Technology Predictions for 2020

Tue, Dec 31, 2019 with tags technology , predictions , deepfakes , ai , ml , election , quantumcomputing , publicpolicy , privacy

In the past, when I had more time to write, I made a nearly yearly event of creating a list of technology predictions for the next year. It’s ironic that as my professional role has taken on more of a leadership component, I’ve stopped doing this as much. However, now that I’m officially responsible for emerging technologies at Verizon, it’s time for me to again publish a list of predictions. Without further ado, here are my technology predictions for 2020.

Deep Fakes and the Election

It’s disturbingly easy to create deep fake vidoes. While we can laugh at Rowan Atkinson being deep faked into a Christian Dior J’adore ad, the fact is that it’s extremely cheap to create a compelling deep fake. Timothy Lee, a writer for Ars Technica, was able to create a moderately compelling deep fake over the span of two weeks for only $552. If we extend this to the resources of presidential campaigns and/or nation state actors, it’s easy to see how a determined individual could use a body double and sufficient resources to create an incriminating video of a presidential candidate. It’s likely that there will be at least one deep fake that is so compelling that it will fool substantial portions of the traditional news media for a cycle or two (there are elements of the news media that choose to be fooled by anything that supports their world view - I’m talking about major news outlets here).

AI/ML as a Presidential Debate Question

2019 has been an adventureous year for AI/ML. There have been a variety of notable failures and questionable uses, some of which have attracted congressional interest, such as the issues with potential gender bias on the Apple Card. We’ve also seen issues wiht self-driving cars, the rise of AI/ML in hiring, and potential dangers from GPT-2.

We’re still a good nine months away from first Presidential Debates of the 2020 election and with the way that AI continues to fail and the way that it’s faults are increasingly being highlighted in the popular press there’s a good chance that we’ll get a question about AI/ML policy in the 2020 debates.

Quantum Computing Hits a Milestone

2019 was the year that Google demonstrated quantum supremacy. Or, maybe they didn’t demonstrate quantum supremacy. It’s a complicated topic that even in the hands of a master communicator, can be difficult to fully understand. With so much money at stake, a potential shift that could replicate the scale of the Internet, it’s likely that we’ll continue to see many confusing announcements from the likes of Google, IBM, D-Wave, Microsoft, NEC, Alibaba, and more. Up until this point, it seems like each of the announcements has been met with extreme skepticism from the other players. My bet is that in 2020 we’ll see a broad announcement from one of the players announcing limited quantum supremacy in such a way that it leaves little room for debate about whether or not this incredible milestone has been achieved.

AI Fairness as a Regulatory Issue

There’s a saying in Washington that the worst thing that you can do is to get your boss' name in the Washington Post. When I worked at a financial institution in the area, we said the worst thing that we could do was get our CEO called in front of Elizabeth Warren. Over the past year we’ve seen a number of technology executives hem and haw their way through congressional testimony - usually providing vague non-answers about what their company has actually done. Given the huge number of spots where AI/ML is being applied and, frankly, how poorly governed almost of these models are, it’s likely that 2020 will see congressional hearings about creating a regulatory basis for AI/ML across a broad swath of industries.

US Continues to Lack a Federal AI/ML Policy

This one is almost too easy. Given the current administration and makeup of congress, the chances are very high that we’ll end 2020 without a clear national vision of how to apply AI and ML to transform the country. This will increasingly allow competitors with a cohesive focus, such as Canada, to catch up and allow China to gain even more of a lead in the AI/ML race.

DNA Privacy

The discovery of the identity of the Golden State Killer was a watershed moment in forensic science. Investigators were able to use publicly avilable DNA databases, which were previously designed for geneology, to narrow down a range of suspects based on familial relations. This brings up dramatic questions about your identity being compromised by the unwitting actions of your family members and the potential impact of a surveillance state.

After the Golden State Killer case there were numerous other cases where forensic geneology was used to identify a smaller set of suspects and cold cases dating back decades were quickly resolved. In some cases the suspects quickly pled guilty, and we have the first conviction in the case of previously unsolved 1987 murders or a Canadian couple traveling in Washington State. There’s a strong case to be made around the Fifth Amendment and we should expect to see some of these cases head to the Supreme Court in 2020.

Streaming Collapse

A couple of years ago the only streaming service that people subscribed to was Netflix. Today we have Hulu, Disney+, Amazon Prime, ATT TV Now, Pluto, YouTube TV, Twitch Prime, Noggin, Sling, BritBox, Apple TV+, and the upcoming Peacock and HBO Max. I know, you’re thinking that I’m missing some there, and you’d be correct. Quite simply, that’s a lot of different services and we’re already seeing pullback with PlayStation Vue shutting down at the end of January 2020.

It’s to the point where a customer needs to subscribe to a number of different solutions to get the content they want, and that’s just financially viable. While the economy is still good, many consumers won’t feel need to downsize their streaming solutions. However, most of these solutions have little sticking power and therefore if the economy even begins to sputter, expect to see a massive number of cancellations (even bigger than what ATT TV Now has been seeing) that will result in the death of a number more streaming solutions. While the collapse of small services is nothing new - I wouldn’t be surprised to see HBO Max, Hulu, YouTube TV, or Sling not make it through the end of the calendar year.

Federal Facial Recognition Bills

2019 saw a number of different facial recognition bills appear across the country. Over time these have been simultaneously expanded and partially rolled back, as was the case with San Francisco who realized that it’s nearly impossible to use modern mobile phones without facial recognition. This creates an incredible patchwork of regulation that will make it nearly impossible to innovate in this space. Expect to see some tech savvy lawmakers attempt to push through a common set of federal guidance for facial recognition systems.

AutoML Failures

AutoML from companies such as Google, H2O.ai, and Data Robot seems like a miracle worker. You can give it some data and the algorithms will poke and prod with different models and feature combinations to uncover amazing insights. Many over-excited C-level executives have bought into this and probably uttered something like “This will democratize AI/ML and make us a true powerhouse!”. And they’re kinda correct, Google’s AutoML beats models created by most Kagglers. This sounds really attractive, but the fact is that these algorithms can only do exactly what they’re programmed to do and without a robust model governance process, companies are introducing extreme risk. During 2020 expect to see a company do something truly galling and to blame it on AutoML. It wouldn’t surprise me if this was a blatantly racist model applied to an area such as lending or security.