The algorithm of innovation
While the European Commission formulates a strategy for the use of artificial intelligence and is studying the complex issues related to this field, professor Tuomas Sandholm emphasizes the importance of investing in artificial intelligence. He believes there are considerable opportunities for growth, and countries that invest in the technology now will profit from the new job opportunities later.
Tuomas Sandholm achieved global fame earlier this year when Libratus became the first artificial intelligence program to beat four of the world’s best poker players during a no-limit Texas Hold’em contest. It is one o’clock in the afternoon when the Helsinki-born Sandholm (1968) walks into the glass-walled office and sits down at the long table. Flown in from Pittsburgh, his home base (“a great place!”), he delivered a 90-minute talk as keynote speaker during the artificial intelligence conference at the Brightlands Smart Services Campus. Topic: Libratus, or more precisely, the algorithms behind the artificial intelligence program for strategic reasoning. After all, Sandholm may be capable of many things, but he does not consider himself to be a good poker player. He says this is a question he gets on a regular basis. “If I answer in the negative, they want to know how I can develop poker strategies.” It is a misconception that Sandholm, a professor affiliated with Carnegie Mellon University, also located in Pittsburgh, comes up with poker or game strategies. He writes algorithms that “automatically create these strategies”. Sandholm worked on Libratus in 2016 together with one of his students, Noam Brown. In actual fact, this is the result of a project lasting over 14 years, focused on “solving imperfect information games”.
A better world
The Libratus breakthrough goes considerably farther than the second Brains vs. Artificial Intelligence battle that made global news. There is potentially a wide variety of new application possibilities for the algorithms, ranging from medical to financial. Sandholm is however dead-set on using the algorithms and artificial intelligence solely to create a better world, something he is regularly heard proclaiming in the media. In this sense, it is often not a question of conflicting interests at all he says today, immediately admitting that scientific knowledge and technological innovations are appealing to different parties and organizations, including criminal. By researching application possibilities or working with others to do this (“I’m not waiting for someone to take it to the world, I’m taking it to the world myself”), Sandholm wants to ensure that his knowledge makes a positive contribution to the world. Within this context, he also founded Strategic Machine, Inc. earlier this year. The company studies a whole laundry list of potential applications in various areas. He sums a few of these up: automated negotiation, strategic portfolio optimization, cybersecurity and strategic pricing. These applications consistently assume the presence of imperfect information, such as a changing market and the lack of information on the competition. The “most radical project” involves possibly exerting influence over T-cell populations in patients. Although tests with human test subjects are not yet on the table, experiments with live mice will soon be conducted in the relatively short term.
These days, artificial intelligence and robotics are high on the European agenda. In February of this year, European MPs made the case to the European Commission for implementing EU-wide regulations for these areas. In their view, these rules are necessary in order to optimize the economic potential of artificial intelligence and robotics and to guarantee a standard level of safety and security, as well as to lay down the liability involved in these fields. The European Union must take the lead in this, according to the European MPs. As the use of robotics grows, so does the number of ethical issues. When asked, a spokeswoman for the European Commission said that the Commission is currently working on a European strategy relating to these “sensitive topics”. In all likelihood, this strategy will be presented early next year. “The Commission has long been aware of the importance of robotics and artificial intelligence, as well as the potential advantages and risks these bring with them.” Among other things, the strategy encompasses encouraging the use of artificial intelligence to benefit the European population and economy. In 2018, the Commission will also delve deeper into the ethical, legal and socio-economic issues that are inherent to the technology.
Invest for employment
Sandholm believes that investing in artificial intelligence is a “fantastic strategy” for countries, including smaller countries such as the Netherlands. The United States is the leader in the field and not easy to dethrone, in his view; China would be the only country capable of taking on this challenge. For this reason, he recommends that smaller countries concentrate primarily on “narrow focus areas”. This might include applying artificial intelligence in the fight against global environmental problems, a link that Techruption has also made. The competition in this area is not as substantial (for the time being), even though the need for solutions is obvious. There are so many growth opportunities for artificial intelligence that enough or even a great deal of success is conceivable, even for those that are not leaders in the field, Sandholm concludes. In short, investing in this technology is always a smart move. This is also the case from an economic standpoint. He says that although the frequently discussed fear of that jobs will be lost because of technological innovation is understandable, it is important not to be blinded or paralyzed by it. “Technological progress inevitably leads to a change in the employment situation, but this doesn’t mean that it only leads to a reduction in the number of jobs. New applications lead to new products, services and earnings models. Countries that invest in artificial intelligence now can profit from this in the future. Anyone lagging behind will draw the shortest straw later. In these countries, both people and business will ultimately suffer.”
The challenges of regulation
When it comes to the regulation of artificial intelligence however, Sandholm feels it is still “way too early”. The scientist points out that the subject matter is extremely complex and thus difficult, if not impossible, for most people to grasp. Formulating good legislation and regulations is therefore an enormous challenge. According to Sandholm, a lack of thorough knowledge on the part of politicians (“if regulators start regulating things they do not understand”) will have incredibly negative consequences; this will drastically limit the (application) possibilities of artificial intelligence and slow down important innovations for people and the world. “Artificial intelligence is simply capable of a lot more than most people think.” In other words, trust in the technology and significantly more knowledge about it on the part of politicians are necessary in order to arrive at well-considered regulations. Sandholm says that after Libratus’ win, he started receiving emails from strangers who weren’t convinced of the triumph. “They wanted to take on Libratus themselves. The crazy thing is that they would never dare claim the opposite. No one had any doubts about the poker players’ skills. People have a remarkable amount of faith in people, but not in artificial intelligence. This has to change. It’s simply inaccurate.”
About Libratus’ victory
For the laymen: this wasn’t the first time that a computer beat a human in a game. However, up until 2017, this had never been done in a so-called “imperfect information game”. A computer playing chess, for example, has all the information (which pieces have been eliminated, the positions of the different pieces, etc.) it needs; this isn’t the case with No-Limit Texas Hold’em. Additionally, in this variation of poker, players may place unlimited bets, which made the challenge for Libratus so great that many people didn’t think it was possible for the computer to win.
About Tuomas Sandholm
Tuomas Sandholm comes from a family of academics. His father was also a professor (pharmacology and toxicology), and his mother got her PhD in periodontics and also ran a dental practice. Sandholm made worldwide news with Libratus this year. However, Tuomas Sandholm was a famous name even before all this; algorithms he has created are used in the US to match kidney patients with live kidney donors. The scientist has an impressive résumé that is an amazing 132 pages long. In the future, he hopes to be able to use his knowledge to take on environmental pollution. This will have to be a high-impact project though; Sandholm doesn’t see the point in investing so much time in it otherwise. Sandholm lives with his wife and two daughters in Pittsburgh. As a professor, he is affiliated with Carnegie Mellon University (computer science department) and speaks four languages, including Finnish, the language of his country of birth. By his own account, he’s not a good poker player, but Sandholm did make it to the top in windsurfing, winning first place in Finland in the 1980s. “What would I do if I had to choose between windsurfing and poker? That’s like asking which daughter I love more, haha. I refuse to make a choice.”