Stellar Muthoni

Transcript with timestamps before Speaker labels. Audio link: project_debator_PA.mp3 [00:00:11] Pakinam Amer: This is a science talk podcast from Scientific American. I'm Pakinam Amer. It's as old as ancient Greece and a practice driven by passion as it is by facts and evidence. It's also exclusively human. Animals cannot debate, machines cannot debate, right? Well, here is the thing. [00:00:31] John Donvan: This truly is a first for us the first time that an artificial intelligence namely Project Debater will be on our stage arguing with a human being, and may the best debater win, and as we like to say at every debate, may civil discourse win as well. [00:00:46] Pakinam Amer: That debate between man and machine happened in 2019 in San Francisco in front of a live audience. My guest today is Noam Slonim, a distinguished engineer at IBM Research Haifa, who along with colleague Ranit Aharonov and others, created the machine that was on that stage. One that can debate a human without a script and occasionally win. One with a supercomputer for a brain. Project Debater is a cloud-based AI system created by IBM, built with an NLP or a Natural Language Processing model, and trained using deep learning and machine learning techniques. It took about seven years to develop, and since, it has proven itself formidable opponent to champion debaters worldwide. Its model can scan over 400 million newspaper articles and Wikipedia pages and in the time it takes you to finish a cup of coffee. It has a synthetic female voice, soft human-like with a metallic edge. Some news outlets have called it Miss Project Debater referring to it with a she-pronoun. I'm personally conflicted about humanizing it until it's on Twitter engaging in low-level debates and foaming at the mouth over which animal, a cat or a dog, is the better pet. For now, it will do. [00:02:00] Project Debater: But I suspect you've never debated a machine. [00:02:04] Pakinam Amer: Noam, who decides if your AI has won a debate or not? How can you measure this? [00:02:09] Noam Slonim: I think a reasonable measure of asking the audience to vote before the debate starts and then to vote again after the debate ends, and then the winner is declared as the side who was able to pull more votes to his side. But we also asked the audience another question. Which side better enriched your knowledge during the debate, and consistently, in almost all the debates that we had with humans, the system was receiving significantly higher scores than humans. I think it was not very surprising, but still, it was reassuring and interesting to observe. [00:02:51] Pakinam Amer: Debate is very complex. It involves arguments and counterarguments, cross-references and analogy. The ability to engage with confidence in a dialog and to judge the quality of a piece of information and whether or not it will further once cause, and finally, to leave an impression on an audience enough to sway them to your side. Debate is also more than the sum of its parts, and that's what I ask Noam about next. How do you train a computational system to engage meaningfully with a human? [00:03:21] Noam Slonim: First of all, at a high level, the system has two major sources of information. One of them is a massive collection of around 400 million newspaper articles. When the debate starts, using different AI components, the system is trying to find short pieces of text that satisfy three criteria. They should be relevant to the topic, they should be argumentative in nature, they should argue something about the topic, and they should support our side of the debate. Once it finds these short pieces of text, the system is trying to use other capabilities in order to glue them together into a compelling narrative. This is one major source of information for the system. The other major source of information for the system is a collection of more principled arguments that try to capture the commonalities between the many different debates that humans are having. These are arguments written by experts. We had thousands of such more principled argumentative elements. When the debate starts, the system is looking for the most relevant principled arguments in this collection in order to use them in the right timing. Just to give an example. What we mean by a principled argument, so if we're debating whether or not to ban the sale of alcohol or whether or not to ban organ trade, in both cases, the opposition may argue that if you ban something, you are at the risk of the emergence of a black market that by itself has a lot of negative implications. The black market argument is a principled one. It can be used similarly in many different contexts, and this is what the system is trying to do with this source of information. By the way, people may naively assume that this is just a keyword matching thing. That if you've done something, we should anticipate the opposition to use the black market argument, but obviously, this is not always the case. Sometimes we give the example of a debate about whether or not we should ban the use of internet cookies. Probably we're not going to see a black market of people selling internet cookies in the street corners or something like that. It is more subtle than that. In this aspect as well, the system needs to develop a more, I would say, nuanced understanding of the human language in order to perform well. This is the second major source of information. Finally, of course, there is the button, which is the most challenging part. We need somehow to respond to the arguments raised by the opposition. This starts by understanding the words articulated by the human opponent, and for that, we simply use the Watson speech recognition capabilities out of the box, but of course, we need to go beyond the box. We need somehow to understand the gist of the arguments of the opposition, and for that, we use an arsenal of techniques. Most of them rely on the same principle of trying to anticipate in advance what the opposition might argue, and then listen to determine whether indeed the opposition were making these arguments