Scientists Suggest Sending AI to Aliens So They Can Talk to It in Real Time
“Aliens could learn one of our languages, ask the LLM questions about us and receive replies that are representative of humanity.”
Speaker of the House
Artificial intelligence is being heralded as the future here on Earth, and according to a pair of scientists invested in the search for extraterrestrial intelligence (SETI), it might even allow aliens to talk to us — or a version of us, at least.
In an editorial for Scientific American, SETI Institute astronomer Franch Marchis and NASA researcher Ignacio G. Lopez-Francois have teamed up together as “alien-curious scientists” to advocate for an AI-infused version of what those in their respective fields refer to as “messaging extraterrestrial intelligence,” or METI for short.
Despite its heady implications, humans have been engaged in METI for as long as we’ve been going to space. Since the early 1960s, we Earthlings have sent everything from music and scientific formulas to Morse Code and maps out into the great beyond in hopes of hearing back — though thus far, those attempts at communicating with ETs have been for naught.
To Marchis and Lopez-Francois, these METI efforts can only be enhanced by introducing something as interactive as an AI large language model (LLM). In particular, they’re arguing that we should send one out into space so it can speak for us.
“Aliens could,” the duo wrote, “learn one of our languages, ask the LLM questions about us and receive replies that are representative of humanity.”
Hello Out There
With the growing body of research about the potentially hundreds of millions of habitable exoplanets in our Milky Way galaxy alone, these alien-oriented scientists believe “several of these worlds could host technological civilizations curious to meet us and learn about us.”
While one might argue that our current LLMs are not even up to snuff enough for humans to get much valuable communication out of them, these SETI scientists said that some open-source models, such as those made by Meta and Mistral, could already be fine-tuned enough to act as human emissaries.
After creating a human simulacrum fit for the stars, these alien-oriented LLMs could then be compressed, the pair explained, via the process of “quantization,” which maps huge sets of data onto smaller sets. They could then be sent out to space using several methods — including radio, laser, or even copper disk communiqués — that could make the transmission of any LLM as rapid as possible (though Marchis and Lopez-Francois admitted that there’s some massive technological hurdles that would have to be tackled first.)
“By sending well-curated large language models into the cosmos,” they concluded, “we will open the door to unprecedented exchanges with extraterrestrial intelligences, ensuring that our legacy endures, even when we might not.”
