Will AI take over the world?
Fake news don't affect people's worldview as much as we might think, said David Sumpter in a lecture he held as part of a symposium organised by Uppsala University in collaboration with Tokyo Tech.
At the lecture, David Sumpter told about a study where researchers examined how well people remember false information fabricated for propaganda purposes, so-called fake news. A group of people was presented a number of old fake news, in order to decide if they remembered the news or not. Some of the fake news had occurred during the US presidential campaign while others were made up by the researchers. The people who participated in the study remembered the true false news at the same degree as the false fake news, the researchers found. According to David Sumpter, it is apparent that fake news do not have a strong impact on people's perception of reality.
Raazesh Sainudiin, who organized the workshop where David Sumpter held his lecture, has recently devoted a significant amount of his time to researching propaganda operations conducted on Twitter before the US election. He has, together with his colleagues, studied how Russian bots have affected American political discourse.
- When we ran our models, it turned out that the bots had no impact at all, said Raazesh Sainudiin.
Do you think artificial intelligence will take over the world?
- No I don't think so. The methods are quite the same as when I was a student and I find it hard to believe that they will change in the foreseeable future, says David Sumpter.
- The answer depends a lot on what is meant by “A” and “I” in AI, says Raazesh Sainudiin, and continues:
- Generally speaking, AI methods today and in near future are statistical in nature and simply “learn” from past historical data. When AI gets applied for certain economic objectives, it can propagate existing status quo, which may not necessarily be the best for society. So, in this sense I'd say certain applications of AI can "take over the world" with possibly deleterious effects on society. On the other hand, we as citizens of a democratic society can shape how AI can serve society - so it’s really a matter of collective human free will and choice.
The workshop was organised in collaboration with Combient and was part of the fifth joint symposium with Tokyo Tech and Uppsala University, which took place at Ångström Laboratory 24-25 September.