CES 2019: A Tech Solution for a Standoff


To say I’m typically not a fan of CES would be an understatement, but this year was very different. A combination of better logistics and fewer people at the show (likely tied to the government shutdown and China drama), coupled with some truly earth-shattering content, made CES a must-attend event this year.

One of the things that jumped out at me — given that the show took place while the controversy heated up over that damn silly wall President Trump wants to build — is that IBM announced a compelling fix. It’s a fix that actually could improve the U.S. government overall.

In addition, I had an interesting chat with Frank Shaw at Microsoft on big takeaways from the show, and I think he offered the best one: that this show not only showcased the death of the tech market, but that its death would be a good thing. That should wake up a few folks.

I’ll cover all that this week and close with my product of the week: the HP Spectre Folio laptop. Although it was a product of the week last year, it made such a huge difference for me at this show that I’m bringing it up again.

CES 2019’s Fix for the Clash Over Trump’s Wall

Overshadowing CES this year was the government shutdown, which not only prevented a lot of government speakers from attending the show, but also raised the specter of many of us running into a nightmare when we headed home (due to a potential TSA strike). The standoff was on my mind when I met with Ranit Aharonov, IBM’s manager of debating technologies, and Noam Slonim, IBM’s senior technical staff member of debating technologies.

I was looking forward to the meeting, because I once trained to be a debater when I was thinking of becoming a lawyer. Sadly, my school’s debate program was killed due to lack of budget. What IBM’s Project Debater does is use a new form of artificial intelligence — kind of a blend of machine learning and deep learning — to capture the best arguments on a subject.

It does this by training the system with humans (which seems more like machine learning) but it doesn’t use data scientists as an interface (which is more like deep learning). The machine trains itself by observing submissions of arguments on a topic from real people, as well as the results of real people comparing two arguments on the same side and choosing which is better.

For instance, let’s say we were debating whether the world was flat. People would submit arguments on both sides and get back arguments on the side they supported and then pick the best ones. They would compare arguments on the same side in order to minimize bias. The system then would trains itself, utilizing its access to a deep repository of true facts, in order to determine the best arguments on a topic.