If you’ve noticed Google doing a better job of understanding what you say using speech recognition on your smartphone lately, you’re not crazy. Google’s voice search has indeed become more accurate, thanks to advances in artificial intelligence, the tech company announced today.
“Today, we’re happy to announce we built even better neural network acoustic models using Connectionist Temporal Classification (CTC) and sequence discriminative training techniques,” Google Speech Team members Haşim Sak, Andrew Senior, Kanishka Rao, Françoise Beaufays and Johan Schalkwyk wrote in a blog post today. “These models are a special extension of recurrent neural networks (RNNs) that are more accurate, especially in noisy environments, and they are blazingly fast!”
The new models are working in the Google app for iOS and Android, as well as dictation on Android, which works inside of some third-party apps, the team members wrote.
Google has reported improvements in voice search not once but twice this year. Clearly the company has been investing in the underlying technology. RNNs are one increasingly popular approach to doing deep learning, a type of artificial intelligence, and Google is widely thought to have a deep bench in deep learning.
“In addition to requiring much lower computational resources, the new models are more accurate, robust to noise, and faster to respond to voice search queries — so give it a try, and happy (voice) searching!” wrote Sak, Senior, Rao, Beaufays, and Schalkwyk.
Read the full blog post for more detail on how the team managed to get the new performance gains.
Hammer & Chisel didn’t have much luck with its pioneering iPad game, Fates Forever. It launched last year as a mobile version of the popular multiplayer online battle arena (MOBA) games. It won a lot of praise, but it didn’t get traction with players.
So Jason Citron, the founder of Hammer & Chisel, made the tough choice of shutting the free-to-play game down and pivoting to something new. It was an emotional time, but the team came through with something new, he said. The result is Discord, a new free voice communication chat app for multiplayer mobile games.
The app is in alpha testing now, and it is getting a lot of traction from players who want a reliable way to talk to team members during a game. Citron hopes it will make a splash in the $ 1.7 billion voice chat market, which is growing within the larger esports professional gaming market.
“We think it’s a necessary, missing piece for the esports movement,” Citron said in an interview with GamesBeat.
Citron said investors have been supportive, in part because Discord is taking off, even though it’s still in alpha testing.
Citron’s Hammer & Chisel went through YouWeb’s 9+ incubator, which provided initial funding. Early investors included Accel, TWI, and IDG. Then the company raised another round of funding earlier this year from Silicon Valley venture capital firm Benchmark Capital and Chinese Internet gaming company Tencent. Citron in particular got a lot of attention for his new startup because he built Aurora Feint, later called OpenFeint, in the early days of the Apple App Store and then sold it to Gree in 2011 for $ 104 million. That was an extraordinary, fast-moving experience for an entrepreneur who had a big pay day at the age of 26.
Above: Discord voice and text chat for mobile games.
Image Credit: Hammer & Chisel
But with Hammer & Chisel, his first move wasn’t a success.
“We were very careful in spending our money,” Citron said. “We are an experienced team. We’ve done this before. We spent enough to get Fates Forever out. It was clear it wasn’t going to be a big business. We didn’t burn a lot of money on marketing.”
Fates Forever had a lot of critical acclaim as a pioneering MOBA on mobile. It had a lot of attention from Apple, which featured it prominently.
“We could theorize about why a game didn’t work, but at the end of the day, it didn’t work,” he said. “Making games is challenging. It’s a combination of novelty, pop culture, and operations. It was a very emotionally difficult time. We all love games. A lot of the most important relationships with my wife and best friends are built around games. A lot of our best memories we have are about playing games together. That’s what motivates me. I want to make things that help people become closer together through games.”
During the process of creating Fates Forever, the team noticed that voice and text chat for mobile hadn’t advanced. Over the past decade, there hadn’t been much progress as voice chat moved from the PC to mobile. So the team started experimenting.
Above: Fates Forever
Image Credit: Hammer & Chisel
“Out of that, we built Discord,” Citron said.
Others have devised ways to handle game communication, such as Curse, Skype, C3, Hipchat, and Teamspeak. But Eros Resmini, who handles business operations at Hammer & Chisel, said that none of them are that satisfying. In fact, you can think of Discord as a mashup of two useful tools.
“If you think Skype and Teamspeak had a baby and it hand all sorts of super powers that is parents didn’t have,” Citron said. “What was basically a skunkworks project appears to be the most promising product we’ve built.”
Citron said that Discord became popular because it has low-latency communication, or minimal delays between when someone says something and another person hears it.
For much of this year, the company has been working on Discord. The networking infrastructure is built in Erlang, a technology that AT&T created in the 1980s for telecommunications. The system is spread across nine data centers around the world. The company has done tests to make sure that the latency is good.
Resmini also noted that esports competitors — or professional gamers who play games for money prizes — were worried about security. With Skype, it’s easy to get somebody else’s personal internet protocol (IP) address because the communication happens peer-to-peer. Citron said that Discord works through server infrastructure, so it’s impossible for anyone to obtain another player’s IP address. If a voice server in the distributed cloud gets attacked by flooding it with data, it shifts the communication to another server.
“We have special hardware on our endpoints that does distributed denial of service (DDoS) protection,” Citron said.
Above: Discord for mobile game chat.
Image Credit: Hammer & Chisel
Discord runs in a browser or as an app. So getting into the chat is easy, as all you have to do is share a link and type in your user name.
“That’s the magic moment,” Citron said. “It’s really easy to get people into a chat.”
Early reaction has been favorable for Discord. The company hasn’t spent money on user acquisition, but it is spreading via word of mouth.
“We’ve thought about doing esports sponsorship deals, but it’s growing so fast on its own that we haven’t yet,” Resmini said.
Over time, Hammer & Chisel hopes to introduce ways to monetize Discord. Those options include offering paid customization to players, who can pay for special “emoji” text chat or other kinds of stickers or decals.
Citron founded Hammer & Chisel as a mobile-first game company in 2012. Eros Resmini, who also worked at OpenFeint, joined to help get Discord off the ground.