- • The dangers of “deepfake” videos
- • A paradigm shift in AI
- • You’ll soon be able to talk to your car
AI-powered talking heads…
Researchers from Samsung and the Skolkovo Institute of Science and Technology used artificial intelligence (AI) to create a talking Mona Lisa. To do this, all they needed was a single image of the painting.
AI-Powered Talking Mona Lisa
Source: Skolkovo Institute of Science and Technology
To animate the talking head, they took an existing video of somebody speaking and mapped the face. They combined that with the Mona Lisa image to generate the movements… And then they refined it to the point where it became realistic.
This is amazing. AI took a single image of Mona Lisa and brought her to life.
Imagine the educational applications. We could take historical figures, say George Washington or Abe Lincoln, and animate their speeches.
This would bring history to life for students. And make learning more interactive and real.
But there is a downside…
If you didn’t know any better, it would look like the talking Mona Lisa was real. It’s very convincing.
And that means this technology could be used to manipulate people and make divisive actions even more powerful.
For example, this could be used to create a fake speech from any politician and put it out on social media. Or it could be used to frame or defame any individual, regardless of whether or not he is a public figure. That’s a disturbing thought.
And one more detail: The research lab that produced this video is in Moscow. Given the Russian government’s penchant for spreading “fake news” over social media, this development should give us pause.
And keep in mind, this is not expensive technology. AI is just software. And once it’s out there, it’s out there. To create your own talking head, all you need is the software, an image, and a voice recording. That’s it.
So, AI can be used for good or for bad. As I said last week, our ethical choices will be the biggest issue society grapples with over the next 30 years. And ethics can vary greatly from one culture to another.
Turning the tables on AI’s conventional wisdom…
The team at Google’s DeepMind just put out research showing how to apply AI to small data sets. That means an AI can now work effectively with minimal data as a starting point.
Previously, it was thought that you needed hundreds of thousands (if not millions) of images to train an AI to “learn” to recognize objects. Dogs… cats… cars… trucks… humans… flowers… basically, the larger the data set, the better for training an AI.
But the team at DeepMind trained an AI on just 13 images, basically 13 examples, of something. And get this: The results were better than those AIs that were trained on massive data sets.
The AI was able to infer and categorize images more efficiently. It was able to “think” more like a human brain to solve problems, rather than having to see nearly every possible example of an object that it was trained to recognize.
This turns the tables on what was common knowledge in the industry. And the implications are massive.
Up to this point, we thought that the only way to train an AI was with large data sets. And that meant the countries and companies with the most data would dominate the field.
But this new research proves that AI can work with small amounts of data… And it can outperform best-in-class AI built from large data sets.
This opens the door to all kinds of innovation. We’ll see advancements at a rapid pace, because small players can get in as well. That’s right, technology like this levels the playing field.
Any early-stage company or entrepreneur will be able to compete with a large corporation with seemingly unlimited resources.
And we’ll see progress in areas that have been difficult to address with AI up to this point.
Applying AI to medical applications comes to mind. That’s been hard up to this point because medical records are messy, there is no common standard, and there is no centralized repository for them.
That has made it difficult to get large amounts of medical data to train an AI… Especially for rare diseases.
But now that we know how to apply AI to small data sets, we can turn it loose in the medical field. That will lead to new cures and treatments. This new approach is going to accelerate the use of AI to address very specific problems and challenges.
Voice assistants are coming to your car…
BMW just announced that Intelligence Personal Assistant, its AI-powered personal assistant, will come standard in new BMWs, starting with the 3 Series.
Intelligence Personal Assistant will allow drivers to control most features with voice commands.
They can control the radio… adjust the temperature… check their gas or battery level… get GPS directions… and more, all with voice commands. Anything you do with your hands now, you will be able to do with your voice.
Now, you may wonder why this technology is just now coming to cars, when we have had things like Alexa, Google Home, and Siri in our homes for years now.
The reason is simple. Car manufacturers start working on a design model three to four years before production. So, they started working on the current 2019 model cars back in 2015 or 2016.
That means there’s always a technology lag time with our cars.
The company powering BMW’s Intelligence Personal Assistant is Nuance Communications… A company long-time Exponential Tech Investor subscribers may remember. Nuance provides AI for voice interface, and one of its industry areas of focus is automotive.
This is a fantastic application of AI. Using a voice interface in an environment where the machine operator (the driver) should have his hands on the wheel is something that should be in every vehicle. It will make for safer drivers and less accidents.
And when self-driving cars start to hit the road, the voice interface will be equally important. Consumers will sit down and say, “Siri, take me to work”… and you’ll be on your way in seconds.
Editor, The Bleeding Edge
P.S. As a reader of The Bleeding Edge, you already know that technologies like advanced AI… genetic editing… and autonomous vehicles aren’t decades away. They’re here. Right now. And this tech will change our society in ways very few can predict.
But on Wednesday, June 12, at 8 p.m. ET, that’s precisely what I’ll do.
I’m hosting an online technology investing summit. During the event, I’ll share my four biggest predictions for the future of technology. And I’ll talk about four bleeding-edge tech companies that could make investors 10 times their money in the years ahead.
The event is free to attend. And if you’re interested in investing in the companies powering the technology we talk about here, you’ll want to be there. Go right here.