- The Daily Update
- Posts
- Microsoft reportedly develops human-level voice tech
Microsoft reportedly develops human-level voice tech
VALL-E 2 will not be released to the public
Jack Shields
July 05, 2024
Happy Friday!
Hope you had a relaxing holiday. We’re capping off an unusually quiet week in AI with major news from Microsoft. Let’s get straight to it.
In today’s Daily Update:
🗞️ Microsoft develops human-level AI voice tech
🤖 Google blames AI for falling short of climate goals
📸 WhatsApp is working on personalized AI avatars
🚨 AI Roundup: Four quick hits
Read time: 2 minutes
TOP STORY
🗞️ Microsoft develops human-level AI voice tech
DALL-E 3
A Microsoft research team has introduced a new synthetic speech model called “VALL-E 2” that will not be released to the public at this time.
What you need to know:
Microsoft claims VALL-E 2 is the first voice model to achieve human-level performance in generating speech from text.
VALL-E 2 is reportedly able to clone voices with just a few seconds of a recording.
This model relies on neural codec language models to generate speech, an approach that boosts VALL-E 2’s understanding of language and ability to process text.
Microsoft says VALL-E 2 is for research purposes only, and has no plans to incorporate the model into any products.
Why it matters: If Microsoft’s claims are true, this is a major advancement in synthetic speech technology that is accompanied by a wide range of potential risks. Human-level speech generation could be an invaluable tool for direct-to-consumer businesses around the world, but an equally useful tool for malefactors and foreign adversaries.
AI INSIGHT
🤖 Google blames AI for falling short of climate goals
DALL-E 3
Earlier this week Google reported that it is falling significantly short of an ambitious plan to combat climate change due to its AI efforts.
Key points:
Three years ago, Google set a goal to go “net-zero” in climate-changing gas emissions by 2030.
Despite that, Google’s emissions increased by 13% in 2023. The company’s emissions are up 48% since 2019.
Google says the growth is due to the massive electricity demands for data centers involved in AI projects.
The International Energy Agency projects that the global data center and electricity demand could double by 2026.
Why it matters: The environmental costs of AI development are often underreported and downplayed by Big Tech. Google’s numbers here are concerning and likely mirror increasing emissions at other major companies throughout the industry.
BUSINESS SPOTLIGHT
📸 WhatsApp is working on personalized AI avatars
DALL-E 3
WhatsApp is developing a new generative AI feature that allows users to create personalized avatars of themselves.
The details:
To generate a personalized avatar, WhatsApp users will have to take photos of themselves once. These images will be used to train Meta AI to generate avatars in the user’s likeness.
The optional feature allows users to imagine themselves “in any setting from the forest to outer space.”
Meta appears to be taking a cautious approach with personalized avatars after experiencing issues with previous AI tools.
Why it matters: On the surface, this is a fun new feature for an instant messaging app. The underlying importance goes back to last week’s story about Synthesia’s upgraded AI avatars. Tech companies are pushing toward digital avatars that can take our place in a variety of online settings. Only time will tell if this is really possible.
MORE TRENDING NEWS
🚨 AI Roundup: Four quick hits
Columbia University researchers develop method that allows multimodal large language models (MLLMs) to reason with images.
Perplexity upgrades its AI search engine despite ongoing investigations into its data scraping practices.
Magic seeks funding at $1.5 billion valuation for its AI coding assistants.
Google will now generate disclosures for political ads that use AI.
THAT’S A WRAP