Deepfake

Deepfake is a type of content that uses ‘deep learning’ tools to convincingly replace one person's likeness and/or voice with that of another in digital media. These tools are readily available online and becoming easier to use. The creative applications are endless, ranging from harmless to nefarious, and around 500,000 deepfakes (video or voice) were estimated to be shared on social media in 2023 – a staggering increase from the 2021 figure of 14,678.

The technology could simply let you insert your face into your favourite films, become a sporting hero or live out dreams of being a rockstar. Or, you could use someone else’s face – and the ethics start to get murkier.

When a parody TikTok account seemed to feature actor Robert Pattinson dancing and performing magic tricks, even friends of his thought it was really him. While these videos weren’t meant to cause any harm, they were made without Pattinson’s knowledge and he found the experience unsettling. Some AI-generated images have become unexpected viral hits: one of the Pope wearing a Balenciaga puffer jacket fooled many and brought significant public awareness to the practice last year. 

There’s potential for more serious misinformation and deception. When a fake photo of an explosion at the Pentagon rapidly spread across social media and investment websites, the S&P 500 briefly dropped 0.3%. Endorsements are no longer a sure bet either; ask the BBC and U.S. news anchors who purportedly promoted a nonexistent investment platform created by Elon Musk. 

Deepfake fraudsters can avoid detection by eagle-eyed members of the public through targeted schemes. Recently, in Hong Kong, a finance employee obeyed orders to transfer out US$25.6m after a video call with his CFO and several other staff members… who all turned out to be ‘clones’. And you thought mic blunders were bad enough on Zoom ($ZM).

Efforts are under way to make the digital landscape more secure, from automated detection systems to stricter laws against the creation and dissemination of AI-generated content. Embedding watermarks is something companies like Adobe ($ADBE), Microsoft ($MSFT) and Google ($GOOG) are working on. Meanwhile, Meta ($META) has delayed the release of its Voicebox speech generator, due to potential misuse.

However, the field is moving so fast that detecting and reducing the spread of deepfakes can feel like a cat and mouse game. And as approximately a quarter of the world’s population heads to voting booths this year, the impact of ‘fake news’ could be severe and long lasting. 

Revealing the truth later doesn’t always undo the damage, either. Case in point: last week a good old-fashioned human error led to Lyft ($LYFT) reporting a profit margin at 10 times its real value, which sent the stock price soaring. The mistake was quickly corrected, but its market value did not completely retreat. If first impressions last, we all need to pay extra attention to what we believe in.