI will change our current job market, I briefly discussed some relevant Peltarion projects. At Peltarion, we have the ambitious goal of operationalizing AI, making the latest techniques available for all, not just the large tech giants. Besides improving tooling for the latest AI techniques and building an operational AI platform, we also have a large data science team. In the team, we conduct research with real industrial projects to properly understand how to make best use of the latest AI techniques.
One project aims to help radiologists build treatment plans for cancer patients. Instead of human experts having to manually examine hundreds of MRI and CAT scan images (like the image above), this part of the task can be automated with AI. In the cancer treatment picture above, a large tumor can be seen on the left side of the brain. With humans and machines working in tandem, the process of building treatment plans can be made significantly cheaper and more efficient.
In addition, people unable to perform advanced tasks such as detecting brain tumors can be empowered by AI to successfully complete these advanced tasks. Perhaps in the future, people who haven’t received extensive radiologist training can together with AI also help out with building treatment plans for cancer patients. This an example of how AI can change our jobs and empower people to perform advanced tasks not possible without the help of AI.
Can Europe catch-up?
As mentioned numerous times during the summit, Europe is unfortunately lagging behind major world players with our efforts to adopt and create value from the latest AI techniques. We need to make substantial efforts and investments to start catching up with the rest of the world. For more details, see the following article in Financial Times: “Europe left playing catch-up in artificial intelligence”.
What if we do it right?
Here is a quote from one of our customers at Peltarion, Oscar Höglund, CEO of Epidemic Sound, an awesome music company in Sweden. If we do it right, making massive use of AI, it has the potential to 10x your business. AI is much more than just a tool or a niche research field, it has the potential to impact our society more than most people can imagine.
As Oscar says, we must be brave (see interview for more details). We need to start to operationalize AI in Europe, and we need to start now.
02/ More on Business & AI
The AI struggle is real: Thoughts from Web Summit 2019
Getting unstuck: How to adopt a creative mindset around AI challenges
How NLP and BERT will change the language game
Taking the pulse on global AI industry: Almedalen 2019
The value-add of deep learning in predictive maintenance
In a time when manufacturing companies are under intense pressure to improve efficiency and productivity, every gain is valuable
AI is not a silver bullet
Towards operational AI
10 applicable areas for AI: Beyond chatbots and customer recommendations
Machine and deep learning: Non-critical deployment
Machine and deep learning: Experimentation stage
You will never understand AI if you only try to understand AI
In every "AI story," there are five main chapters. You need to start from the end.
Reducing the skills gap in deep learning
Five reasons why you are not data-driven
Peltarion rocks Slush – the leading startup event in Europe
Increasing the deep learning footprint: From niche to mass adoption
How to get your business AI-ready
Shifting the 90/10 factor to elevate AI project productivity
Mapping the challenges of deep learning for a strategic approach
The dangers of AI purgatory for any new project
Three takeaways on trust in AI from the Almedalen Week 2018