Last week, like a lot of you, I imagine, I watched the State of the Union and tried to figure out what was true and what was fiction. Tied into several of the live-streamed press fact-checking streams, I found that the comments validating or invalidating what the president said came in so far after the comment was made that it would have been better to skip the speech and wait until the next day when the talk and the feedback were better matched.
Even then, it wasn’t linear, it was just collections of comments that were biased either for or against the talk. That’s not reality — that’s just mind-numbing pablum for those on either the left or right who really don’t want to think. I like to make up my own mind not have a bunch of folks on either side who basically are commenting from the sidelines force me into their increasingly delusional world view.
This week IBM Think takes place in San Francisco, and one of the programs that will be covered is its Project Debater. What makes this AI project unique is that it melds human interaction with deep learning to create a very different kind of solution that would be uniquely valuable to both litigation and politics. It could make the court system much fairer, and it also could make it far more difficult for a politician or biased media outlet to confuse or mislead us.
On this day before IBM Think, I’ll explain how this new hybrid AI platform that IBM is showcasing could make politics and litigation more beneficial and then close with my product of the week. This week it is the new Dell XPS 13 in Frost White, which makes an impressive statement with design, portability, materials and technology.
Creating a Hybrid AI
One of IBM’s unique positions, which really shouldn’t be unique, is that AI technology should be used to enhance rather than replace humans. There are lot of reasons for this. The most compelling for me is that if robots take over everything, who is going to buy stuff? No, wait, perhaps the most compelling is that in IBM’s world we aren’t obsolete, and the threat of becoming obsolete as a race is something I’d hope more of us could wrap our heads around.
Now we do have the concept of the singularity, in which man and machine become inseparable, but that is unlikely in my lifetime. If it were to arrive early, I’m pretty sure it would end my life, so an alternative path would be significantly more attractive.
Given that most AIs, even those designed to work with people, are created largely as isolated standalone systems that could, with a flip of a switch, no longer need their human operators, a different approach is needed to address this human obsolescence problem.
IBM’s Project Debater
This brings us to IBM’s Project Debater. This revolutionary approach uses a blend of crowdsourcing and deep learning AI to come up with something far better than either component can be alone. Project Debater uses the crowdsource model to define the quality of an argument, which then is used to train the deep learning AI engine and optimize what is initially a debate response.
This blends the AI’s affinity for facts and numbers, the objective part of the solution, to the human subjective view of quality. In an ideal case, what you get is potentially a self-evolving decision system better able to resist problems like confirmation bias, and make fact-based decisions that aren’t devoid of feeling and humanity.
Once again, looking at this ideally, it would be a blend of objective and subjective decision criteria, still based on verified facts, that is focused on human defined high-quality outcomes.
The foundation of the movie series Terminator would be a showcase of an objective AI-driven future without the human elements that Project Debater brings to the table. Of course, the world we currently live in is defined by an excess of subjective decision methods that are focused more on status and seeming right than on being right. Either extreme doesn’t bode well for our survival.
The State of the Union
In watching the State of the Union, I observed three problems with trying to evaluate the veracity of the information. One was that the fact checkers couldn’t keep up. Two, they had three classes of response, “right, wrong, misleading,” and the most common, misleading, was always nuanced negative, which made the media coverage look biased. Three, the response from the Democrats had little to do with the actual talk (not a surprise, given that it was written before the event and the speaker didn’t watch the address). This last is like watching a debate between two people who can’t hear each other’s positions. It was less a real response and more a “Look, I too can be on TV!”
What Project Debater does is use people to establish the quality of arguments on both sides of an issue to get to the strongest arguments. People in the crowdsourced pool vote only on the arguments they agree with, so they don’t compromise the result because of preset biases.
This would allow a real-time rating of the arguments on both sides prior to the address. For instance, let’s take the wall. The most powerful argument for it is likely the reduction of criminals coming into the U.S. The most powerful argument against it is that spending money on enhancing existing border security and services like the Coast Guard would have a greater impact on illegal immigration and increase the likelihood of catching the criminals at the border. Both positions would be backed by validated statistics to help the audience pick the better side.
So, while hearing the president talk, an observer would see both the strongest argument for his position on the wall and the strongest argument against, and then could form an opinion on which position to support. Unlike the media, or either side of the argument, these would be unbiased and fact-based.
I do think that this implementation likely would force the use of this tool in the decision-making feeding the State of the Union and that smarter presidents, not wanting to look foolish on TV, would use it to set their positions in the first place. If they didn’t, we would have the objective tools necessary to identify their incompetence and the need to remove them from office.
