Obama was a natural subject for this kind of experiment because there are so many readily available, high-quality video clips of him speaking. In order to make a photo-realistic mouth texture, researchers had to input many, many examples of Obama speaking--layering that data atop a more basic mouth shape. The researchers used what's called a recurrent neural network to synthesize the mouth shape from the audio. Recurrent neural networks are also used for facial recognition and speech recognition.)
As part of that, they've promised to "bring highly automated driving functions to market as a core competency from 2021." They announced they're rolling out "Level 3" automation--which means a car that can drive itself some of the time--in the A8 model this year with promises to bring fully autonomous vehicles to market in 2020. On the electric side, the company has promised a sporty little electric vehicle called the I.D. Instead, the company's engineers had built them to run artificially well under testing conditions (and only under testing conditions).
"Planes fly roughly 99 percent of the miles that they fly by computer. It's now to the place that it is not safe for humans to fly in a lot of conditions. If you could have a robotic surgeon that makes one mistake in 10,000, or a human that made one mistake in 1,000, are you really going to go under the knife with the human? As a counterpoint, however, there are lots of Americans who choose to drive rather than fly, fearing the latter more despite knowing that it is statistically much safer.
In the report, researchers at the Facebook Artificial Intelligence Research lab describe using machine learning to train their "dialog agents" to negotiate. At one point, the researchers write, they had to tweak one of their models because otherwise the bot-to-bot conversation "led to divergence from human language as the agents developed their own language for negotiating." In other words, the model that allowed two bots to have a conversation--and use machine learning to constantly iterate strategies for that conversation along the way--led to those bots communicating in their own non-human language. Already, there's a good deal of guesswork involved in machine learning research, which often involves feeding a neural net a huge pile of data then examining the output to try to understand how the machine thinks.
The world we experience is not the real world. Which raises the question: How would our world change if we had new and different senses? More recently, researchers in the emerging field of "sensory enhancement" have begun developing tools to give people additional senses--ones that imitate those of other animals, or that add capabilities nature never imagined. Researchers are working on other technologies that could restore sight or touch to those who lack it.
Despite the recent emergence of browser-based transcription aids, transcription's an area of drudgery in the modern Western economy where machines can't quite squeeze human beings out of the equation. That is until last year, when Microsoft built one that could. Automatic speech recognition, or ASR, is an area that has gripped the firm's chief speech scientist, Xuedong Huang, since he entered a doctoral program at Scotland's Edinburgh University. Huang and his colleagues used their software to transcribe the NIST 2000 CTS test set, a bundle of recorded conversations that's served as the benchmark for speech recognition work for more than 20 years.
It now provides data for 60 million homes across the United States that it has already assessed with its algorithms. For the past two years, Project Sunroof has walked people through all the information-gathering steps of installing solar panels: After you tell it where you live, its algorithms estimate how much solar energy falls on your roof, calculate how much solar panels would reduce your electricity bill, and deliver estimates from local installation firms like Solar City. Google created the data for this feature in-house, training a machine-learning algorithm on the common appearance of rooftop solar panels and then letting it loose on the cities and towns that Project Sunroof already covers. Right now, the company has analyzed installations on about 60 million buildings in the United States; it hopes to get to the remaining 40 million buildings in the next few years.
Uber didn't immediately respond to The Atlantic's request for comment on Sunday. His termination was tied to a federal lawsuit in which Waymo--the self-driving car company that spun out of Google--is accusing Uber and Levandowski of stealing its design secrets. Uber's request for arbitration would have allowed the ride-sharing giant to defend itself behind closed doors. Instead, Uber will have to go to trial to answer Waymo's accusations in what is sure to be an ugly showdown.
Without once saying the words "artificial intelligence," a stream of Apple executives described a vision of the near future in which Siri, the company's AI avatar, stitches together the company's many hardware products. After a strangely dystopian video in which Apple's apps go away and the world plunges into post-apocalyptic violence, Apple CEO Tim Cook led off the company's keynote at its big gathering for coders, the Worldwide Developers Conference, in San Jose. Through the demonstrations and talks, Apple's vision for Siri became clearer: It is an all-purpose stand-in for predictive, helpful intelligence across all Apple devices. "With Siri intelligence, it understands context.
Recurrent neural networks "learn" by repeatedly processing the data given to them. Instead of a typical computer program, which runs certain pre-set functions on a large data set, neural networks learn probabilistically what the set "looks" like. As they prepare this model, they spit out new approximations of the data set--data that wasn't included in the original set, but which could be. In the case of the type of program that Shane uses, it learns to model character-by-character: It figures out which character are is most likely for a certain spot, then it moves on to the next, and the next after that.