Австралийн Cortical Labs байгууллагын судлаачид лабораторид ургуулсан хүний мэдрэлийн эсийг робот нохойн системд холбож, түүнийг тодорхой цэг рүү чиглүүлэх туршилт амжилттай гүйцэтгэлээ.
Cortical Labs-ийн гүйцэтгэх захирал Брэтт Каганы тайлбарласнаар, энэхүү туршилт нь биологийн систем дэх дэвшилтэт технологи бодит ертөнцөд хэрхэн хэрэгжиж болохыг харуулсан анхны алхам юм. Судлаачид чип дээр ургуулсан мэдрэлийн эсүүдийг электродтой холбож, цахилгаан дохио дамжуулах хоёр талын сувгийг бий болгожээ. Энэхүү санал хүсэлтийн гогцоо буюу давтамжтай цахилгаан дохионы тусламжтайгаар роботыг хүссэн чиглэлдээ хөдөлгөх боломжтой болсон байна.
Honda Research Institute Europe-той хамтран хийсэн тусдаа туршилтаар эрдэмтэд 4,000 гаруй цагийн турш 1,300 орчим өөр төрлийн дохионы хослолыг туршиж, виртуал орчинд агентыг удирдах дасгал хийжээ. Хэдийгээр энэ нь туршилт, алдааны замаар явагдсан хүндрэлтэй ажил байсан ч биологийн эсийг тооцооллын үндсэн бүрэлдэхүүн хэсэг болгон ашиглах нь эрчим хүчний хэмнэлт, дасан зохицох чадвар болон системийн бат бөх байдлыг сайжруулах том боломж гэдгийг судлаач Даниел Таннеберг онцолсон юм.
Уламжлалт хиймэл оюун ухаантай харьцуулахад хүний тархины эсүүд нь маш бага эрчим хүч зарцуулж, танил бус орчинд илүү сайн дасан зохицдог. Гэсэн хэдий ч энэхүү технологи нь одоогоор эхэн шатандаа байгаа бөгөөд илүү нарийн төвөгтэй даалгавруудыг богино хугацаанд гүйцэтгэх, биологийн санах ойн механизмыг ойлгох зэрэг сорилтууд тулгарч байна. Судлаач Жонсон Жоугийн тэмдэглэснээр, гол бэрхшээл нь зөвхөн мэдрэлийн эс ургуулахдаа бус, харин тэдгээр системтэй найдвартай харилцаж, тэдгээрийг сургах тогтвортой арга замыг олоход оршиж байна.
Дэлгэрэнгүйг эх сурвалжаас харах
↓Эх сурвалжийг нээх ↓
Researchers at Cortical Labs in Australia have attached lab-grown neurons to a robotic dog and used them to steer it toward a goal on a simple map. The demonstration shows living cells directing a machine in the physical world, rather than only inside a computer.
Computer chips built on living neurons are drawing attention because the human brain runs on far less power than sophisticated AI and remains much better at handling the unfamiliar. The robot dog is the latest demonstration of that idea, and it follows earlier experiments with the same technology.
The work sits at an early stage. It is not particularly fast or efficient yet, but it gives a tangible picture of what biological computing might one day do.
How the Neurons Drive the Robot?
The neurons are grown right on the chip, which is fitted with electrodes that carry electrical signals in both directions. That two-way channel is what lets the team turn activity in a small group of neurons into movement for the attached robot. Feedback loops of electrical signals were used to get the dog to its target.
“The robotic dog is a simple demonstration of how progress in biological systems can also be applied into the real world, but it helps make the opportunity tangible,” says Brett Kagan, co-founder and CEO of Cortical Labs.
He adds that, longer term, the approach could support robots and autonomous systems that must make decisions in unpredictable, real-world environments.
How Honda Is Shaping The Training?
Cortical Labs has also run a similar demonstration with Honda Research Institute Europe. Everything there happened inside a virtual system, but the aim was the same: directing an agent around a basic grid in search of a reward.
The setup mirrored the robot dog. Electrical signals stimulated neurons on a chip, and the chip sent electrical messages back. By varying the characteristics of the signal and supplying feedback on the agent’s moves, the researchers could train the living chip and interpret its movement.
Over more than 4,000 hours of interactions with the chip and agent, the team tested almost 1,300 different signal combinations. There was still, it should be said, a fair amount of trial and error involved. Daniel Tanneberg, a Senior Scientist at Honda Research Institute Europe, says using living cells as the core computational substrate brings many challenges but also offers great potential in energy efficiency, robustness and adaptation.
What We Still Don’t Know
Cortical Labs had already shown neuron clusters completing computer tasks, including playing Pong and Doom, and is now extending the idea into robotics. Kagan points to a gap in conventional machines.
“Robots that are becoming remarkably capable machines at their physical control, still struggle with the higher order control required to learn quickly or adapt when conditions change,” according to him.
However striking a robot’s walking, running or acrobatics may be, he notes, knowing where and when to act remains difficult with traditional techniques.
The next steps are to scale up the chips to take on more complex tasks in a shorter time, since robots in the real world will need to do much more than cross a simple grid. Another aim is to learn more about how these neuron-based chips retain information, and what kind of “memory” feature they might have. Johnson Zhou, a Research Software Engineer at Cortical Labs, frames the challenge plainly.
“The challenge is not simply growing neurons on silicon: It is finding reliable ways to communicate with biological systems, train them and understand which approaches produce consistent results.”
Enjoyed this article? Subscribe to our free newsletter for engaging stories, exclusive content, and the latest news.
