The terminology NASA once used to refer to artificial general intelligence has changed, the space agency said in response to FedScoop's questions about emails obtained through a public records request, suggesting the science-focused federal agency may be discussing the emerging technology in the age of generative AI.
While building artificial general intelligence (AI that is theoretically as powerful as humans) remains a distant goal, it remains an important one for companies like OpenAI and Meta. It is also a topic of intense debate and discussion among technology researchers and civil society, with some believing it could distract attention from more pressing AI risks, such as bias, privacy, and cybersecurity.
NASA is one of the few government agencies to have expressed special interest in the AGI problem, while many federal agencies remain focused on more direct applications of AI, such as using machine learning to process documents. In a statement to FedScoop, Jennifer Doelen, deputy news officer at NASA, said the agency is “committed to formalizing protocols and processes for our use of AI and expanding our efforts to foster AI innovation across the agency.”
NASA's framework for the ethical use of artificial intelligence, released in April 2021, refers to both artificial general intelligence and artificial superintelligence. In response to FedScoop's questions about the status of this work, NASA said the “terminology of AI” has changed, pointing to its treatment of generative artificial intelligence, including the large-scale language models that power systems like ChatGPT. (Whether systems like ChatGPT will ultimately be the basis for AGI remains a topic of debate among researchers.)
“NASA looks at artificial intelligence as a whole, not just subparts,” Doelen says. “Terminology from past frameworks has evolved. For example, the terms AGI and ASI have evolved into what we think of today as generative AI (genAI).”
NASA also highlighted a new working group focused on ethical artificial intelligence and the agency's efforts to achieve the goals outlined in President Joe Biden's executive order on AI last October. NASA also hosted a public town hall meeting on AI capabilities last month.
However, the apparent retreat from the term artificial general intelligence is notable given some of the future concerns outlined in the 2021 Framework. For example, one of the goals outlined in the document was to “lay the foundations for a successful, peaceful, and potentially symbiotic coexistence between humans and machines.” The Framework notes that while AGI has not yet been achieved, there is growing confidence that there may be a “tipping point” in AI capabilities that will fundamentally change how humans interact with technology.
Experts often divide artificial intelligence into a few categories: artificial narrow intelligence, which is an AI use case designed with a specific application in mind, and artificial general intelligence, which refers to an AI system that matches the capabilities of a human user. The NASA framework also talks about artificial superintelligence, which describes AI capabilities that “exceed” human capabilities.
The document stipulates that NASA should “early adopt” domestic and global best practices for these advanced technologies. While noting that many AI systems will not reach the level of AGI or ASI, it urges NASA to consider the potential impacts of these technologies. Many of the considerations outlined in the report are still far from reality, but range from analyzing the possibility of encoding morality into advanced AI systems (a potentially impossible task) to “merging” astronauts with artificial intelligence.
“Creating a perfect moral code that applies in all cases remains a challenging task that NASA experts will need to work on in collaboration with other national and international experts,” the document said. “As humanity moves toward long-term spaceflight, technology may need to consider the benefits and implications of merging humans with AI machines, most notably adaptations that enable survival during long-term spaceflight but the challenges of returning to Earth.”
As part of this framework, NASA's interest in studying the effects of AGI was also discussed in emails obtained by FedScoop earlier this year.
