Viv Uses Deep Learning to Teach Computers How to Learn
We’ve talked before about voice recognition software and even wrote an entire article using the Nuance (NASDAQ:NUAN) voice recognition software as a sort of experiment to see just how far voice recognition has come along. We weren’t blown away by the experience but we could see that voice recognition has come a long way. However, every time we have to yell “speak to a human” into our phone when speaking with the robotic voice recognition interface at our local bank, we wonder just how long it will actually be until computers can understand us. What is stopping us from just talking naturally to our computers and having them understand us? This is where we need to start distinguishing between voice recognition and the ability for a computer to comprehend what we are actually saying. This is where deep learning comes into play. Deep learning can teach computers how to read things and comprehend them. Deep learning can teach computers how to learn.
Voice recognition is the ability for a computer to transcribe the words that come out of our mouths with near 100% accuracy. For the purposes of where we are at today, Nuance (NASDAQ:NUAN) can pretty much do that. Their entire company is focused on improving voice recognition, and the fact that Apple chose them to power Siri should speak loads as to how good their technology is. Voice recognition is not the problem here. Understanding speech is the problem. This is why chatting with chatbots these days is generally a waste of time. What we need to do is create an interface that understands what humans are saying whether that comes in the form of text typed into a chatbox or voices directed into a speaker. What we need to do is teach computers how to comprehend humans by using deep learning. This is exactly what a company called Viv is doing.
Founded in 2012, Viv has taken in $30 million in funding to build a technology that goes about “radically simplifying the world by providing an intelligent interface to everything“. What Viv is really trying to do is make your computer comprehend what you are saying when you speak to it or type commands to it. If Viv has its way, soon everyone will be able to talk to everything.
You might never have heard of the founder, Dag Kittlaus, but he’s actually the co-founder of Siri which was acquired by Apple in 2010. It should come as no surprise then that Viv is also using Nuance voice recognition technology for the “voice recognition” part of their technology. Voice recognition isn’t what Viv is all about. Viv is all about using deep learning for comprehension.
Viv’s claim to fame is a computer science breakthrough called “dynamic program generation” which is essentially “software that writes itself”. You don’t need to provide any context because the computer is figuring all of that out on the fly. Below you can see a (rather blurry) example of how the technology can break down the structure of sentences in just milliseconds:
Essentially, the computer is learning how to learn. Mr. Kittlaus recently demoed his Viv technology to a live audience at a TechCrunch Disrupt event and gave it the following challenges:
- Will it be warmer than 70 degrees near the Golden Gate Bridge after 5PM the day after tomorrow? (It actually answered that question.)
- Send Adam 20 bucks for the drinks last night (It pulled up a Venmo to Adam and then put in the comments that it was for drinks last night)
- Send my mom some flowers for her birthday. (It pulled up a bunch of flowers, then he asked for tulips, and it pulled up tulips. It then knew his mom’s address and in one click it paid for and sent them.)
- Get me a nice room in Palm Springs for Labor Day Weekend (He booked a $1,000 a night hotel room in about 10 seconds which was kind of slick. Is it just us or is that just way too damn much to spend on a hotel room?)
- I need a ride for six people from my office to Madison Square garden (It pulled up a few Uber rides nearby that could handle the lot of them).
The fact that you can quickly pay people, book $1,000 a night hotel rooms, or send your mom flowers isn’t what is exciting here. What is exciting is the level of comprehension being demonstrated such as in the first example above.
Viv isn’t the only company trying to use deep learning to help computers understand humans.
- A Canadian company called Maluuba has taken in $11 million to develop a natural language technology that has been already deployed across 50 million devices. One of their Board Members, Dave Grannan, was previously the CEO at Vlingo which was a speech recognition service for mobile phones that was acquired by Nuance in 2012. Maluuba is nowhere near human levels of comprehension but they are showing some promising results recently besting Facebook, Google, and IBM in computer reading comprehension.
- While not directly related to computer comprehension, one company called ObEN is doing something quite interesting. ObEN has taken in an undisclosed amount of funding to build a technology that can take any voice and turn it into a “human voiceprint”. For example, you might read 5 minutes of text to the program and then after that it will be able to duplicate your voice. Kind of creepy, but useful for hardcore VR simulations perhaps.
- Lastly, a Chinese company called Mobvoi is said to be the only firm in China that is offering a Chinese voice recognition solution that is so impressive even Google invested in it. Mobvoi has taken in $71.6 million so far and recently debuted a “fully Chinese” smartwatch called Ticwatch 2.
No doubt there are many other bright minds trying to teach computers how to learn. We should inevitably see one winning company emerge that can provide a computer learning API that is used by every kind of user interface in the future using an “Intel inside” business model. We see Viv as having an advantage here because of the pedigree of the people behind it. These guys built and sold a successful company to Apple, and then left their nice jobs at Apple to build something grand. We can’t wait to see what form the first implementation of their technology will take. While retail investors may not be able to invest in Viv yet, they can buy shares today in voice recognition company Nuance (NASDAQ:NUAN) as a sort of “picks and shovels” play on this theme.