AI Predicts the Last Selfies Taken on Earth Before the Planet’s Collapse
The singular relationship between humanity and artificial intelligence
--
Humanity’s relationship with artificial intelligence is singular. Our deepest fear is that a higher form of intelligence could end our species. But at the same time, it is our deepest hope for accelerating progress and solving science’s biggest challenges.
Decades ago, we asked artificial intelligence to predict the end of our civilization. Today, we ask it to imagine and depict the end of our species.
Back in 1972, for the first time in history, a team of MIT researchers built a machine learning model to anticipate the collapse of modern civilization.
The program predicted the collapse of modern society by 2040, charting a trajectory based on population, pollution, and natural resource use.
From another perspective, artificial intelligence has been our first fear as a threat to our species. Stephen Hawking first, Elon Musk later, as well as numerous experts, have warned us of the risks that an intelligence superior to our own could pose.
All of these voices align with the perspective of Nick Bostrom, author of the bestseller Superintelligence, which has received endorsements from Elon Musk and other prominent people in Tech.
Bostrom has repeatedly argued the risks of extinction are little known and grossly underestimated. Artificial intelligence poses a threat worse than any precedent technology –if its development does not have proper precautions.
The existential risk posed by a general form of artificial intelligence is an irreparable catastrophe –at worst, extinction.
Bostrom’s fear, as well as those who share his views, is evolutionary. The most intelligent species rules the world; with an explosion of Intelligence, AI would gain the ability to improve itself in a short time, surpassing human intelligence potential by many orders of magnitude.
And should AI achieve “superintelligence,” surpassing humans in intelligence, then it could become impossible to control.
As unlikely as it sounds, humanity may one day be wiped out, down to the last individual. And in general, anthropogenic extinction risks are more significant than natural ones.
A growing body of scientific literature asserts most existential risks originate from human activity. And more and more studies argue the threats behind artificial intelligence and biotechnology.
In the past two centuries, humans have introduced more sources with a high probability of extinction than in 200,000 thousand years of Homo Sapiens survival. A nuclear conflict is an obvious example of this risk.
But even more worrying for Bostrom is the set of existential risks we will introduce through technological innovations during the current century.
Faced with this prospect, it is curious when someone asks the artificial intelligence program DALL-E to imagine the extinction of the human race.
DALL-E is the neural network of the artificial intelligence lab OpenAI that creates images from textual input. A team of seven researchers spent two years developing the program, and in recent weeks the company began giving access to the beta version.
OpenAI plans to offer the program to professional artists, content creators, journalists, and publishers to create digital images and provide creative insights.
Researchers have trained DALL-E to create digital artwork in any style. And the latest version can produce high-resolution images that can be mistaken for photos.
When someone provides DALL-E with a description for an image, a neural network determines a set of core features, and another neural network generates the pixels needed to make those features. Once the script is typed in, the system creates ten distinct images.
DALL-E can also start with an initial image and make changes on demand. A user can provide an image, select an area, and tell the AI how to edit it.
After having a waiting list of over a million people for weeks, in late July, OpenAI began giving access to DALL-E to the first user groups, and the web was flooded with controversial images: cheeseburger lamps, peanut butter sandwich Rubik’s cube, greener cities, and tennis heads.
One user asked to generate a plate of alien seafood from another planet; another one, a dinosaur in a tuxedo looking in the mirror; and yet another one requested DALL-E to imagine the last selfie taken on Earth.
In images that first appeared on TikTok, DALL-E depicted men with grotesque, zombie-looking faces, with backgrounds of destruction, cloudy skies, and explosions behind them.
