With goals of scaling down the size of artificial intelligence devices with funding from the military, scientific efforts have reached new research milestones by studying bee brains.
This is not the plot of a Black Mirror episode, but an ASU research project intended to uncover rules of brain miniaturization from bees and incorporate them into neuromorphic computing, AI software that mimics the structure of the brain.
Funded by the Defense Advanced Research Projects Agency, Ira A. Fulton Schools of Engineering professor Yu Cao and assistant professor Ted Pavlic are joining professors in ASU’s School of Life Sciences and at the University of California San Diego to study stingless bees.
Jon Harrison, a professor at the School of Life Sciences, said the grant began in March 2019.
The work continued in June, during ASU’s Tropical Biological Research in Panama study abroad program, where students worked on independent research projects in the rainforest.
Harrison said long before the Panama project, he has been personally interested in what keeps bees small and the consequences of gigantism, which can be applied to the technological world.
“You compare a cellphone to what a mosquito can do, and it's not very good,” Harrison said, commenting on the efficiency of current technology. “These systems could potentially impact everything we do.”
He said the ultimate goal of this project is to “understand the principles that allow animals with small brains to have as good cognitive processing as animals with big brains.”
Pavlic said their research is focused on a taxonomic group of stingless bees that live in the tropics that are so small that an entire colony can fit inside a coffee cup.
“The reason we focus on these, as opposed to some arbitrary insects, is that there are about 15 different species of these bees, and so within this group of 15, we can study ones that are really, really small and some that are just merely small, and that they differ in their evolutionary trajectory and how they got smaller,” Pavlic said.
The project’s website says the bees’ brains “miniaturize from 4 to 0.2 (milligrams) without measurable loss of sensory and learning capacities.”
Artificial Intelligence as an application
Many people encounter AI everyday, whether it is through Amazon’s Alexa, Apple’s Siri or self-driving cars, but Cao said there is yet to exist a device with the same mental capacity as bees at such a small size.
“The connection from the biological discovery to the computational model to the sort of engineering system altogether is fascinating,” Cao said. “It’s an interesting path, but not an easy one.”
In self-driving cars, for example, the AI has to be smart enough to take different pieces of data and prioritize them to make a decision. But small microprocessors do not yet have this capability.
“It is so important — not only perception, but also the reasoning of a system,” Cao said. “If I see a ball while driving, I’ll be more cautious. But if I see a ball popping out from a field and I see a boy, that’s a high importance object and threat to a vehicle. (The device) must be able to correlate all these different objects to make a final decision.”
Pavlic said the first half of the project required studying the brains of these bees. Now they’re moving into the second half: computational modeling, a form of problem-solving with the use of simulation and optimization.
As stated on DARPA’s website, “AI advances are beginning to have a deep impact in digital computing processes.” This means that the letters and symbols expressed in binary code are being changed, and trade-offs between computational capability and size will “become increasingly critical in the near future.”
When asked about the motivation and desired outcome of the million dollar grant, which is expected to end September 2020, DARPA declined to comment.
A look into the military’s involvement
This is not the first of ASU’s research to be funded by DARPA, as the organization has given the University tens of millions of dollars in grants that range from research on lasers to waveform transmission capabilities.
Cao said there can be misconceptions about military-funded projects, but that his goal is to help develop the next generation of engineering.
“People may think this is defense-oriented, but I don’t really think that’s the case,” Cao said. “If the results from the research get to something thought of as progress, they may connect it to defense eventually. But for the University, we focus on the fundamental side of it.”
Their team has not received direct instructions from DARPA, but that the goal of mimicking the efficiency of stingless bees' brains is shared between DARPA and the University, Pavlic said.
Pavlic said the agency has not communicated specific information as to how they would like to apply the research, but said he thinks there are a plethora of possible uses.
“They have communicated to us that they are very worried about the term limitation of silicon based transistors and the architectures we're using to put them into scale,” Pavlic said. “It seems crazy to them that animals can scale and apparently lose none of their function on a small scale, whereas we are really limited to these large computers where the only way we're getting faster is by actually putting them in parallel.”
Pavlic said he is excited for the public to begin seeing biology as a form of technological advancement.
“The Defense Department is interested in understanding how to make computation smaller. They feel that the state of the art and what we know about computation is so limited that the only next place to go is to look at insects,” Pavlic said.