- Scientists at US Department of Energy explore concept of an “Exocortex”
- It is pitched as an extension of the scientist’s brain with a word-based HCI
- It is uncharted territory as one member of the team puts it
AI is already a big part of our lives, shaping healthcare, transportation, education, customer service, retail, and creative industries, to give just a few examples.
Researchers at the US Department of Energy’s Brookhaven National Laboratory are currently exploring the concept of a science “exocortex,” the integration of artificial intelligence with human cognitive capabilities, viewed as an extension of the scientist’s brain.
The idea is to increase the cognitive abilities of researchers by providing a bridge between the human mind and a network of AI agents.
Kevin Yager, Electronic Nanomaterials Group leader at the Center for Functional Nanomaterials (CFN), detailed the idea in Digital Discovery. The proposed exocortex is certainly ambitious in its scope, aiming to streamline specific research tasks through a swarm of inter-communicating agents.
Each AI agent would specialize in particular tasks – be it handling scientific literature, orchestrating experiments, or data synthesis – and their collective behavior could offer a synergy that significantly extends the cognitive reach of human scientists.
The exocortex could also aid scientific inspiration and imagination by leaning into hallucinations.
As Yager explains, “Although hallucinations are generally undesirable, their existence is intrinsic, and there is a tradeoff between hallucinations and creativity. In other words, some amount of hallucination is desirable to enhance creativity and communication. More broadly, evaluations of LLM creativity suggest that they can generate outputs that are non-trivially novel and useful to humans.”
This vision is not without its challenges, naturally enough, and will require further advancements in AI, particularly in terms of reliability and efficiency across highly technical domains.
Yager says, “Research is required to determine how best to exploit LLMs to generate agentic modules that can perform tasks autonomously (over short timescales) by iterating on a problem.”
The development of a seamless word-based human-computer interface (HCI) will also be needed to make these interactions feel like natural extensions of the human thought process.
While still in a speculative stage, the concept of an exocortex feels like a natural evolution for AI.
As Yager says, “Interaction between a swarm of AI agents – each responsible for intelligently mediating access to a suite of research capabilities – and a human researcher should lead to the emergence of enhanced human capabilities. By expanding the researcher’s intelligence into the exocortex, the researcher can accomplish more, as they are able to intuitively and seamlessly weave myriad physical, computational, and cognitive systems into their intellectual work.”
In an article on Tech Xplore, written by some of his colleagues, Yager issues a call to arms, stating: “We’re entering unchartered territory with tremendous potential benefits for nanoscience and beyond. But no one person can do it alone. We need a community.”
