Written by Zinnia Lee
At Cortical Labs, CEO Hon Weng Chong and his team have trained human brain cells on computer chips to play the world's first video game, Pong. Zhong is now turning the Australian startup's technology into a biocomputer, which he hopes will not only surpass artificial intelligence, but also drive growth for other companies.
Hongwen Zhong, CEO of Cortical Labs. Image credit: Cortical Labs/Forbes Asia
Last Christmas, Amazon called Zhong Hongwen to say its chief technology officer wanted to visit Cortical Labs' base in Melbourne, which gave him a Christmas surprise. Werner Vogels, chief technology officer at the world's largest cloud provider, is intrigued by the sci-fi innovation of the four-year-old startup: a computer chip powered by human brain cells, called DishBrain, that has learned to play Atari's game Pong using the energy equivalent of a pocket calculator.
"Warner tells us that the biggest cost to any data center or cloud provider is the energy they pay to run the equipment and cooling system." Hongwen Zhong, co-founder and CEO of Cortical Labs, recalls. "Our system consumes almost no energy and outputs very little heat. The unit economics are completely reversed, especially when you start training it to perform AI tasks. ”
Training lab-grown living brain cells on a chip to play tennis-like video games is just the beginning. By combining the extraordinary learning power of the human brain with the processing power of silicon chips, Zhong is building biological computers that he claims could be smarter and more efficient than today's artificial intelligence. The 35-year-old doctor-turned-entrepreneur is commercializing the innovation in the hope that one day it will disrupt applications — from testing new drugs to treat brain diseases to cutting huge energy bills needed to train artificial intelligence.
In a video interview from his Melbourne office, Zhong said: "The ultimate goal of Cortical Labs is to let other users be creative with this technology, just like NVIDIA. In five years, we want five to ten successful companies or startups running their own businesses, but using our technology. ”
Amazon CTO Warner Vogels (left) visits Cortical Labs' base in Melbourne in early 2023. Image credit: Cortical Labs
In April, Cortical Labs raised $10 million in a funding round led by Horizons Ventures, a private investment arm owned by Li Ka-shing, Hong Kong's richest man. Also joining the round were Blackbird Ventures, one of Australia's largest venture capital funds, and In-Q-Tel, a venture capital arm owned by the CIA.
With the new capital, Cortical Labs aims to start generating revenue when it starts selling its biocomputers by the end of this year. Zhong said the company had signed up with Bit Bio, a spin-off company at the University of Cambridge, to experiment with its products. Bit Bio provides human cells for research. He added that by the end of 2024, Cortical Labs will begin offering cloud services with clusters of 120 biocomputers for companies to program brain cells for different tasks. Zhong also revealed that Cortical Labs is in talks with several U.S. cloud service providers to access its services.
Of course, Zhong Hongwen faces a difficult task. For two decades, scientists have been trying to fuse living cells with silicon, hoping they can build an AI system that works like a human brain while advancing applications in medicine and other fields. Back in 2004, a biomedical engineer at the University of Florida said he had trained a set of rat brain cells arranged on a grid of electrodes to control a flight simulator of the F-22 jet. So far, however, no one has succeeded in building a biological computer that can compete with traditional computers.
Cortical Labs has gone further than most companies in commercializing biological computers. A rare example is California-based Koniku, which combines living cells with computer chips to create sensors that mimic the ability of sniffer dogs to detect explosives and drugs. The company has partnered with European aircraft manufacturer Airbus to deploy its equipment for aviation safety.
But not everyone will be impressed. Madeline Lancaster, a researcher at the University of Cambridge's Laboratory of Molecular Biology and a pioneer in the lab's creation of mini-human brains, points out that playing Pong was something that high school calculators could have done decades ago. In a written response, she said: "The neurons in the petri dish (information messengers in the human brain) appear to be stuck and there has been no significant improvement in their computational power in 20 years. It's unclear how Cortical Labs plans to overcome this hurdle and achieve the kind of exponential growth that computers need, so it's hard for me to see real viability. ”
In addition to the technical hurdles, Cortical Labs faces ethical questions — whether lab-grown brain cells are conscious, whether they can feel pain and pleasure, and so on. The company says its brain cells are "sentient" and defines it as "responding to sensory impressions."
Cortical Labs has been working closely with bioethicists and will continue to engage in dialogue with society on ethical issues, he said. However, on the technology side, Zhong acknowledges that there is a "huge technological moat" for anyone entering the field, and that the field is full of uncertainty.
