Lab-grown human brain cells drive virtual butterfly in simulation
Could organoid-driven computing be the future of AI power?
Researchers affiliated with the neuroscience platform FinalSpark have devised a 3D simulation depicting a butterfly that's directed by human brain cells.
"It represents a significant step towards the realization of concepts previously confined to science fiction, such as The Matrix, and opens up new avenues for research in cognitive preservation and mind uploading," said software developer Daniel Burger in a summary of the project and an associated video.
FinalSpark, a Switzerland-based startup, offers its Neuroplatform so researchers interested in biocomputing can write code that interacts with brain organoids, which are pea-sized mini-brains derived from induced pluripotent stem cells.
According to Burger, these cells are maintained in incubators at 37°C – more cell sauna than The Matrix. While the cells lack the complexity of the human brain, they'll nonetheless respond to stimuli through a multi-electrode array (MEA) that supports two-way electrical communication.
As a platform demonstration, Burger and colleagues created a virtual world with a butterfly model. The brain organoids watch for human input signals, which in this case take the form of clicks on the virtual world scene. The associated Python software computes whether the click event occurred within the butterfly's field of view. Then the brain neurons respond to the stimuli by telling the butterfly model to fly toward the click event location or not.
A simplified version of the Python decision-making code looks like this, with the movement handled by separate vector calculations:
function decideButterflyAction(isTargetVisible: boolean, elicitedSpikeCount: number): boolean {
const SPIKE_THRESHOLD = 5; // exemplary
const TIME_WINDOW = 200; // milliseconds, also exemplary
if (isTargetVisible && elicitedSpikeCount > SPIKE_THRESHOLD) {
return true; // Move towards target
}
return false; // Move randomly
}
"It’s crucial to emphasize that while these movement functions are implemented in software, the decision to use one or the other is driven by the brain organoid’s response to stimulation," said Burger.
- Spectre flaws continue to haunt Intel and AMD as researchers find fresh attack method
- Open source LLM tool primed to sniff out Python zero-days
- Gary Marcus proposes generative AI boycott to push for regulation, tame Silicon Valley
- Big browsers are about to throw a wrench in your ad-free paradise
Though met by the YouTube commentariat with a fitting internet meme reference – "At long last, we have created the Torment Nexus from the classic sci-fi novel Don't Create the Torment Nexus" – Burger maintains his butterfly "controlled by a lab-grown mini-human brain remotely connected to the internet" promises several benefits if the technology can be successfully refined.
Biological neural networks (BNNs), Burger argues, have potential advantages that "include significantly lower energy consumption" – he notes that the human brain operates on only about 20 watts – "truly higher cognitive and adaptive behavior such as creativity, true zero-shot learning capabilities, superior pattern recognition and generalization, better handling of ambiguity and noise, and the potential for self-repair and neuroplasticity."
A point of comparison, FinalSpark cites the Hewlett Packard Enterprise Frontier supercomputer, which operates at 21 megawatts, or 8–30 megawatts by another measure.
Burger however acknowledged that there's additional traditional computing infrastructure involved that needs to be considered. "We currently compare CPU vs. organoid processing directly," he told The Register. "We still have energy consumption from supporting hardware like incubators and electrical stimulation systems. We don't yet have an exact 1:1 comparison of CPU vs. organoid including all supporting hardware, but this is something we're working on quantifying in the near future."
He also said while the need for a software layer is expected to continue, it may become less significant over time. "Based on our current work, it appears that software interfaces will remain necessary when brain organoids interact with human-built systems like computers," he said. "Much like how we as biological intelligent beings use software to communicate (e.g., you writing to me), this intermediary layer will likely always be present. However, we anticipate the supporting software may become more streamlined in the future."
Burger sees applications for "zero-shot learning capabilities for entirely novel scenarios, tasks demanding intuition and creativity, and pattern generalization across a wide range of tasks, rather than narrow, optimized tasks."
At a time when major tech companies with AI commitments are exploring nuclear power plants to keep AI models running, the utility of living neurons, which consume 1 million times less energy than current digital processors, becomes obvious.
The ethical implications are perhaps less clear, which is why Burger points out that these organoids lack the capacity for high-level consciousness and should be thought of more like plants than people.
And speaking of consciousness, Burger contends that this technology – at some point, some day – could lead to the development of "ectopic cognitive preservation."
"This concept includes theoretical and practical methodologies for capturing and maintaining the structural and functional integrity of human cognitive processes beyond the limits set by biology, including aging, cognitive decline, and, ultimately, lifespan extension," he explained.
If you want a picture of the future, imagine billionaires sharing their thoughts forever. ®