Fake BBM’s voice

“Far scarier, however, is the fact that most ordinary people largely can’t tell the difference between fake and real."
Fake BBM’s voice

In all likelihood, it had cost nothing to fake President Ferdinand Marcos Jr.’s voice in a viral YouTube video and audio clip that alarmed diplomatic and security officials last week.

In dissecting the fakery, a computer expert said Marcos’s faked voice — known technically as “voice cloning” — was likely done using an artificial intelligence (AI) system available for free.

In fact, several American companies are offering free AI-based voice cloning tools. One of the most widely known is called “Descript.” The company boasted its AI-powered system can clone any person’s voice.

Descript also offers other audio and video production services. The company’s website says users can avail of a free voice-cloning plan utilizing 1,000 words. For $24 (roughly P1,300), users can get unlimited voice generation.

It is probable that whoever faked Marcos’ voice used Descript’s free tool since the fake Marcos voice used only 26 Filipino words to seemingly urge military action against China.

The fake audio clip, since deleted after the Marcos government launched an investigation, circulated on YouTube early last month in the relatively obscure “Dapat Balita” YouTube channel.

Except for the obvious fact that the channel was set up recently, nothing much is known about “Dapat Balita.” Its YouTube channel, however, had nearly 127,000 subscribers and had aired 13 video and audio clips, all focusing on the country’s strained relations with China before it disappeared.

In flagging the fake Marcos audio, the Presidential Communications Office (PCO) said a “foreign actor” was likely behind the “deepfake.” “Voice cloning” is a subcategory of “deepfake.”

(A “deepfake” is essentially putting words into someone else’s mouth in a very believable way. Deepfakes also use deceptive images wherein a target is purportedly shown in compromising or scandalous situations. Deepfakes have also become more sophisticated by the day and are getting increasingly harder to spot. Besides Marcos’s faked voice, notable recent examples of deepfakes include nude images of Taylor Swift and a video of Ukrainian President Volodymyr Zelenskyy calling on his troops to lay down their arms.)

The PCO, however, didn’t categorically identify the “foreign actor” likely behind the deepfake. However, analysts suggested Chinese-influenced bad actors were the culprits.

This could be because the fake audio circulated as tensions further heightened between the Philippines and China in the West Philippine Sea.

Additionally, computer giant Microsoft recently reported that Chinese-influenced actors were experimenting “with new media” and were refining “AI-generated or AI-enhanced content” to spread conspiratorial narratives all over the world.

At any rate, no matter the “foreign actor” involved, the fear is that, as a foreign affairs department official said, the deceptive audio affected the Marcos government’s foreign policy on China.

The official didn’t elaborate, but it is possible the fake audio was a clickbait attempt to portray the Philippines as the aggressor in the WPS instead of China.

Still, whatever the intention, it doesn’t change the scary fact that the manipulation was easy and cheap to do.

Far scarier, however, is the fact that most ordinary people largely can’t tell the difference between fake and real.

In a series of man-on-the-street interviews, for instance, a TV news program played the fake video and audio clip to passersby and asked if the voice was that of Marcos. Most of those interviewed said they heard the President.

It was a reasonable reaction. Despite the crudely done audio clip exhibiting tell-tale clues, the voice was fake.

This is reasonable because, as one book-length study on the problem noted, “Seeing is believing: it is pretty convincing for us to believe in what we see and listen to from our ears. It is unlikely not to believe in the things you have observed yourself. It is now much easier to trick the brain’s visual system with misperception by utilizing these booming latest (AI) technologies.”

Multiply those individual human failings by hundreds of thousands and we have an enormous problem.

Related Stories

No stories found.
logo
Daily Tribune
tribune.net.ph