Slashdot Mirror
New IBM Robot Holds Its Own In a Debate With a Human (nbcnews.com)
PolygamousRanchKid shares a report: The human brain may be the ultimate super computer, but artificial intelligence is catching up so fast, it can now hold a substantive debate with a human, according to audience feedback. IBM's Project Debater made its public debut in San Francisco Monday afternoon, where it squared off against Noa Ovadia, the 2016 Israeli debate champion and in a second debate, Dan Zafrir, a nationally renowned debater in Israel. The AI is the latest grand challenge from IBM, which previously created Deep Blue, technology that beat chess champion Garry Kasparov and Watson, which bested humans on the game show Jeopardy.
In its first public outing, Project Debater turned out to be a formidable opponent, scanning the hundreds of millions of newspaper and journal articles in its memory to quickly synthesize an argument on a topic and position it was assigned on the spot. "Project Debater could be the ultimate fact-based sounding board without the bias that often comes from humans," said Arvind Krishna, director of IBM Research. An audience survey taken before and after each debate found that Project Debater better enriched the audience's knowledge as it argued in favor of subsidies for space exploration and in favor of telemedicine, but that the human debaters did a better job delivering their speeches.
The AI isn't trained on topics -- it's trained on the art of debate. For the most part, Project Debater spoke in natural language, choosing the same words and sentence structures as a native English speaker. It even dropped the odd joke, but with the expected robotic delivery. IBM's engineers know the AI isn't perfect. Just like humans, it makes mistakes and at times, repeats itself. However, the company believes it could have a broad impact in the future as people now have to be more skeptical as they sort out fact and fiction. "Project Debater must adapt to human rationale and propose lines of argument that people can follow," Krishna said in a blog post. "In debate, AI must learn to navigate our messy, unstructured human world as it is -- not by using a pre-defined set of rules, as in a board game."
In its first public outing, Project Debater turned out to be a formidable opponent, scanning the hundreds of millions of newspaper and journal articles in its memory to quickly synthesize an argument on a topic and position it was assigned on the spot. "Project Debater could be the ultimate fact-based sounding board without the bias that often comes from humans," said Arvind Krishna, director of IBM Research. An audience survey taken before and after each debate found that Project Debater better enriched the audience's knowledge as it argued in favor of subsidies for space exploration and in favor of telemedicine, but that the human debaters did a better job delivering their speeches.
The AI isn't trained on topics -- it's trained on the art of debate. For the most part, Project Debater spoke in natural language, choosing the same words and sentence structures as a native English speaker. It even dropped the odd joke, but with the expected robotic delivery. IBM's engineers know the AI isn't perfect. Just like humans, it makes mistakes and at times, repeats itself. However, the company believes it could have a broad impact in the future as people now have to be more skeptical as they sort out fact and fiction. "Project Debater must adapt to human rationale and propose lines of argument that people can follow," Krishna said in a blog post. "In debate, AI must learn to navigate our messy, unstructured human world as it is -- not by using a pre-defined set of rules, as in a board game."
Completely fake. The topics were prearranged, and yes they were "assigned on the spot" but there was a predetermined list. IBM is desperately trying to sell their AI snakeoil. If AI worked, why not have it solve REAL problems that people will pay for, rather than parlor tricks like plying Go, and Chess and other games?
Please define "actually knowing." The machine appears to have sifted through information, extracted bits relevant to the topic, and then presented arguments supporting its position. At some level, it does know its topic. What it lacks is a value judgement of whether it cares about this position or not. That value judgement seems to me to be a critical part of calling it sentient, but it does seem to know the topic. In many ways, the machine knew more about the topic than the human it was debating given the amount of data that it had absorbed and organized internally into information.
So, it scans human-generated content, and then builds a plausible sounding argument to support whatever position you give it.
In this way, it works like a "lawyer", and not like a "scientist":
“there are two ways to get at the truth: the way of the scientist and the way of the lawyer. Scientists gather evidence, look for regularities, form theories explaining their observations, and test them. Attorneys begin with a conclusion they want to convince others of and then seek evidence that supports it, while also attempting to discredit evidence that doesn’t.” Leonard Mlodinow, Subliminal: How Your Unconscious Mind Rules Your Behavior
It sounds like IBM has created a lawyer, not a scientist.
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
But that isn't how debates ought to take place. Debates should start with premises and mutually agreed facts and then reach conclusions via reason and logic.
First of all the world is full of complex systems where you can't directly link cause and effect, predictions of the future, other people's actions and reactions and so on that can't be proven like a science experiment. Even when we agree on the facts, we disagree on the significance and meaning of the facts or even the overall model or ideology that they fit into. A question like "Are Trump's import tariffs good for the American economy?" could probably fill volumes of economic journals without a definitive answer in sight. Even in retrospect 10 years from now they'll still be arguing how much it actually mattered and how much would have happened anyway and certainly a lot of guesswork on the alternatives, so ending in conclusions is wildly optimistic. And that's when they don't have a self-interest in disagreeing with it.
Most public and political debates aren't actual debates, they're more like elevator pitches. You get two minutes in the spotlight to tell people who have no clue about the topic why your idea is great and their idea sucks. They will do the same. What you're looking for is buried deep down in committees, reports, propositions and whatnot where people decide that maybe goods of type X and not Y should be included or the rate should be 25% and not 22%. That's the kind of debate you take when you're preaching to the choir or have an expert group or something. When you're pitching to the general public the goal is simply to convince them that you're the person they should follow.
Live today, because you never know what tomorrow brings
Not knowledge, however this thing can process vast amounts of knowledge and provide a response which has a very high probability of being correct and/or at the very least relevant, and it can do this in real-time.
If users could follow directions regarding computers, or even have a two-way conversation with this machine, then this thing could take over a third of my duties in I.T. (Handholding, and transliterating). As it stands the machine may reduce my time to research answers, a problem which Google has largely solved.
I suppose this could provide a high level of research to those who need to learn how to research, making them ineffective at their jobs without technology, and unable to accurately access the quality of the technology leading to stagnation and the collapse of modern society.
Deploying this thing in a field of debate shows an effective alternative to reducing the value of intelligence. If we were to capitalize on this for educational and testing purposes, somehow outlawing and blocking its use in commercial environments, it could prove an incredible asset for building a brighter future. We could potentially use this machine to sharpen our brains rather than to dull them.
To extend the usefulness of the "training dummy" - how many board meetings, etc. could benefit from a participant with the ability to translate real-time access to the majority of the relevant data into coherent arguments? If you could set the thing in "Devil's Advocate" mode (i.e. argue against anything proposed) you could potentially kill a lot of bad ideas very early in their formation, and steer more plausible ideas past many potential pitfalls. Heck, get two of them arguing against each other in "for and against" mode to potentially cut to the heart of a lot of issues, especially if they can integrate input from human debaters on the fly. Heck, just interjecting "That statement does not appear to have any supporting evidence" would go a long way.
--- Most topics have many sides worth arguing, allow me to take one opposite you.
I don't think there would be a winner in a Trump vs machine debate. Trumps "wins" in debates come from undercutting arguments with completely fabricated bullshit, and if that doesn't work he'll start throwing insults to throw the opponent off their game. That track just won't work with a machine, because they wouldn't have the emotional base to get upset enough to lose their cool.
I too would like to see this debate. The machine *MIGHT* release the blue smoke from the never ending stream of made-up day-dreams, but it'd still be fun to watch the first few rounds as Trump was confronted with actual facts.