The presence of Artificial Intelligence (AI) avatars participating in meetings instead of real people may seem very futuristic. But for Sam Liang, CEO of AI-powered transcription software company Otter, that could happen by the end of the year.
According to Liang, Otter has already made advances in this area and a prototype avatar that could replace a person in a remote meeting in virtual rooms may be close to becoming a reality.
Otter is developing an AI capable of answering 90% of questions
Imagem: Shutterstock / ST.art e CanvaPro / wannawit
Otter’s idea is to train AI avatars with information from recorded meeting notes and the voice data of the specific people they are trying to replicate. In addition to being able to act and talk like these people, AI would, in theory, be able to participate in conversations and answer questions based on the worker’s perspectives.
According to Liang, who revealed that he has up to 10 meetings a day, Otter’s expectation is to achieve an AI work personality capable of answering 90% of the questions asked during meetings. The remaining 10% can be answered by the human worker, who can receive questions by email after the meeting.
“A prototype could work this year,” Liang said in an interview with Business Insider.
If the project is successful, replacing people with AI in meetings could represent a huge time saving and productivity gain. Another example given by the CEO is the possibility of using virtual replicas during employee vacation periods.
Limitations and difficulties
Otter launched an AI tool that can answer questions and provide information about recorded meetings (Image: reproduction / Otter)
There are, of course, limitations. The initial idea is for AI to participate in meetings where existing information is exchanged. This would be the case of meetings related to customer support, sales and team status updates.
Meetings that require brainstorming new ideas will probably need the presence of human employees from each sector involved. Furthermore, there are also technical and social obstacles on the way to developing this technology.
According to Liang, it is “very challenging”, for example, to achieve a model that knows when to speak or intervene, a technology that understands the tone to use or that identifies emotional signals to raise your voice or not.
“You need to have knowledge and emotional intelligence to participate productively,” Liang said of AI.
For the executive, “several stages” will be necessary until AI becomes fully operational with this objective. However, he also believes that the technology to create this type of avatar already exists, citing what platforms like Character.ai and MetaAI already do when replicating personalities.
Still aiming to help save time with meetings, Otter launched a new AI chatbot this month that can answer questions and provide information about recorded meetings.