Back in September 2016, Google purchased Speaktoit, a company specializing in natural language technology. The company's main service at the time was API.AI, an SDK that allowed creating voice interfaces for mobile apps. Google re-purposed the tool to create Google Assistant Apps (formerly known as 'Actions on Google'). Read More
When Google acquired British artificial intelligence startup DeepMind back in 2014 nobody was really sure why. It was a secretive company without any consumer-facing products, but it clearly had some technology of interest in Mountain View. DeepMind is now part of the wider Alphabet structure, and now a little bit more is known about what goes on there, including the work it does for Google.
One of the latest projects to come out of DeepMind is called WaveNet, a deep neural network for generating more natural artificial speech recordings, and its use case at Google is obvious. Google Assistant should sound much more human now that WaveNet has been incorporated into it. Read More
Google has been using CAPTCHA and reCAPTCHA for years now to crowdsource information from images it has found online. On the one side, you prove to Google you're a human, on the other, Google gets humans to see small images of text from a book, numbers on a house, or various objects and animals and provide OCR for the text and numbers plus logical grouping and categorization for the objects/animals.
That's the main dig that XKCD's comic today is based on. It's a CAPTCHA but it's asking the user to say if there's a stop sign, just so it can properly use that data to control its self-driving vehicle. Read More
The madness of IFA 2017 may have settled down, but that hasn't stopped Huawei from unveiling its new SoC, the Kirin 970. Under its in-house HiSilicon brand, the Chinese giant seeks to push the boundaries of what is possible with mobile AI. And yes, there are the typical SoC improvements here, too. Read More
Robots are cool and, when designed and executed properly, robots can be quite endearing and awesome. But at $2800 a pop with very little added benefits, this one is definitely straddling the line to absurd territory.
Announced at IFA by QIHAN Technology (apparently a company specialized in AI and robotics), the Sanbot Nano is a 2.7 feet tall (848mm) robot that tries hard to look like a little kid. It has rotating head and arms, and is decked with sensors (IR, PIR, obstacle detection, touch sensor), 7 microphones, 250 lights, 2 tweeters and a subwoofer, a 3D camera, and a 10.1" touchscreen display. Read More
The Astro email app launched earlier in the year with the promise of helping you to clean up your inbox using AI. It's available for Office 365 and Gmail accounts and had garnered praise from users of both services. There's an integration that lets you manage your emails from Slack, and since June it's been possible to manage your emails with Astrobot Voice via an Alexa Skill on Amazon's Echo speaker. That same voice assistant functionality is now available directly from the app. Read More
Watermarks are the most common way to prevent images from being used without licensing, but they're not nearly as potent as you'd imagine. Research at Google recently published a paper called "On the Effectiveness of Visual Watermarks" regarding how easily watermarks can be removed, and how this can be prevented. Read More
According to a report from Bloomberg, Facebook is building a video calling device for the home, in what is essentially a new product category. It apparently has a laptop-sized screen and smart camera technology that gives users the impression of being in the same room as one another. Read More
Masayoshi Son, the founder of SoftBank, has made no secret of his plans to become "the biggest investor in technology over the next decade". Last year the Japanese corporation acquired ARM Holdings, and it already had a controlling stake in the US carrier Sprint. Son has even set up a special Billion Vision Fund, which aims to raise $100 billion for the purpose of making strategic technology investments. It's now come to light that SoftBank has used some of the capital to amass a roughly $4 billion stake in the chipmaker Nvidia. Read More
Google has been improving Gboard with the same type of tools it uses for speech recognition: machine learning. The budding technology is rapidly becoming a ubiquitous method for improving results and performance. If a network can be trained to accurately accomplish something in a performant way, odds are you'll see it introduced to any product it can be applied to. Gboard and text-input as a whole are no different, and we are reaping the benefits of improved corrections and predictions every time we swipe out a low-accuracy message to a friend. But how do these improvements work? Read More