Microsoft beefs-up its AI assistant with voice, vision

Microsoft on Tuesday doubled down on deploying artificial intelligence to consumers, releasing an updated version of its Copilot chatbot that can hold voice conversations and interpret images.

“Copilot will be there for you, in your corner, by your side and always strongly aligned with your interests,” said Mustafa Suleyman, executive vice president and CEO of Microsoft AI in a blog post introducing the upgraded chatbot.

Suleyman was a former leader at AI pioneer Google DeepMind—and was poached by Microsoft in March from his startup to head up its consumer AI division.

Microsoft was an early mover in releasing generative AI technology to everyday users, largely using the models from its $13 billion partnership with OpenAI, the maker of ChatGPT.

It has since been caught by Google and soon Apple, which are using their dominance of the global smartphone market to get their own AI tools to consumers.

Meta meanwhile is riding on the popularity of Instagram, Facebook and WhatsApp to promote its Meta AI chatbot to the company’s billions of users.

Microsoft said Copilot’s voice feature allows a more intuitive way to interact with your AI companion, offering four voice options for brainstorming, quick queries, or even emotional companionship.

The voice features will mainly be available in English-speaking countries while the vision tool is in a test phase and will require users to “opt-in.”

Microsoft stressed that safety will be paramount, and that data garnered from its new vision feature would be discarded after use.

According to the company, the new power will “see” what the user sees on a web page and “answer questions about its content, suggest next steps and help you without disrupting your workflow.”

The feature would start “with a limited list of popular websites to help ensure it’s a safe experience for everyone,” the company said.

Microsoft also said it was releasing a new feature called “Think Deeper,” that would give Copilot the ability to reason out more complex problems.

The addition comes after OpenAI released its o1 model last month to solve trickier problems in science, coding and mathematics—something that earlier models have been criticized for failing to provide consistently.

Microsoft said it was also going to make Copilot more personalized through a Discover feature that will make the tool “further personalized over time according to your conversation history.”

However, that upgrade would not yet be available in the EU and Britain, which have tighter restrictions on the use of personal data by large platforms.

© 2024 AFP