#37: Brought back to life, as a deepfake

1-2-3 Tech Miscellany

Hi, I’m Scott Bryan and you’ve (most likely) signed up for 1-2-3 Tech Miscellany, a weekly newsletter about ethics in tech.

Each week, I’ll introduce you to 1 ethical dilemma, 2 good news stories and 3 random quotes, photos or other oddities.

If you find yourself here via some other means, feel free to subscribe below.

Of course, if you no longer wish to receive this weekly email, you’re welcome to unsubscribe using the link at the bottom of the page.


Reading time: 3-4 minutes

Part 1 - Ethical Dilemma

Should deepfake technology be regulated?

“I believe in death, but in the death of Dali, absolutely not.” —  Salvador Dali

Dali wasn’t known for his modesty. Mind you, he wasn’t renowned for his prescience either, but his prophecy came to pass.

He’s been brought back to life — as a deepfake.

To reproduce Dali’s likeness, the Dali Museum in Florida fed an AI algorithm one thousand hours of archival footage of the painter. From these videos, the AI learned every facial detail, making Dali, well, Dali.

Watch this video. The experience is, ahem, surreal.

Dear Prudence

Deepfake tech, the machine-learning-powered system for swapping faces and doctoring videos, is usually associated with fake celebrity porn, or the potential to make world leaders say volatile things. Since the technology is readily available for anyone to use, many wonder why deepfakes haven’t taken off as a propaganda tool. Some argue it’s not as dangerous as we’re led to believe. Others take comfort in the knowledge, even if it does, deepfake detectors will catch them.

Assuming this is all true (it's not), imagine the exciting potential this technology represents. We could resurrect other historical figures. We might even bring back Grandma! Deepfake tech has even shown potential to help the chronically ill.

These are the positive angles Tenecent are presenting to the world. Tenecent are a Chinese tech giant and world leader in developing deepfake technology. 

They fear regulation could smother their lofty ambitions, so they just released a white paper insisting deepfakes are, overall, a good thing. 

In it, they argue, “It's not just about ‘faking’ and ‘deceiving,’ but [it’s] a highly creative and groundbreaking technology.” Tencent urge regulators to “be prudent” and to recognise its potential societal benefits.

They outline some of them. Here are three:

  1. Better shopping

    • Generate a precise model of yourself online and use it to try on clothes for size

  2. Better movies

    • Overdub and perfectly lip-sync, the original actor's voice into multiple languages.

  3. Better entertainment

    • Face-swap your head onto your favourite character in a video game or movie.

These are fun features no doubt. But in the midst of a climate crisis, are these the societal benefits industry should be focusing their energies on?

For good or ill, this stuff is coming. The question remains, who’s steering it?

Regulators, Mount Up!

Governments have, to date, been hands off with tech companies. They’ve said, “Regulate yourselves. Write it all down. Tell us what you come up with.”

We only figured out cars needed seatbelts after we started driving them. And that’s what seems to be happening here too.

This approach simply won’t fly anymore.

Furthermore, technology moves too fast for policy to keep up. 

This was fine when Alexander Graham Bell invented the telephone. It took decades to codify communication laws, but it also took an age before phones were ubiquitous too.

Today, by comparison, when someone invents a new technology, it’s everywhere in jig time.

Think back to the early days of drones. Tech folk argued regulating them would kill a nascent industry, but before you knew it, kids were getting them for Christmas. By that point, it was too late. Drones were everywhere, causing problems of their own.

Ring Of Fire

I would love to see deep fake tech technology give chronically ill patients peace and dignity. I am keen to visit the Dali Museum too.

But no matter how many amazing things deepfake technology does for us, or how accurate anti-deepfake technology evolves to combat the dangerous stuff, damage will still be caused by humans.

That’s why such technology needs oversight.

We need to get smart and agile about technological developments. We can’t let powerful companies run rings around everyone, or burn new paths far ahead of everybody else.

Tenecent might insist deepfake tech “will not topple society’s truths, much less pose a threat to the world order,” but that’s easy for them to say!

Maybe it will, maybe it won’t.

Either way, we must be conscious of deepfake’s negative impacts. At the same time, we should “look to the better angels of AI to harness deepfake’s potential to create new forms of communication” — ones that can genuinely improve life on planet earth for everyone.

And that involves regulation.

Share


Part 2 - Good News Stories

The AutoDerm 3000

A study by the Medical School Vienna, has found AI is better at detecting skin cancer than dermatologists.

This doesn’t mean the algorithm replaces doctors, it just helps them do a better job. Computers aren’t even close to being able to carry out many of the functions of a skin specialist — for now. 

Read more on Wired.

Recreate Your Voice

Motor Neurone Disease, is a dreadful progressive neurodegenerative disease. It can take away a person's ability to speak.

Using the same technology to create deepfake audio, it is now possible to build a complete digital voice clone. This will allow MNS suffers to continue communicating in their own voice.

Read more on ProjectVoice.

Share


Part 3 - Quotes, Tweets And Other Oddities

“Think about Facebook as a seller of meat products.

Most of the meat is produced by others, and some of the cuts are delicious and uncontaminated. But tainted meat — say, Trump steaks — also gets out the door in ever increasing amounts and without regulatory oversight.

The argument from the head butcher (Zuckerberg) is this: People should be free to eat rotten hamburger, even if it wreaks havoc on their gastrointestinal tract, and the seller of the meat should not be the one to tell them which meat is good and which is bad (even though the butcher can tell in most cases).”

Kara Swisher

  1. Vicariously.io “offers the easiest way to create Twitter lists based on the follows of other users."

  1. Who Knew?

Source: Michael Zelenko


Thanks so much for reading.

Feel free to use the link below, to share this newsletter.

Have a great week.

Scott