Ilya Gelfenbeyn, CEO of Api.ai, on AI and the IoT
Artificial Intelligence is a fascinating topic for many people nowadays, no matter if they are a consumer or an influencer. Today, I’m happy to be joined by Ilya Gelfenbeyn, CEO and co-founder of Api.ai, a conversational UX platform used to embed natural language understanding capabilities into connected devices, apps and services. Regular readers of SitePoint may recognize the service, as we have covered Api.ai in the past with a series earlier this year on getting started with the platform.
Ilya has a background in machine learning, natural language processing and conversational interfaces.
Elio Qoshi: Thanks for being with us today for this interview Ilya!
Ilya Gelfenbeyn: My pleasure.
Elio: We have covered Api.ai in the past, but could you briefly explain the concept behind it?
Ilya: Sure. Imagine having an actual conversation with a product, just as you would with a human. Api.ai is a conversational user experience platform for building natural language interactions for bots, applications, services and devices. The Api.ai platform lets developers seamlessly integrate conversational chatbots and intelligent voice command into their products and services. Developers can use Api.ai for speech recognition, context awareness, and conversational management to quickly and easily differentiate their business, increase satisfaction, and improve business processes.
Elio: What are some use specific cases Api.ai can be a great fit for?
Ilya: It fits a landscape of horizontal use cases. We are seeing our technology applied in innovative, creative ways for travel, customer service, e-commerce, the Internet of Things, gaming, automotive, finance, and more. Building more engaging and personal user experiences improves customer retention, increases revenue, reduces operation costs, and promotes productivity. Conversational user experiences let us feel more connected to products, companies, and devices in a more human way while allowing us to automate for efficiency.
Elio: What do developers need to know to be able to use Api.ai? Any prerequisites?
Ilya: Api.ai makes it easy for developers (and non-coders) to design and integrate intelligent and sophisticated conversational user interfaces into their products. Once you create your bot or agent, you can quickly and easily deploy it across various pre-integrated platforms, such as Facebook Messenger, Slack, Twilio, Cisco Tropo and Spark, Skype, Kik, Telegram and more.
You can leverage several pre-built knowledge packages created for a variety of popular topics based on over two and a half billion user queries processed by the system. When enabled, your agent can understand thousands of diverse requests out of the box – no coding required. Additionally, Api.ai has a robust library of SDKs and integrations with several popular platforms and technologies.
Elio: What’s Api.ai’s origin story?
Ilya: Speaktoit was co-founded by Artem Goncharuk, Pavel Sirotin, and myself; the team specializes in human-computer interaction technology based on natural language conversations and deep neural learning. In 2011 we launched Assistant, an intelligent personal assistant app six months before Siri was released. Assistant is one of the highest rated Android apps with over 30 million subscribers.
In 2014, we decided to open the platform up to developers as Api.ai to leverage the knowledge, data, and technology from the Assistant.
Elio: How does Api.ai compare to other bot development services out there?
Ilya: There are two types of bot development services. One is just for building integration kits, of which we are partners. The other is Natural Language Understanding (NLU) based. For NLU, Api.ai has the most sophisticated tools for designing conversational experiences.
Elio: How well does Api.ai integrate with products like Slack and the Amazon Echo?
Ilya: You can easily port your Agents across Api.ai supported platforms such as Twilio Programmable SMS, Slack, Facebook Messenger, Telegram, Kik, and others. With just one click, you can instantly expose your Agent to a channel of new users. We add new integration channels regularly.
Elio: What new features can we expect in the coming months or into 2017?
Ilya: We are constantly working on the developer experience and making it fast and easy to create and deploy advanced conversational scenarios. We are also always working to improve the AI that powers the Natural Language Understanding engine.
Elio: That sounds rather promising! Let’s talk a bit more about the Internet of Things. What benefits does Api.ai believe that AI will bring into the IoT space?
Ilya: AI in general will help devices to better understand people and predict their needs, intents and actions. Conversational AI will bring a universal interface for communicating with (and across) devices. There will no longer be a need to learn how to work with devices; people can simply express what they want to happen in natural language. Devices will share context and user preferences so that there is an immediate baseline of understanding.
Elio: What prerequisites or knowledge does one need to have to start setting up smart devices with your platform?
Ilya: After creating an agent in Api.ai, you’ll need some programming skills to connect it to a device. We offer a variety of SDKs for the most popular platforms to simplify this as much as possible.
Elio: How does Api.ai support integrations with Smart Home devices like Amazon Echo, Siri and Google Home?
Ilya: Google Home has not launched yet, and Siri is closed for the most part. For Amazon Echo, there is an agent-export functionality in Api.ai.
Elio: If people enjoy your service, but want to switch to another service (or use multiple services) at some point, how would the transition work out?
Ilya: People can export the intent definitions they created in Api.ai, but it would be hard to achieve the same level of natural language understanding quality. So, data is not limited to just our platform. Developers can leverage data to other services if that’s what they want to do.
Elio: The IoT and cloud-based services are heavily reliant on a strong web connection; what measures does Api.ai take for its users in case servers are down?
Ilya: Api.ai is hosted with AWS and other reliable (five nines) cloud service providers. We also provide an offline or hybrid implementation. Basically, you can export the agent and run it on the device locally. This functionality is offered for large-scale customers with special needs (like privacy or connectivity concerns). You can also make it work in a hybrid mode where it would run in connected mode when possible, but will automatically switch to local mode if needed. An example of this would be a connected car where internet service might not be available in all geographical locations.
Elio: IoT and robotics could lead to a whole new level of home automation. Are there many use cases of Api.ai being used for robotics in the home? Is it ready for wider adoption and how do you see Api.ai shaping that path?
Ilya: We have a handful of real world use cases where our natural language processing platform is being used in robotics for the home. However, we cannot cite these customers just yet. We are seeing an increasing number of projects created by developers with this use case, though. The more developers we have in this field, the better Api.ai understands users in this area. Api.ai is the only real development platform for such use cases. We are looking forward to being the driver of this movement.
Elio: I am sure we can expect many innovative things in the near future! Thanks so much for answering our questions Ilya, I am sure we will be hearing a lot about Api.ai in the IoT scene!
For those interested in trying out Api.ai or investigating its capabilities further, have a read of Patrick Catanzariti’s series on building an AI assistant with Api.ai.