
The sleek, faceless android struts on metal legs down a corridor lined with autonomous robotic arms, each softly whirring as it operates pipettes and all manner of lab equipment. Its reflective black faceplate watches the human researchers mill about the automated laboratory , tracking their performance and adherence to protocol. Deviate — misplace a vial or mis-pipette an RNA sample — and “Supervisor” will know.
The scene feels straight out of science fiction. But Supervisor is very real — and it’s learning more every day. Within six months, its creator believes the artificial intelligence-powered machine may become so well-versed in scientific procedure it could replace a human lab technician.
At a housing construction site in Gananoque, Ont., Val spends nearly two months laying concrete, used to build 26 stacked townhouse units. At a housing construction site in Gananoque, Ont.
, Val spends nearly two months laying concrete, used to build 26 stacked townhouse units. The world’s first AI-powered, android scientist Today, what is believed to be the world’s first AI-driven, humanoid research robot is already at work supervising human workers and giving tours at Insilico Medicine’s automated LifeStar1 laboratory, located in Suzhou, China. It still has much to learn from its human peers, said Alex Zhavoronkov, the Canadian founder and CEO of Insilico, a biotechnology company devoted to automating research.
His scientists, kitted out with motion-tracking gear and virtual reality goggles, now spend much of their time controlling the humanoid and teaching it to perform standard protocols like handling microchips, navigating the lab and pipetting microlitres of fluid. While the human technicians explains to the robot what each procedure is achieving, the AI-powered lab is itself learning to adjust to the android’s presence and actions through sensors and cameras. All this information, as well as hundreds of thousands of motion-captured recordings of the human workers, are being fed into the machine’s AI system, slowly training the humanoid how to perform science.
This “extremely valuable” data can then be sold to other humanoid robotics manufacturers. “The current stage of the humanoid development is we are creating a data factory,” Zhavoronkov told the Star. “We are training the humanoid to perform many, many laboratory scientist tasks .
.. every time we train the embodied AI model and upload it into the humanoid, it will become more and more capable.
” The government has boasted that Canada has 10 per cent of “the world’s top-tier AI researchers, the second most in the world.” The government has boasted that Canada has 10 per cent of “the world’s top-tier AI researchers, the second most in the world.” Insilico’s lab is already largely automatic, using a blend of AI and robotics to autonomously discover new drugs combating cancer, aging and numerous diseases around the clock.
A number of the company’s drugs are already being tested on humans — one of their pharmaceuticals designed to treat a devastating, chronic lung disease recently passed Phase IIa clinical trials, for example. But an actual human-shaped robot is expected to be a game-changer, Zhavoronkov said. “The advantage of a humanoid is that it is future-friendly .
.. You want to have a device in the lab that is as or more capable than a real human,” he said.
Most lab equipment are designed to be used by humans, without automation in mind, he said. It’s also difficult for non-humanoid machines to perform regular maintenance on lab devices. But if all goes as planned, Supervisor would be capable of doing anything a human lab technician can.
Will robots replace human workers? Zhavoronkov expects the robot to one day replace human lab workers in performing basic tasks. “For laboratory scientists, for technicians, I think that definitely they need to be replaced and improved,” he said. This should serve as motivation for human workers to “learn and start inventing new things” instead of becoming complacent in their skills, he argued.
“Humans need to continuously invent. If you don’t invent, you are going to be obsolete and replaced by robots,” Zhavoronkov said. It’s been more than a year since OpenAI released its first demo of ChatGPT, kicking off a tech arms race.
It’s been more than a year since OpenAI released its first demo of ChatGPT, kicking off a tech arms race. The idea has become a growing concern for Alán Aspuru-Guzik, director of the Acceleration Consortium — an organization focused on building self-driving laboratories — and a professor of chemistry and computer science at the University of Toronto. “We always have to think about this,” he said.
“We want to be very careful ...
If I am directing a laboratory, I will never replace a person with a robot. Rather, (I would) use a robot so that the person can do more sophisticated things.” The key is to upskill — to train humans to work alongside robots, and to ensure robots will support and boost human workers’ productivity, not the other way around, Aspuru-Guzik said.
The problem occurs when a company replaces its human workers with a robotic crew entirely, he said. “I think if we don’t pair it with upskilling and we don’t pair it with things like universal basic income in the future as well, then we might be in trouble,” he said. “I do think about this and I get scared about it .
.. because there are some people that won’t be able to upskill.
” That said, and although he is not affiliated with Insilico’s project, Aspuru-Guzik’s team is also “considering procuring a humanoid robot” for one of its labs. “We really want to increase our productivity without sacrificing a single person. We want to make sure everybody has a job, always.
” he said. “But that’s just me.” A robotic supervisor Our columnist worries about the lack of oversight for emerging tech.
Our columnist worries about the lack of oversight for emerging tech. According to Zhavoronkov, his robot makes for the ideal lab supervisor. “Humans can get sloppy when they are not supervised, and very often they can do things that are maybe outside protocol,” he said.
It’s possible to closely monitor their performance through CCTV surveillance, he mused, but noted that “people don’t like that.” On the other hand, after thousands of hours of training, Zhavoronkov’s android should have become so good at by-protocol tasks, it could track and point out errors in his workers’ performance, he said. “Sometimes, an angry human boss also may miss mistakes and not exactly help the scientists,” he said.
“I think that humanoids will replace human supervisors.” At the moment, Supervisor remains more of a pilot study. But once its training is complete, “we want to scale globally” to labs around the world, Zhavoronkov said.
“The idea is to accelerate the discovery of new medicines and enable seamless workflows that do not require humans,” he said, like automatically discovering vaccines and therapeutics for the next pandemic. “That will be the future in, let’s say, the next two-and-a-half years,” he said. “It’s very difficult to predict what’s going to happen in the next five.
”.