Cisco AI Emerges
Artificial Intelligence (AI) is clearly one of the hottest topics these days, and I recently had the opportunity to see and hear about some of the things the Cisco Collaboration Group is doing in this area. At a recent Cisco “AI Meets Collaboration” event for press and analysts with Cisco’s Emerge Lab, the company discussed how it’s using AI to “improve the way our customer’s employees work together.”
As Rowan Trollope, SVP and GM of IoT and Applications at Cisco pointed out, “We’re mainly focused on helping teams get work done – we offer a wide range of technologies to facilitate this.” He then noted, “How do we help teams get work done better – it’s not about conferencing or messaging – it’s beyond that. Ambient computing and AI will provide radically better experiences.”
Trollope went on to describe the differences between AI in the home (think Alexa or Siri) and in the workplace. There are four key areas of difference, including identifying users, understanding context, the environment, and security. For example, Trollope noted that context for virtual assistants in the home is about functions such as ordering tickets, asking about the weather, or playing a song. Context is more difficult and more relevant for businesses, as each business has its own lingo, requiring a deep domain knowledge for each industry. When it comes to identity, it’s much easier for a smartphone or a device like Amazon Echo in the family room to identify the user, but it’s much more challenging in a meeting room where there are different users throughout the day.
Jonathan Rosenberg, CTO for Cisco's Collaboration Business, explained how Cisco is “creating AI for work, which is different than solving the problem for the home.” In this video interview, Rosenberg explains how Cisco is differentiating itself in AI by focusing on meetings and teams working together within the enterprise, as well as the key initiatives Cisco is working on.
Cisco’s new Spark Assistant was the day’s main attraction. Spark Assistant is based on the intelligent conversational interfaces developed by MindMeld, which Cisco acquired in May. The voice assistant is built on MindMeld’s natural language understanding platform, enabling users to talk to it in a natural way. Unlike consumer-based solutions like Alexa or Siri, Spark Assistant is designed for automating tasks at work to help meeting participants be more productive, and is designed for meetings and controlling Cisco room-based video collaboration devices such as a Spark-enabled videoconferencing system. Users can simply say “Hey Spark,” and then speak commands such as “Join my meeting,” “Mute the mic,” “Place a call to John,” and so on. It can automate device functions such as adjusting the volume or recording a meeting.
In this video interview, Tim Tuttle, MindMeld Founder and CEO (now CTO in Cisco’s Cognitive Collaboration Group), talks about Cisco Spark Assistant and how it helps with common meeting tasks based on simple voice commands (Note: sorry about the background noise during the video).
Spark Assistant will ship early next year and will work with Cisco room-based video conferencing devices, over time it will be available on phones and other types of devices. Spark Assistant will initially be available for the Cisco Spark Room Series, which includes the new Cisco Spark Room 70 system, an update to the MX800.
When asked about future plans for individual productivity rather than teams, Trollope explained that Spark Assistant will be available everywhere – in the Spark Client, on mobile devices, etc., but will have a different look and feel and a different set of purposes. For example, Spark Assistant on conference room devices looks like Apple TV, while as a desktop assistant it will more likely have a chatbot interface with long-running conversations. However, Trollope noted that there are currently so many pain points to solve when it comes to the meeting room, that Cisco wants to focus on this area for the near term. We all know how painful joining and participating in a video meeting can be, so eliminating the frustration and making this a better experience will be welcomed by most enterprises.In order to have a useful assistant for meetings, the assistant needs to know who’s in the room, who’s talking, and what’s going on in the room. This requires not only the conversational AI capabilities that MindMeld offers, but also intelligent hardware and an advanced compute platform that can do real-time facial recognition, speech recognition, and language understanding. Meeting rooms with many people participating require purpose-built hardware for image detection, noise suppression, and other capabilities. Cisco is working with Nvidia’s Jetson platform and investing in AI technology for capabilities such as intelligent proximity detection, as well as speaker tracking and facial recognition to recognize and determine who’s in the room and who’s speaking.
Cisco’s next-gen video system uses AI to identify distracting noises during meetings and videoconferences, such as dogs barking or keyboards typing, and can block those noises from the meeting. The system was trained on sample data for various dog breeds, different sirens from around the world, multiple keyboard brands, and other distracting noises.
Machine learning and AI are being used in Cisco video room devices to perform functions such as camera tracking, face recognition, and noise cancellation. For example, to create a better experience, the camera angle can be adapted and changed for when people enter and leave a room. During video meetings, the system learns and identifies who’s in the room, and the camera zooms in on the active speaker, or pans out when additional people enter the room, capturing the view of the room digitally. Cisco’s video room system can also recognize faces, insert “name tags” on the video to identify them.
As Snorre Kjesbu, general manager and vice president of Cisco Collaboration, notes in the following video, Cisco is using machine learning to make the meeting room experience more intuitive.
VR and More
In addition to Spark Assistant, we also saw demos of Cisco TeamTV, and Cisco Spark in Virtual Reality. TeamTV, described as an always-on distributed TV channel for groups, is made for “small, random, informal, ‘unmeeting’ encounters that turn teams into tribes.” It’s an always-listening system that lets participants informally interact on projects, or just chit chat (the AI equivalent of the water cooler or coffee maker?). The demo showed how the AI assistant in TeamTV is always listening so that when a team was discussing a certain topic, the AI assistant heard the key words in the discussion and pulled up relevant information about the topic, providing additional resources that may be useful to the team.
While I don’t think TeamTV will be a huge hit any time soon, Trollope eased my concerns about privacy by explaining that although the TV is always on and always listening, it’s easy to turn it on or off using handwaves so as to ensure privacy.
Virtual reality was also on display, as Cisco demonstrated Spark in VR, which is currently available in the Oculus store. Users can grab, open, share, and present files, draw on whiteboards, interact with 3D objects, and participate and collaborate in meetings without physically being in the meeting room. With the use of the AI assistant, the system was able to quickly learn about different topics and bring in relevant resources. While it’s certainly early days for VR for collaboration, this is clearly the direction we’re headed, and Cisco is forging ahead.
All in all, I was very impressed with what I saw and heard at from the Cisco Emerge Lab, and it’s clear that Cisco has a vision and is committed to investing in and forwarding the use of AI for enterprises. While not quite as sexy as consumer AI, using AI to improve the way business users meet and collaborate is certainly an admirable goal. Cisco is in a great position to be a leader in this area, as demonstrated by Spark Assistant, and we’re only at the beginning. I give the company credit for staking out its place in this nascent area, making some smart acquisitions, and showing its leadership and vision.