In 2022, a paper published in the journal Neuron described something that sounded impossible: a small dish of human brain cells, grown on a silicon chip in a Melbourne laboratory, had learned to play the video game Pong.

Not simulated brain cells. Not a brain-inspired algorithm. Actual living human neurons, cultured from stem cells, connected to electrodes, receiving electrical signals representing the position of the ball: and responding with signals that moved the paddle.

The company behind it was Cortical Labs. And what they’re building may be one of the most significant: and most unusual: technology bets in Australian history.

What Cortical Labs Is Building

Cortical Labs is developing what it calls biological intelligence: computing systems that use living neurons rather than silicon transistors as their processing substrate.

The concept: human brain cells are extraordinarily energy-efficient learning machines. A biological neuron uses roughly one millionth of the energy of a silicon transistor performing a comparable function. The human brain: with approximately 86 billion neurons: runs on about 20 watts of power. The world’s most powerful AI supercomputers consume megawatts.

If you could harness that biological efficiency for computing, the implications for AI would be profound: vastly lower energy consumption, potentially faster learning from less data, and new capabilities that silicon-based architectures struggle to replicate.

That’s the bet Cortical Labs is making.

How the Pong Experiment Actually Worked

The Pong demonstration: published in Neuron and widely covered internationally: worked like this:

Human neurons (derived from stem cells) and mouse neurons were grown on a multi-electrode array: a chip studded with hundreds of tiny electrodes capable of both stimulating and recording the cells.
The position of the Pong ball was encoded as electrical signals sent to the neurons through the electrodes.
The neurons’ electrical responses were decoded and used to move a paddle in the game.
When the paddle missed the ball, the system sent a “perturbation” signal: random, disorganised stimulation that the neurons found uncomfortable. When the paddle connected, stimulation was more predictable.
The neurons learned, over minutes to hours, to move the paddle more reliably: not because they were programmed to, but because biological neurons naturally seek to minimise unpredictable stimulation. It’s a property called predictive processing.

The system: which Cortical Labs calls DishBrain: wasn’t playing Pong perfectly. But it was learning. And it was doing so in a way that no silicon system had replicated with comparable energy and data efficiency.

The Science Behind It

The theoretical framework underpinning Cortical Labs’ work draws on the Free Energy Principle: a theory of brain function developed by neuroscientist Karl Friston, who has collaborated with the company. The principle holds that biological systems act to minimise surprise: to make the world more predictable from their perspective.

This turns out to be a powerful and general learning principle. By designing stimulation protocols that exploit this property, Cortical Labs can get neurons to learn tasks without explicit programming: the learning emerges from the biology.

The Team and the Backing

Cortical Labs was founded in Melbourne by Hon Weng Chong (CEO), a medical doctor turned entrepreneur, alongside co-founders with backgrounds in neuroscience, bioengineering, and software.

The company has raised significant funding from Australian and international investors, including backing from Radar Ventures and participation from investors in the US and Asia. It has also received support from Australian government innovation programs.

Its scientific advisory board includes some of the world’s leading neuroscientists and AI researchers: a credibility signal that this is serious science, not speculative theatre.

Where It’s Headed

Cortical Labs’ roadmap focuses on two near-term directions:

Drug discovery and neuroscience research: DishBrain systems offer pharmaceutical companies a new way to test how drugs affect neural activity: cheaper, faster, and more human-relevant than animal models. This is likely the first commercial application.
Biological computing platforms: Longer-term, the company is developing systems where biological neural networks handle specific computing tasks: particularly learning and adaptation tasks where their energy efficiency and sample efficiency advantages over silicon are most pronounced.

The company published a follow-up paper in 2024 demonstrating improvements in learning speed and task complexity, and has been expanding its laboratory capabilities in Melbourne.

The Ethical Questions

Growing human brain cells and using them for computation raises genuine ethical questions: ones that Cortical Labs has engaged with publicly rather than avoided.

The neurons used are not conscious or sentient in any meaningful sense: they’re small clusters of cells, not anything approaching a brain. But as the technology scales, the ethical questions become more complex. Do larger neural organoids have experiences? What obligations do researchers have to biological computing substrates? These are questions the field is actively working through, with Cortical Labs participating in the conversation.

It’s to the company’s credit that they’re treating these questions as real rather than dismissing them as science fiction concerns.

Why It Matters for Australia

Cortical Labs is doing foundational science in Melbourne that would be at home in the world’s top research universities. The fact that it’s a startup: with commercial ambitions and private funding: is itself notable.

Australia has historically struggled to commercialise deep science. The path from breakthrough research to funded company to market product has been well-trodden in the US, less so here. Cortical Labs is an example of that path being navigated locally: and if it works, it will be one of the most significant technology companies Australia has produced.

🦅 The small business relevance: Cortical Labs isn’t a tool you’ll use in your café or accounting practice anytime soon. But it represents something important: the leading edge of what AI is becoming. The efficiency gap between biological and silicon intelligence is real, and the companies that bridge it will reshape what’s computationally possible. That reshaping will eventually reach every business: probably within a decade.

The Bottom Line

A Melbourne startup grew human brain cells on a chip and taught them to play Pong. That’s not a headline from a speculative fiction magazine. It happened, it was peer-reviewed, and it was published in one of the world’s most respected neuroscience journals.

Cortical Labs is doing something genuinely unprecedented: and doing it in Australia. Whatever you think about the commercial timeline or the ethical complexity, it deserves a place in any serious account of what Australian AI looks like in 2026.

Cortical Labs is headquartered in Melbourne. Learn more at corticallabs.com.