"At Cortical Labs, it's very difficult because we're not only having to make hardware like Nvidia, but also write software like OpenAI. Most importantly, we had to prove that this biological computer could be scaled even further, and we could let it do more than just play Pong. ”
Zhong joked that Cortical Labs' biological computer is "a body in a box." The human brain cells it uses are made from engineered stem cells extracted from adult skin or blood. They are then integrated into a chip and placed in a device the size of a shoebox, which contains a system that provides nutrients and air to the cells and removes waste products from the cells. Brain cells are essentially like the central processing unit (CPU) of a traditional computer.
Image credit: Cortical Labs/Forbes Asia
To train brain cells to play Pong, Cortical Labs connected a hybrid chip to a computer running the video game. The computer sends electronic signals that show where the bouncing ball is and how far away it is from the racket. These brain cells decide for themselves how to move and learn to improve their performance through feedback from electrical signals.
Cortical Labs' biocomputers, he said, could help test the efficacy and side effects of drugs that target neurological disorders such as dementia and epilepsy. This, he says, can be achieved through the representation of human brain cells in games like Pong. The former doctor hopes to work with pharmaceutical companies such as Biogen and Eli Lilly to replace animal cells with human cells in drug tests.
"In the short term, Cortical Labs' biocomputers lead to compelling in vitro solutions that will allow us to further our understanding of how drugs affect neurons." Jonathan Tam, an investor in Victoria Harbor Investments that has invested in Cortical Labs, said in a written response. He is responsible for investments related to artificial intelligence. "In the long term, these synthetic neurons will provide unprecedented insights into our brain function, which will open avenues for treating previously unsolvable neurological diseases."
Cortical Labs' biological computers could also help solve a more practical problem: high electricity bills. Zhong Hongwen said that biological computers can significantly reduce the energy cost of training artificial intelligence, while using cloud computing requires a lot of electricity because it needs to read a large amount of data. In addition, it requires a lot of coolant to keep the learning process going.
For example, according to a 2021 study, the training of GPT-3, the brain behind OpenAI chatbots, consumed 1287 gigawatt-hours of electricity. That's equivalent to the electricity consumption of about 120 U.S. homes in 2021. On the other hand, the human brain is said to operate at about 20 watts, which is equivalent to the energy that powers an LED bulb.
Image credit: Cortical Labs/Forbes Asia
Zhong Hongwen said that through the innovation of Cortical Labs, the cloud computing platform of artificial intelligence can reduce energy consumption "exponentially to the order of 6 to 8 watts." "We really hope that with this technology, energy will no longer be a bottleneck," he said. It may also be more sustainable, because instead of feeding brain cells with electricity, we use a simple sugar, glucose, a compound like that. ”
This is not the first time Zhong has implemented ambitious plans. When he was a student at Melbourne Medical School, he saw the need for diagnosis in places with few medical resources and developed a low-cost stethoscope that could be plugged into a smartphone.
After working as a doctor for a while after graduating, he came up with the idea again in 2014 and co-founded CliniCloud to commercialize a digital stethoscope that allows anyone to listen at home and consult a doctor online. The medtech startup has received investment from Chinese tech giant Tencent and Ping An Group's venture capital arm.
However, a lack of health awareness prior to the pandemic, combined with uncertainty over former President Donald Trump's administration reforming the U.S. healthcare system, led to the collapse of CliniCloud's business in 2018. "ClinicCloud was far ahead of its time. If we could live a few more years, we would really take off because of Covid. In fact, at the beginning of the pandemic, there were a lot of people calling me and asking for a thermometer and stethoscope, and literally writing me a blank check. "
By then, Zhong Hongwen had activated his AI fulcrum. Inspired by a research paper by Alphabet's Google DeepMind, Zhong co-founded Cortical Labs in 2019 with Andy Kitchen, former head of AI research at CliniCloud. The startup's vision of a future of building artificial intelligence systems using human brain cells attracted a seed investment of A$1 million ($673,000) from Blackbird Ventures, followed by an expansion investment of S$1.2 million ($895,000) from investors including Singapore's January Capital.
Now, amid the AI boom, Zhong is busy getting Cortical Labs' biological computers into production. He said many companies ask him what else brain cells can do. One of the "ridiculous" questions is whether cells can learn to trade bitcoin, Zhong laughs. As crazy as it sounds, Zhong doesn't dismiss the idea. Instead, he offered to write a software code so that the inquirer could send the bitcoin price to brain cells and conduct experiments.
"If it works, it's great. It's your technology, your intellectual property. Make a lot of money this way. We shouldn't be completely dismissive because a lot of ideas sound really silly at first, but you never know where it's going to go. ”
This article is translated from https://www.forbes.com/sites/zinnialee/2023/06/21/cortical-labs-brain-computer/?sh=2ce380d6c42d