Given the hoopla around generative AI, you might think most people are using chatbots every day to get answers, write emails and dream up new ideas that empower humanity. But it turns out a majority of folks haven't even used the most popular chatbot, according to the Pew Research Center, which said earlier this year that only about a fourth of Americans had even tried OpenAI's ChatGPT.
That's why AI makers Meta, OpenAI and ElevenLabs all announced new voice additions to their tools last week. They're betting that you're going to be more interested in their chatbots if they talk back to you using a famous, notable, or distinctive voice like Apple's Siri or Amazon's Alexa.
Meta announced that it's licensed voices for its Meta AI chatbot from actors Awkwafina, Dame Judi Dench, John Cena, Keegan Michael Key and Kristen Bell to answer questions, "explain something you're curious about" or tell you a joke. "Voice is going to be a way more natural way of interacting with AI than text," Meta CEO Mark Zuckerberg said.
ElevenLabs, a tool for turning text into audio (including cloning your own voice based on a small sample), added New Age author Deepak Chopra's voice (listen here) to the list of iconic voices in its free Reader App. Chopra joins a list of "legendary stars" added in July, including Judy Garland, James Dean, Burt Reynolds and Sir Laurence Olivier, whose voices are used to read aloud books, articles, PDFs and other text. The app offers hundreds of voices, in 32 languages.
And after hitting pause on its AI Voices in May, when actor Scarlett Johansson accused the company of copying her distinctive tones for one of its audio personalities, OpenAI announced an update that adds five new voices (for a total of nine) with improved accents. They're part of the Advanced Voice Mode feature available to ChatGPT Plus and Team subscribers (plans start at $20 a month).
"It can also say 'Sorry I'm late' in over 50 languages," OpenAI said in an X post announcing the additional voices.
Given the controversy over Voice Mode, maybe just how to say "Sorry" in 50 languages would've been enough.
Here are the other doings in AI worth your attention.
AI can now beat 'I'm not a robot' captcha tests
Online services using captcha puzzles — you know: Pick out all the photos with bicycles to show us that you're a human — will now have to rethink how they'll go about verifying that whoever's asking for access isn't actually a bot.
That's because researchers in Switzerland were able to train an AI model to "solve Google's reCAPTCHA v2 image challenge," ZDNet editor Radhika Rakjumar reported. "The researchers trained the model — named YOLO for 'You Only Look Once' — on images of the usual reCAPTCHA fodder, meaning mostly road vehicles, traffic lights, and other related environmental objects."
Their Sept. 13 report, posted on Cornell University's Computer Science site, is called, simply, Breaking reCAPTCHAv2. Though no AI (or human) is perfect, the AI model was able to trick the image identification quiz often.
Will captcha tests go away? Not necessarily. But they may get harder, which Rakjumar notes wouldn't be great for "existing accessibility concerns for the visually impaired." Google may be experimenting with other ways to stop spam, along with content scrapers and other malicious actors using AI bots — and those methods could include fingerprint or retinal scans to prove you aren't a robot.
OpenAI, Jony Ive working on an AI hardware something-or-other
OpenAI came up in the news a lot last week, proving that sometimes things happen in threes.
First, CEO Sam Altman confirmed in an interview with Axios that the AI hardware project he's working on with Jony Ive, Apple's famous former design chief, isn't a phone. That came after Ive confirmed in a Sept. 21 New York Times interview that he was working with Altman and OpenAI.
So what's he working on? A wearable? (Altman was an investor in the failed Humane AI pin.) Glasses? An AI umbrella? Altman didn't say, but he did note that whatever it is, "It's a long way away."
Second, Altman got some buzz for writing (maybe? Or maybe he used ChatGPT?) a blog post called The Intelligence Age. In it, the AI startup chief who's hoping to make billions by convincing everyone they should be using his chatbot, touted his optimistic view of AI. He predicted, "Our children will have virtual tutors who can provide personalized instruction in any subject, in any language, and at whatever pace they need" and, "AI models will soon serve as autonomous personal assistants who carry out specific tasks on our behalf like coordinating medical care."
He also thinks, "We will have a superintelligence in a few thousand days (!)" — or not. "It may take longer," he added. I guess ChatGPT couldn't forecast when we'll see an artificial general intelligence.
Altman also opined that, "We will be able to do things that would have seemed like magic to our grandparents." (Who wants to tell him that our grandparents are already amazed by computers, Venmo, and the ability to video chat with their grandchildren across the world on a smartphone?)
The third bit of news was a surprise to even Altman — that OpenAI CTO Mira Murati decided to resign. Murati, who joined OpenAI six and a half years ago, said in an X post that she wants "to create the time and space to do my own exploration."
Industry observers noted that Murati's exit comes after other top executives have resigned, with some citing safety concerns over the pace of OpenAI's development. Altman is also reportedly working with investors to figure out how to transform the nonprofit into a for-profit company and raise the capital he needs to grow and fend off rivals including Google, Meta and xAI.
Also worth knowing...
Microsoft's need for energy to power its computer servers — and its AI software — may be prompting it to buy the nuclear reactor at Three Mile Island that was the site of the worst nuclear accident in US history. The plant would be back in service by 2028.
Forget wedding vows. If you're looking for a practical use for ChatGPT, consider using the chatbot to negotiate monthly utility bills and random fees. CNET's Amanda Smith explains how.
Meta made several other AI-related announcements at its AR/VR developer conference, including an experiment with integrating custom AI content into your Instagram and Facebook feeds. Katelyn Chedraoui, who reviews AI image generators for CNET, calls the experiment "a terrible idea."
If you'd rather not wait however long for Jony Ive and OpenAI to give you your futuristic-gadget fix, check in with CNET's Scott Stein for his take on Meta's updated Ray-Ban glasses, which offer new AI features including live translation and camera recognition of QR codes.