Tag Archives: capabilities

How to use Audiobox, Meta AI’s new sound and voice cloning tool – Android Police

Posted on by

Meta introduced its generative AI model for speech, Voicebox, in mid-2023. Meta aims to take AI sound generation to the next level with Audiobox, Voicebox's successor. The innovative tool generates sound effects from text prompts, eliminates noise from speech recordings, creates a restyled voice, generates speech in the style of an audio clip, and more. Before we take it for a spin, let's learn more about Meta's Audiobox.

The Audiobox demo is available on the web only. Try it on your Mac, Windows desktop, or a top Chromebook.

Creating high-quality audio can be a challenging process. Not everyone is a sound engineer and has access to extensive tools to create audio. Here's where Meta's Audiobox comes into play. It's a sound-generation tool from Facebook AI Research (FAIR). Meta's latest offering generates audio and sound effects using voice inputs, text prompts, and a combination of both.

With Audiobox, Meta aims to lower the barrier of audio creation and make it easy for general users to create high-quality sound samples. Whether you want to create audio for a podcast, YouTube video, audiobook, or video game, Audiobox can be your helping hand to get the job done.

Generative AI has made audio creation and voice cloning popular. There is no shortage of such tools. Meta's Audiobox easily stands out from the crowd due to its unique capabilities. Here's what you can do with it:

All Audiobox features are available to try from the company's official website. You can generate audio samples, check previews, and download them to your device.

You can also move to the Sound Effects menu and describe the sound sample you want to create. Add enough details to get astute results from Audiobox. We ran several text prompts and were impressed with the generated sound effects.

Audiobox can produce sound samples that are close to how people speak naturally. It has led to concerns about AI-powered deepfakes. Especially since the US presidential elections are around the corner, you can't rule out misuse of such AI tools. Meta implements automatic audio watermarking on audio generated by Audiobox.

The embedded signal in the generated audio is negligible to the human ear but can be tracked to the frame level. Meta will also add a voice authentication to prevent impersonation. The person must speak a voice prompt while registering their voice. The text prompt refreshes every 50 seconds, so playing someone else's pre-recorded voice is difficult.

Meta decided against making the AI model open source to prevent potential misuse.

Meta has done a remarkable job with Audiobox. It's accurate and very good. Try it with different prompts and voice samples, and check the results. Besides Facebook, tech giants like Google and Microsoft are exploring generative artificial intelligence to create content.

The search giant recently launched Google Bard to take on Open AI's (and Microsoft) ChatGPT. Read our dedicated post to learn more about Google Bard. We also compared Google Bard with ChatGPT to find their capabilities, limitations, and potential.

See the original post:

How to use Audiobox, Meta AI's new sound and voice cloning tool - Android Police

Busted! Drive-Thru Run by "AI" Actually Operated by Humans in the Philippines

Posted on by

The AI, which takes orders from drive-thru customers at Checkers and Carl's Jr, relies on humans for most of its customer interactions.

Mechanical Turk

An AI drive-thru system used at the fast-food chains Checkers and Carl's Jr isn't the perfectly autonomous tech it's been made out to be. The reality, Bloomberg reports, is that the AI heavily relies on a backbone of outsourced laborers who regularly have to intervene so that it takes customers' orders correctly.

Presto Automation, the company that provides the drive-thru systems, admitted in recent filings with the US Securities and Exchange Commission that it employs "off-site agents" in countries like the Philippines who help its "Presto Voice" chatbots in over 70 percent of customer interactions.

That's a lot of intervening for something that claims to provide "automation," and is yet another example of tech companies exaggerating the capabilities of their AI systems to belie the technology's true human cost.

"There’s so much hype around AI that everyone is misunderstanding what this tool is," Shelly Palmer, who runs a tech consulting firm, told Bloomberg. "Everybody thinks that AI is some kind of magic."

Change of Tune

According to Bloomberg, the SEC informed Presto in July that it was being investigated for claims "regarding certain aspects of its AI technology."

Beyond that, no other details have been made public about the investigation. What we do know, though, is that the probe has coincided with some revealing changes in Presto's marketing.

In August, Presto's website claimed that its AI could take over 95 percent of drive-thru orders "without any human intervention" — clearly not true, given what we know now. In a show of transparency, that was changed in November to claim 95 percent "without any restaurant or staff intervention," which is technically true, yes, but still seems dishonest.

That shift is part of Presto's overall pivot to its new "humans in the loop" marketing shtick, which upholds its behind the scenes laborers as lightening the workload for the actual restaurant workers. The whole AI thing, it would seem, is just packing it comes in, and the mouthpiece that frustrated customers have to deal with.

"Our human agents enter, review, validate and correct orders," Presto CEO Xavier Casanova told investors during a recent earnings call, as quoted by Bloomberg. "Human agents will always play a role in ensuring order accuracy."

Know Its Limits

The huge hype around AI can obfuscate both its capabilities and the amount of labor behind it. Many tech firms probably don't want you to know that they rely on millions of poorly paid workers in the developing world so that their AI systems can properly function.

Even OpenAI's ChatGPT relies on an army of "grunts" who help the chatbot learn. But tell that to the starry-eyed investors who have collectively sunk over $90 billion into the industry this year without necessarily understanding what they're getting into.

"It highlights the importance of investors really understanding what an AI company can and cannot do," Brian Dobson, an analyst at Chardan Capital Marketts, told Bloomberg.

More on AI: Nicki Minaj Fans Are Using AI to Create "Gag City"

The post Busted! Drive-Thru Run by "AI" Actually Operated by Humans in the Philippines appeared first on Futurism.

Read the original post:
Busted! Drive-Thru Run by "AI" Actually Operated by Humans in the Philippines