You can no longer trust videos – 02/18/2024 – Ronaldo Lemos

Last week, the world watched in bewilderment as videos created by the Sora platform, the latest project from artificial intelligence company OpenAI. The images show a woman walking in Tokyo, dogs playing in the snow and a house on a cliff facing the sea. All were synthesized by AI from a mere paragraph of text with instructions. They are all convincing as if they were reality itself (at least to those who look inattentively).

Videos are for us humans the most reliable proof that something happened. In recent years we have learned that texts and images are easily manipulated. But what about the video? For most people this is what shows that something actually happened. All of this now falls apart. You can no longer trust any video.

Imagine someone using AI with the following instruction: “Create a video to bypass a bank’s facial recognition. Turn your head left and right, look up, the way the bank requests it for authentication.” Several banks, including in Brazil, use facial recognition as a security mechanism. With artificial intelligence tools it could be easy to fool everyone. Being sure who is on the other side of the screen will become more expensive and difficult.

The problem is far from just that. With the launch of Sora, OpenAI wants to send a message: the company is the future of audiovisual creation.

Today the games, film and series industry has a monopoly on creating the most attractive and valuable images on the planet. This is done with expensive and difficult-to-realize productions, filmed based on reality or arduously programmed. OpenAI wants to show that in the future (perhaps near) it will be able to do the same with synthetic images, generated by AI.

More than that, it will deliver this Promethean fire to anyone, who will be able to conjure AI to create everything from unprecedented monumental works to pornography, fake news and interference in electoral campaigns.

The curious thing is that the company no longer reveals how Sora (which means “sky” in Japanese) was trained, due to concerns about copyright issues. This would have been done with all types of internet videos, films, series and synthetic data generated by video games. When I met with Sam Altman in Brazil last semester, he emphasized the importance of synthetic data for training today’s AIs.

Sora is the newest “slide” that Altman shows to the world as part of his plan to raise US$7 trillion (R$34.79 trillion). If successful, it will be the largest capital raising in history. The objective would be to achieve “general artificial intelligence”, dominating the chip market.

The thesis is that, with the arrival of artificial intelligence, many economic sectors become “legacies”, relics of the past. This includes everything from Hollywood to Netflix, the financial market, most of the areas of health, engineering, medicine, law, accounting and management, not to mention the military and military equipment sectors. What would make the $7 trillion cheap close to the prize sought

There is also the thesis that this is all a trick, and we are being collectively fooled, among other things, by videos of women walking in Tokyo and dogs playing in the snow.

Gone – Bitcoin

It is already – Bitcoin

Coming – Bitcoin?

LINK PRESENT: Did you like this text? Subscribers can access five free accesses from any link per day. Just click the blue F below.