and then respond accordingly. [00:07:00] Pakinam Amer: You may have noticed that Noam referred to Watson. That's not a reference to IBM's founder, Thomas J. Watson, but a supercomputer, a predecessor of Project Debater which, okay, was named after Thomas J. Watson, you win. Watson, the computer and not the industrialist, made his debut on Jeopardy, the IBM challenge in 2011, and it killed it. Scorching the human competition and wrapping up with a score of over $77,000. The AI was also featured in a collaborative series between Bloomberg and Intelligence Squared US. Bade for by IBM, the series was the first time that the AI really flexed its power to analyze the same issues that are typically taken on by human debaters using crowdsourced information. At one point, IBM Watson sifter through over 5,000 submissions from the public all written in natural human language, and analyzed the arguments within them. It filtered out irrelevant data and clustered the rest into four and against categories. It was then able to weave a narrative the plate of the strength of each side of the argument. Arguably, not a bad tool for policymakers. Perhaps Project Debater can weigh in on the merit of its big brother one day without bias, of course. Speaking of bias, the past few years have proven that algorithms are as biased as the people who created them. What is Project Debater if not a bundle of algorithms? It also seems that Project Debater, without sentiments and inherent moral compass or a sense of right or wrong, is as neutral and fair as the information it siphons into a live debate. How do we guarantee it's not pulling from the wrong sources, sources that are biased or malicious? [00:08:48] Noam Slonim: If the data is biased, the system might be sensitive to that. If you're considering a particular controversy, and then you want the algorithms and you end up with 1,000 arguments that support the motion and only five arguments that contest the motion, you immediately understand that there is bias in the data that you're considering in favor of one side versus the other. Now, whether this bias is justified or not, this is a different question, but this is at least a way to quantify and understand that the bias exists. [00:09:27] Pakinam Amer: In other words, while Project Debater cannot remove bias from the sources, it can, somehow, recognize it and so far, it's not pulling from the entire internet but from a verified library of resources that includes scholarly journals and credible news. While Project Debater is indeed clever and a feat of computer engineering, really it is, it's still missing the very obvious human elements that can make or break some debates, the tone and cadence of a person's voice, charisma, passion of belief, and what a person is arguing for or against. Debater is meanwhile faceless. [00:10:06] Noam Slonim: You are right. Even if we go back to the ancient Greeks and Aristotle, he was thinking about rhetorics, and he defined three fundamental pillars for rhetorics. We have the logos and ethos and pathos. You are right that the system that we developed is more focused on logos. That said, I think it has elements also for the other pillars. First of all, I think it's an interesting question. To what extent an artificial intelligence system has an ethos? I think it does because when the audience listens to a machine quoting numbers and specific facts that are related to the debate, the audience understands that this is a machine I believe, consider the data being presented by the machine as reliable. In addition, regarding pathos, this is again another highly important element during a debate, and obviously, humans are much better with that, but still, we invested in this aspect as well and we were trying to make the voice of the system to be more expressive. Not too much, but still more expressive for the purposes of the debate. We were trying, to some extent, to consider all three aspects, but indeed, we focused more on logos, and this may explain some of the results of the debate, but there are teams in academia that are actually considering the other aspects. This is a very active field of research. [00:11:38] Pakinam Amer: The team that brought Project Debater to life didn't just involve computer scientists and technologists. According to Noam, it was at the intersection of many disciplines. At one point, they even had an author, a philosophy student, and a world champion debater on the team. I couldn't help wonder if someday, Project Debater will find its way into our social media as a mediator or as misinformation police. [00:12:04] Noam Slonim: Over the last year, we started to consider more, I would say, interactive forms of dialog systems that benefit from the notions and the assets and the technologies that we developed as part of Project Debater. We demonstrated the system in a debate which is interactive but has a very clear structure. In a free dialog, this is a different scenario that we have started to explore with collaborators in academia. How can we take these notions into a more fully dialog body system? It's very interesting to observe the difference between this scenario and a debate because, in a debate, you really try to defeat the opponent. Here, I believe this is a different situation and it has a lot of implications because if you just keep shooting evidence at the other side intended to prove the other side that they're wrong, this will not be that beneficial. Chances are that you will just cause the other side to be more protective. We are looking at something that will require new capabilities. It's also about being able somehow to listen, somehow to reflect to the side their concerns. So hopefully, we will have interesting results to share in the upcoming months. [00:13:31] Pakinam Amer: I dare say that even Project Debater cannot argue against that. Noam says that this is their next challenge, making the AI system work in a free interactive dialog. You've heard from Noam Slonim from IBM Research Labs in Haifa. That was your science talk, and this is your host, Pakinam Amer. Thank you for listening. [music]