Van Bryan here, Jeff Brown’s longtime managing editor. Over the next few days, we’re taking stock of the year that was and looking forward to the year ahead.
Yesterday, we learned what Jeff sees ahead for 5G. Today, we turn to the world of artificial intelligence. As readers know, AI is progressing faster than most believed possible. That will have ramifications – good and bad – for society.
Read on to discover why Jeff believes AI will become an integral part of our world in 2020.
– Van Bryan
Van Bryan: Jeff, today we’re talking artificial intelligence (AI). But before we get to what you see happening next year, could you recap the biggest AI stories you saw in 2019?
Jeff Brown: There were really two major trends in AI that stood out this year. First, we saw a proliferation of early stage hardware companies – specifically, semiconductor companies – release application-specific AI semiconductors. In other words, we saw a flood of hardware that was designed to run AI applications. And, of course, anybody could purchase these devices.
Researchers, companies, academic institutions, and even individuals could buy these AI semiconductors and begin experimenting with AI applications. That’s why 2019 was such a huge year for AI. The hardware needed to run AI was readily available.
Van: You mentioned there were two trends. What was the second?
Jeff: The second big advancement for AI was that we saw tremendous strides in utilizing smaller data sets. Traditionally, an AI had to “learn” by studying massive data sets. That’s how it improved.
A good example is Google’s AI training model called “Bert” (Bidirectional Encoder Representations from Transformers). This AI was trained with 340 million data parameters.
But just this year, Google’s researchers produced a data set that was 60 times smaller than Bert’s. And the results from the AI that trained on it were about as good.
This was just one example, but it’s happening across the industry. And the result of this is that AI is being “democratized.” That means that artificial intelligence isn’t just for large corporations anymore.
Because AI semiconductors are readily available – and because these AI applications can run on smaller data sets – smaller companies and even smaller governments can have access to AI.
That’s why we’ve seen such an explosion of AI applications in 2019.
Van: Great insights, Jeff. That was 2019. Let’s talk 2020. What do you see happening with AI in the new year?
Jeff: I have a few predictions around AI. The first is straightforward. In 2020, we will see the first fully autonomous passenger car be available for purchase. I’m talking about a self-driving car you can buy or lease for your own personal use. It will be able to take you from point A to point B. And the car will drive 99% of the way by itself.
Van: We’ve been covering the progress of self-driving cars in the pages of The Bleeding Edge. What makes you so confident that 2020 will be the year that autonomous vehicles go up for sale?
Jeff: Remember that self-driving vehicles rely on artificial intelligence. They must “learn” how to be safe drivers. And the underlying semiconductor technology used for AI will double in power over the next 12 months. The amount of data that’s being collected from fully autonomous vehicles will more than double year-over-year.
And the AI algorithms – the software that makes self-driving cars possible – is improving significantly every month. So the AI algorithms won’t be twice as good from 2019 to 2020. They’ll be about five times as good as what we have today.
That means we’re getting to the point where these cars are essentially fully autonomous. It’s just a matter of time before a company puts one of these cars up for sale. I believe that will happen in 2020.
Van: You mentioned you had several predictions around AI. What were the others?
Jeff: The second prediction is that we will continue to witness many more examples of artificial intelligence outperforming the very best humans at specific tasks.
An AI has already beaten the world’s greatest human Go player, which is an ancient Chinese strategy game. An AI has also bested humans at Dota 2, which is a massively multiplayer online game. And AIs are even beginning to outperform human physicians and radiologists when diagnosing disease from X-rays and MRIs.
This trend isn’t going to slow down. It’s going to speed up. In 2020, many of the world’s top “experts” in a task won’t be humans, they will be artificial intelligences.
That might make some of us uncomfortable. But that’s certainly the direction we’re headed. And speaking of AI in health care, I have one more prediction for artificial intelligence in 2020.
Van: Please, go on…
Jeff: In 2020, I predict an AI will discover a completely new drug therapy without any assistance from a human. In other words, a human will provide the data for the AI to study. But artificial intelligence will develop the new drug therapy all on its own. I believe that will happen sometime in the next 12 months.
And this lines up with what I learned when I attended the 2019 STAT summit in November.
(To catch up on Jeff’s insights from the STAT biotech conference, go here.)
At STAT, I listened to a panel on artificial intelligence. The big question up for discussion was how long it would take an AI to find new drug candidates without human involvement.
Hal Barron, chief scientific officer at pharma giant GlaxoSmithKline said that we would see this happen in one to two years. He later rephrased his comment, probably hedging slightly. He said…
“I think we’ll have drug targets discovered in one to two years… And I would predict within the next two to three years, we’ll have examples of drug targets discovered by machine learning that unravel biology.”
I agree with Barron. But I don’t think it will take two or three years. AI will discover a new drug therapy for humans sometime in 2020. And because AI can come up with more precise therapeutic solutions, there won’t be as much trial and error. The Food and Drug Administration (FDA) approval process will be sped up.
AI is going to have a profound impact in 2020, especially in health care.
Van: Thanks for sharing your predictions, Jeff.
Jeff: Happy to do it for readers of The Bleeding Edge.
P.S. Van here again. Tomorrow – Christmas Day – we’ll continue our 2020 prediction series by looking at another tech trend on Jeff’s radar: CRISPR genetic editing technology.
Of course, readers will want to spend time with family in friends. But if you find time between opening presents and preparing the Christmas dinner, I encourage you to tune in to The Bleeding Edge for your daily dose of tech.
Like what you’re reading? Send your thoughts to [email protected].