Scientist: AI will allow humans to communicate with alien civilizations
Do aliens exist? Can humans communicate with them? Aliens have always existed in artistic creations, strange stories, or unofficial histories. For example, the well-known movie "Close Encounters of the Third Kind" shows the imagination of using music to communicate with aliens, and the mysterious U.S. military base "Area 51" is a must-know holy place for alien lovers.
However, with the rapid development of large language model (LLM) technology, in the future, humans will no longer have to rely on imagination. Is it possible to communicate with aliens through AI?
Experts from the Search for Extraterrestrial Intelligence (SETI) and the National Aeronautics and Space Administration (NASA) recently published an article proposing the idea of using artificial intelligence (AI) to promote dialogue between humans and extraterrestrial civilizations. The article, published on the Science Guy website, was co-authored by Franck Marchis, director of citizen science at SETI, and Ignacio Lopez-Francos, principal research engineer in NASA's Intelligent Systems Division.
Try sending a large language model into space
Scientists pointed out in the article that AI has been widely used on earth, from satellites to IBM Watson on the International Space Station, to future space probes, and AI technology is rapidly expanding into the field of space exploration.
So scientists thought that if a large language model (LLM) was sent to space, it might be an important step in breaking through the communication barriers across galaxies:
“We’re not just transmitting music, math, or a short self-description, we’re transmitting something meaningful: a large, carefully curated language model that encompasses the diverse nature of humans and the world we live in.”
In this way, alien civilizations can talk to us indirectly and receive answers on behalf of humans through LLM, without being limited by "communication delay" issues such as interstellar distance and human lifespan.
Scientists mention 2 open source language models
The scientists mentioned that some small open source language models, such as Meta’s Llama-3-70B and Mistral AI’s Mixtral 8x22B, will allow for training using curated data sets.
Although the size of Llama-3-70B (approximately 130 GB) is a bit large for interstellar communications, significantly compressing the size through a technology called quantization can maintain performance while also Works on its own without an internet connection.
But scientists also point out that this idea faces technical challenges. Current transmission technology can only perform interstellar communication at a speed of 100 bits per second, which means that if a moderately sized LLM is transmitted, it may take centuries to arrive. But through data reduction technology, interstellar communication time is expected to be shortened to about 20 years.
Interstellar communications may have potential risks
Scientists say communicating with aliens is a radical and potentially risky idea because unfriendly aliens could misuse the information humans provide, but it's still worth discussing in light of recent discoveries.
Several space telescopes have revealed that the Milky Way is teeming with exoplanets, with an estimated 300 million exoplanets that are similar to Earth and may harbor liquid water.
Scientists believe that some of these planets may be home to beings curious to meet humans and learn about our technological civilization.
The question is, which AI model can represent human nature?
So the question is, what kind of AI model can represent human nature? The scientist's idea is that if humans want to use a well-designed LLM to communicate with aliens, its functions can include generating text, images and sounds, etc., and the content, personality and tone should be determined by researchers, philosophers, historians and Other experts decided to represent the entire human race.
It can be said to be difficult for AI to imitate humans. Lisa Feldman-Barrett, a psychology professor at Northeastern University, pointed out in 2021 that even if AI technology has developed to the point where it can create components that are close to human complex systems, humans The joys, sorrows, joys and emotion-driven behaviors are still beyond the reach of machines.
"CryptoCity" asked ChatGPT, and it also stated that it cannot represent humans or express humanity because it has no self-awareness, emotions, or independent values, and still needs human supervision and guidance. However, if in the future it is really up to a few experts to decide on the AI model that represents humans, then it is still unknown whether it can be recognized by "human beings all over the earth".
Source: ChatGPT