x
Jarvis.ai is an artificial intelligence (AI) powered personal assistant designed to help you take control of your life and achieve success in the digital age. From its base as a fully-featured traditional personal assistant, Jarvis.ai introduces several innovative features into this space that we believe will transform how people interact with their personal assistants in the future. These features include:
Full integration with chat platforms, allowing your personal assistant to communicate and interact with you wherever you are, using whatever device(s) that you have at hand.
An innovative monetization system that rewards users for interacting with their smartphone apps using Jarvis.ai—a revolutionary app store business model that we believe will change how apps are built and used in the future.
An adaptive learning system that constantly learns from your personal interactions, proactively providing you with information before you need it—eliminating the need for humans to manually train personal assistants.
Jarvis.ai is a product of Jarvis Labs Inc., headquartered in Toronto, Canada.
Can you provide a little background on Jarvis.ai and the history behind its creation?
Jarvis.ai is the product of more than five years of research and development into virtual assistants. Our team has extensive experience in artificial intelligence, mobile software engineering, mobile marketing, and personal services. We've also been closely following advancements in virtual assistants, including Amazon Echo's Alexa. The most recent breakthroughs in artificial intelligence technology have been achieved through the application of deep learning algorithms – a branch of machine learning that allows computers to learn how to understand and analyze complex patterns in data – to problems related to natural language processing (NLP). Jarvis.ai leverages several deep learning technologies such as recurrent neural networks (RNNs) and long short-term memory units (LSTMs) to allow its users to communicate with it using natural language. This feature will be available on our flagship app, Jarvis Assistant.
We've been working towards the launch of Jarvis.ai since October 2013, when we raised a seed round of funding ($500K), and officially began operations in September 2014.
The name Jarvis is obviously a reference to Iron Man's AI "Jarvis." Where did the inspiration come from?
To be honest, we really don't know where the inspiration for the name came from! We were originally using the working title of "Jarvis," before we had settled on Jarvis.ai, and when we came up with the concept for the AI behind our product, we just decided to stick with that name. But now that it's out there, someone told me to check out an ebook titled "I am Iron Man," which I did. It's a futuristic sci-fi story about an entrepreneur who invents a new worldwide currency called Jarvis. I haven't reached the part where it mentions our AI yet, but based on what I've read so far, there's nothing to be worried about!
How does Jarvis.ai improve upon Alexa and Siri?
Jarvis Assistant is designed to go beyond the capabilities of traditional personal assistants, by providing users with the ability to talk and chat to Jarvis from any device, anywhere. In addition to traditional voice recognition features, Jarvis.ai's natural language processing technology is capable of understanding a wider variety of inputs than other virtual assistants currently on the market—including words and phrases in different languages.
Jarvis will also be able to understand context through the use of long short-term memory units (LSTMs), and will be able to provide users with a personal experience based on their preferences. In other words, as our user interacts with Jarvis – both directly, and through the features that it provides to them – it will learn how they like things done, and act accordingly.
Siri's approach is more closed off—it only allows one device to communicate with it at a time using SiriKit APIs. Users are limited in terms of what apps they can access and use. Alexa has taken an "open" approach, making its SDK available to third-party developers so that anyone can develop for Alexa; however, there is currently no marketplace or revenue sharing available for developers. Jarvis will offer the best of both worlds to third party developers: a unique platform for monetization, with access to our revenue sharing program once we begin charging users.
Tell me about your experience in the virtual assistant space and what inspired you to become an entrepreneur in this field?
The founding members of Jarvis have been involved in AI research since 2010. Our backgrounds include post-docs at Toronto's renowned Vector Institute, as well as work on deep learning neural networks with teams at Facebook and Microsoft Research Connections (MSRC). We had all worked together previously on different projects while working towards our PhDs—one day we started talking about how cool it would be if we could create a smart home that followed us around where we went, and we came up with the idea for Jarvis—which led to us starting our company.
Jarvis.ai was among the first companies to take a deep learning approach to virtual assistants (Alexa was around before us, but they use more traditional approaches). Amazon has just recently announced that they are adding LSTMs to the Alexa platform. Of course, this is what we've been doing since day 1! We're pretty far ahead of everyone else in terms of natural language processing capabilities; however, being early means you have no data on which to base your models—which makes it really hard for computer scientists to do anything useful without years of training on large datasets! For example: imagine trying to write an email client by hand, and having to have all of the emails from every user that you'd ever want to support; and then building a model for each one – but each email is different!
