Cisco has announced a new artifical intelligence strategy for its Webex videoconferencing platform, including an AI-powered assistant that can summarize meeting content and answer questions, and generative AI capabilities to improve the quality of video meetings.

The announcements, which were made Wednesday at the company's WebexOne conference, focus on three areas: audio intelligence, video intelligence and language intelligence, according to Keith Griffin, a Cisco distinguished engineer. "Our intent here is to is to apply this across the entire Webex suite, from our devices, to our contact center and the Webex connect control hub, really everywhere you would need an assisted experience," he said.

Users will be able to prompt the Webex AI Assistant with natural language questions and have the tool generate answers in real time. The new tool can also recommend changes to tone, format, phrasing in Webex Messaging and Slido, and suggest responses that are informed by the context of the conversation.

AI-generated meeting summaries are designed to  help users catch up on missed portions of events or entire meetings by providing easily digestible summaries of what was missed. Meeting summaries will be organized into chapters and highlights to allow users to skip through to find topics most relevant to them. Webex AI Assistant can also recap and summarize messages, and generate topic summaries in Slido.

These new capabilities are in various stages of availability, with shipping planned to begin before the end of 2023.

Improving audio and visual quality in video meetings

Cisco is  also helping users tackle network issues that still frequently plague video meetings through the introduction of a new AI Codec, a generative AI solution that the company  claims will solve poor audio quality. A codec is a device or computer program that encodes or decodes a data stream or signal.

The AI Codec allows for massive transmission redundancy to recover from network packet loss. Webex and has built-in speech enhancement functions like noise removal, de-reverberation and bandwidth extension, to deliver what the company says is unprecedented audio clarity.

Cisco said it will also apply machine learning techniques to improve video quality using super resolution, an industry-standard technique for delivering high-definition meetings with high  video quality regardless of bandwidth conditions.

"The codec is fully optimized for speech and requires just a fraction of the bandwidth of other codecs," said Griffin. "This means that even when bandwidth is extremely limited, you get the highest possible speech quality output."

Webex's AI Codec uses up to 16 times less bandwidth than an industry standard audio codec, he said.

Finally, the introduction of Cisco-developed Real-Time Media Models (RMMs) will further enhance the audio and video quality users experience on calls. The models have the ability to take multiple media streams and produce multiple outputs, such as people and object recognition and action analytics like movement and gestures, said Griffin, explaining how RMMs work.

These RMMs will also allow audio and video channels to be used as signals of context in traditional text-based capabilities, like meeting summaries and highlights.

"The whole world has been talking about large language models and generative AI but in the collaboration business, there's more [to these capabilities] than text," said Griffin, adding that while text is one component of Webex's offering, its also focused on speech, video, and other audio and computer vision that inputs.

"That's a pretty exciting place for us to be and we're continuing to seek out new problems and new solutions in those areas," he said.