It’s not every day you get a call from the President of your country.
Or at least, that’s what somewhere between 5,000 and 25,000 residents of the great state of New Hampshire thought, prior to the January presidential primary.
The phone calls were clearly in President Biden’s voice.
He was urging voters not to vote in the primary, saying that “it’s important that you save your vote for the November election.”
Biden went further, saying...
“Voting this Tuesday only enables the Republicans in their quest to elect Donald Trump again. Your vote makes a difference in November, not this Tuesday.”
It all seemed so real…
Only it wasn’t.
It turns out that the calls were automated robocalls.
And the lifelike voice of Biden was created using artificial intelligence (AI) by synthesizing his real voice.
It didn’t sound artificial or synthetic either. It sounded like the real thing — quirks and all.
The media appeared coordinated, as usual...
“It’s obviously somebody who wants to hurt Joe Biden.”
And the widely used classic of the last few years...
“This is an attack on democracy.”
But something felt off to me.
After all, Biden’s name wasn’t on the ballot, a conscious decision made by the Biden campaign… which wanted his name removed from the ballot.
There wasn’t anything at stake for the Biden campaign.
Oddly, despite the remarkable capabilities of U.S. law enforcement, not much is known about the calls…
Other than that they were traced back to a tiny company called Life Corporation, which holds a small, shared office located in a strip mall in Arlington, Texas.
Was it someone behind the Dean Phillips campaign, who was running against Biden as a Democrat… and who wanted to suppress Democratic voters who might write in Biden’s name?
Or was it a government operation, designed to create a crisis… and prime voters for more of what’s to come?
Yet another hoax?
While it might be fun to speculate about the real motives behind these fake Biden phone calls…
This is just a taste of what’s to come.
The incredibly realistic voice calls weren’t created by an imitator who could speak like Biden, or by some bleeding edge technology in a government lab.
Whoever is behind this spoof used off-the-shelf artificial intelligence (AI) software from AI company ElevenLabs.
This is an interesting company with powerful technology capable of generating lifelike speech from text.
For those interested, you can go to the website landing page and type in some text, and then choose the name of the voice you’d like to use and click on the play button.
And the technology does even more than that.
ElevenLabs AI is capable of ingesting voice recordings of an actual person, processing that data, and then creating a lifelike synthesized voice.
This is exactly what “they” did to achieve their ends in the New Hampshire primary.
It wasn’t ElevenLabs’ fault that its technology was misused.
Once the company discovered the account linked to the people/person behind the robocalls, it suspended the account.
But it highlights how easy — and quick — artificial speech can be created for anyone. In just a matter of minutes.
If I had to guess, it probably took less than 30 minutes to create the voice for the robocall. And for someone with experience, I’m sure it could be done in less than 10.
Don't believe me? Here, please have a listen to this.
This is my own voice clone, which I created in a matter of seconds using a free voice-cloning service.
In the above clip, I just spoke into my computer microphone and read out a pre-determined clip provided by the software provider. But the software also allows us to upload a voice recording of our own.
A user can easily “steal” someone’s voice by uploading a YouTube video, or an audio recording of a speech. Someone could even upload an audio recording of a voicemail message. It really is that easy to make a voice clone now.
As you listen to it, you may pick up on the subtle unnaturalness in the above clip. The pauses between sentences will probably feel a little too short. But the voice absolutely sounds like me.
And remember... this is just a free tool and a cloned recording created in 3 minutes with no optimization at all.
The paid-for versions of this technology are far better, offering ways to easily correct for "tone color," emotion, accent, and more.
It can all be finessed, and not just in one language.
This month, ElevenLabs expanded its capabilities to support 29 languages.
Anyone can now generate text and have it converted into speech in one of the 29 languages shown above.
And if that weren’t cool enough, ElevenLabs can now transform speech in one of those languages into any one of the other languages.
Just like that. With a single click. Powerful.
Technology like this gives us both a sense of excitement, as well as trepidation.
It’s fun. It’s an incredible productivity boost…
And it will make a mess of everything, perpetrated by those who wish to spread lies and manipulate us.
When the bad actors combine this generative speech AI with generative video AI, we get the most lifelike deepfakes we’ve ever seen.
We’ve already seen this — fake videos of celebrities and politicians saying things that they never said — using earlier versions of this technology over the last couple of years.
Most of the time, it has been easy to spot the distortions.
But not this year.
Voice and video generative AI has gotten so good, almost everyone will be fooled.
The reality is that we won’t be able to trust much of anything on social media. If we care about a certain topic, it will be upon us to do our own research, cross check, and determine if something was fake or not.
Common sense and critical thinking have become our most valuable assets at a time like this.
Tech companies and research labs are already developing other forms of AI capable of spotting AI-generated content.
Whether certain technology and media companies use that technology or not, is a very different story.
New reader? Welcome to the Outer Limits! We encourage you to visit our FAQ, which you can access right here.
If you have any questions, comments, or feedback, we always welcome them. We read every email and address the most common threads in the Friday AMA. Please write to us here.