We're currently raising money to hire talent who can develop our models further. Our goal is to reach human parity in this field within 24 months. The average person won't know much about machine learning, but they will be able to talk with Jarvis.ai just like they would another person (most people already do this with Siri). They'll throw it curveballs, ask bizarre questions, say things backwards or in other languages; we are intent on making sure Jarvis understands everything—and responds appropriately.
Did you encounter any hurdles when developing your product? If so, what were they?
The biggest hurdle to developing Jarvis was the fact that we had no training data. It's hard to get people to use a product if it doesn't work very well! In order to collect this data we created a mobile app called "Jarvis Beta" which helps us collect user feedback and model our responses using deep learning. We did not anticipate how excited people would be about working with us. Our servers melted down within 5 days of being online—we thought that maybe 50 or 100 people might want to play around with a beta app but there was huge demand for it, hence why we had so many problems with server performance.
Now that you've developed your solution, how do you plan on capturing a market share from giants like Amazon with their Alexa / Echo offering?
Jarvis.ai will have several things which make it very attractive to users, and hence why we believe that we are not directly competing against Amazon in the same space:
A) Jarvis uses deep learning where as Alexa does not! We're working towards human parity within 24 months — right now, our most advanced model can understand random questions better than a 4-year old child. On top of this, we are able to assist you with all sorts of tasks; whereas Alexa is limited (currently!) to just answering questions about itself or the weather – eventually though, once we reach human parity then Alexa will be able to do everything Jarvis does. Beyond that, Amazon has a lot of catching up to do in terms of natural language processing and machine learning — but given that they now have access to every utterance ever said through Alexa (all the data is stored centrally so it's accessible to all developers for training) then I imagine that they'll catch up much more quickly than we did!
B) Jarvis.ai has been around for 2+ years and because we've built a mobile app which helps us collect training data — this means that when someone interacts with Jarvis, we can capture that data on our servers. We also know how many times each user asked Jarvis a question, the different ways people talk / type as well as what we responded successfully and not-so-successfully—all of this helps to improve the product every day.
C) Jarvis doesn't require any networking in order for it to work; all that's required is access to a power-point outlet and an internet connection. It also has amazing privacy features which address privacy concerns (you can turn of our ability to record you if you don't want us do so, or you can delete your recordings). We're also very open about what information we send off of your device – everything is logged with time-stamps and user data (including IP address) — meaning that users have full transparency into how Jarvis works and what's being sent off their phone. You'll never get an "I'm sorry" from Alexa if she fails at something — whereas Jarvis will tell you what went wrong and why. Our transparency can only help with gaining users confidence in using Jarvis!
D) Jarvis doesn't require a monthly fee in order to use it each month — unlike Alexa which charges $100/year for using the service.
E) Jarvis is open-source! This means that anyone from anywhere in the world can improve its functionality. The Alexa Skills Kit is proprietary — meaning that you have to pay Amazon in order to create a skill for their platform. We opened source our entire product on Github and invited the community to improve it!
If you want to see how Jarvis works then feel free to take the code and do whatever you like with it.
F) Jarvis is much more reliable than Alexa will ever be right now — there are so many things that can go wrong when using a voice interface for the first time, so we have tons of feedback on what not to say to it! This means that if someone asks to switch off the lights and Alexa messes up, you can trigger the light switch manually. If Jarvis messes up then it will have a better chance of figuring out what went wrong and correcting instead of just failing — this is especially important for voice interfaces because we're still in the early stages where mistakes will be made, but we want to give users the ability to communicate with Jarvis on its own terms which is partially enabled by deep learning (self-learning).
G) Jarvis is more than just an API — it's a full-service solution which can be integrated with a lot of different platforms. Alexa has only recently added support for IoT platforms like the Raspberry Pi, and that's still limited to very few actions — whereas Jarvis can do it all: it can control your lights, your home-automation system, and even connect with other software platforms. Jarvis is also a fully functioning virtual assistant which can do things like schedule meetings, help you buy stuff from Amazon (as in the comic above), set alarms and reminders — but we've built these features to be modular so they can be added to the platform for others to use.
H) Jarvis can talk back! He's not just some "wake word" which you have to shout in order for it to do something — instead he's a fully fledged character who has its own personality and moods. For example: if you're angry at him then he'll take that into consideration and respond in kind, whereas if you ask him to do something and he's busy then he'll tell you so instead of just ignoring you. These are all skills which Alexa doesn't have right now, but we've built the ability to do it all into Jarvis!