It is my view that this could significantly repair the existing world political process, which is currently more about what the left or right want, and less about doing what is right.
One other thing that an AI could add is a ranking of the arguments based on our unique circumstances. For instance, an AI could, and increasingly likely does, know more about you than you know about yourself.
We often don’t realize that the positions we take are against our own best interests, because we are convinced to take our positions by charismatic folks who don’t even know us, let alone have our best interests at heart. This could help fuel a stronger bipartisan approach in government, because it should be more interesting to us if something is to our benefit than if it is a left or right idea.
This wouldn’t improve government alone — it could have a massive positive impact on litigation.
More Effective Litigation
When I was watching the FTC/Qualcomm case a couple of weeks back, I was struck with what a difficult job the judge had, which was made more difficult given that the FTC couldn’t seem to place its evidence into any kind of decision matrix.
A judge typically wants to do a good job and reach the right decision, but this is at cross-purposes with the folks presenting their cases, who want to herd a judge to a decision favoring their side. In this kind of a case, each side typically consists of teams of attorneys, and there is only one judge, which makes it nearly impossible for the judge to see through the smoke. It seems more likely that judges will end up deciding the case on something other than the facts in evidence.
What the legal profession has is miles of documented arguments and positions, along with records over which side prevailed and at what level. In effect, much of the crowdsourcing part of the effort is cooked — it just needs to be fed into the AI so that a judge can get a real-time sense of whether what is being argued is both consistent with the evidence and consistent with case law and precedence.
It not only would force the two legal teams to up their game, but also would provide the judge with a near superpower in terms of determining the differences between facts and fabrications. In short, it potentially would turn a process that is often more performance art than anything else into something that would be far more likely to determine guilt or innocence accurately. It would particularly benefit those who are overmatched by their opponents in terms of resources.
IBM’s Watson had the legal profession as an initial target. This application not only would be extremely powerful but also would increase significantly the integrity of the legal process. IBM says that its Think Conference is about changing the world. This year, with Project Debater, that seems surprisingly accurate.
Finding ways to use the subjective skills we have while improving the outcome by ensuring that those decisions have a solid objective foundation would benefit the world. From the proliferation of nuclear weapons, to wars, to what to do about global warming, the proper blend of human/machine could make the world a far better place to live. It could make our products better focused, and our governments more responsive and less likely to get us killed.
Those outcomes become increasingly possible as more companies look at out-of-the-box approaches like Project Debater, and the world shifts from being about us vs. them to being just about making things measurably better over time. We have lot of things that need fixing. AI is both part of the problem and a huge part of the solution, and Project Debater is on the latter path. For those going to IBM Think this week, I’ll see you there.
The Dell XPS line is Dell’s equivalent to GM’s Cadillac or Ford’s Lincoln. It is a premium line still at the top end of affordable, which focuses on design, quality and experience over pure performance, price or durability.
The line came about, much like Microsoft’s Surface line, as a counter to Apple’s MacBook, but as Apple’s interest in PCs waned, it has shifted to more of a standout product on its own merits and less of a MacBook clone.
The XPS 13 Dell sent me was top of the line in Frost White. With an Intel Core 7 processor, beautiful 4K display, 512-GB NVMe SSD, and Windows 10 Pro, it costs just over US$2K. Battery life should be well over 10 hours, and it has a small pocketable color-matching power supply, so I don’t have to carry a backpack and I still have plenty of battery life. (My personal goal is to no longer have to lug my heavy backpack everyplace I go.)
One of the changes this year is the screen bezels are smaller, and the camera has been moved from the bottom of the screen to the top, which means it isn’t as good at focusing the folks you are video conferencing with on the state of your nose hair grooming (something I’m not going to miss).
With three thunderbolt/charging USB ports, it can support up to two additional 4K monitors (many of us have started to travel with one or two small portable displays so we can better work on the road), and it has a biometric power button that fires the machine up and ensures the person using it is you.
At around 2.7 pounds, this is a light laptop, and its Frost White color (I typically favor dark colors like black) is growing on me, as it looks like it came from the white science fiction future portrayed in many movies. (It would look impressively good in an advanced weapons or bio lab.)
With a 90 percent recyclable design, it is also surprisingly green, which is further supported by a Bronze EPEAT ranking for sustainability. It has Dell’s impressive Cinema solution for watching TV and movies, which would be far stronger if Amazon would finally allow us to download and watch Prime movies on a laptop (that is where most of my video library is).
In short, the XPS 13 is another showcase from Dell that blends design and performance into something you can be proud to see on your desk, and it is my product of the week.