An engineer released an AI voice generator that can mimic the voices of famous people. Brandon Thomas developed Vocodes as a "pandemic side project." Users type what they want the AI to say, and the system speaks it out loud in the voices of celebrities like Mark Zuckerberg, Bill Gates, Betty White, and Christopher Lee.

More:

• Sam Altman and Sir David Attenborough are rated as the highest quality voices, with Barack Obama and Arnold Schwarzenegger listed as the worst.
• HT Tech tried out Vocodes and said it couldn't read everything. When it can't pronounce a word, the AI leaves it out.
• The Next Web writer Mix notes that it raises ethical and legal questions about how deepfakes could be used for nefarious purposes, like imitating celebrities without their permission.
• Thomas told Mix that he believes the government shouldn't be "overly protective" of machine learning applications, as they will play a large role in creating all virtual musicians and influencers. He acknowledged we might need legal protections against fraud or slander.

Podcast Notes: Lt. Gen. Jack Shanahan talks AI in the military.

Every Thursday, we summarize an AI podcast to save you time and keep you up-to-date. This week features Lt. Gen. Jack Shanahan, who recently retired as director of the Pentagon’s Joint Artificial Intelligence Center, or JAIC.

Jack speaks with Craig S. Smith of the Eye on A.I. podcast about the challenges of fostering innovation in the military and the future of AI projects [Note: Questions and answers were edited for brevity and clarity.]

Craig: What are the most important implementations of machine learning in the military over the next five years?

Jack doesn't know any aspect of the Department of Defense that would not benefit from artificial intelligence in some way. The question is, what can you do in what sequence to show return on investment?

The JAIC is interested in making the operations of all the different military services more efficient and effective. The approach is to start with the lower risk AI use cases first, like understanding data labeling and model training. This can be integrated into some weapons system and sustained over time.

A legal policy review and ethical principles are also needed. The JAIC worked on those for predictive maintenance on an H60 helicopter, as well as humanitarian assistance and disaster relief. Jack envisions using AI in mapping wildfire perimeters, post-hurricane building damage assessment, and road obstructions, which are important but lower risk than trying to put AI capabilities into an autonomous weapons system...

QUICK HITS

• Intel, along with MIT and the Georgia Institute of Technology, created an automated engine that can identify when two pieces of code do similar tasks.
• The U.K.'s information regulator said AI projects need to protect personal data.
• A Microsoft Hackathon led to an AI collaboration to remove plastic from the world's oceans and rivers.
• LinkedIn released its Qualified Applicant AI system that matches candidates to careers.
• Google debuted its Model Card Toolkit for better reporting of AI model transparency.
• DARPA hopes to expand on its AI projects and improve their reliability.
• WhizAI, an AI-based conversational platform for businesses, raised $4m in seed funding from Healthy ventures, Firebolt Ventures, and Bling Capital. 
• Experts such as Wikimedia Director of Machine Learning Chris Albon shared their top picks for best machine learning books.
• CEOs and CFOs talk the winners and losers of the office of the future.*

*This is a sponsored post