Brains vs Machines - Can wetware make AI energy-efficient? A sketchnote
A sketchnote exploration
This sketchnote dives into a fascinating question: Why is the human brain still vastly more energy-efficient than today’s AI systems? And could wetware — a term for bio-inspired computing made from living tissue or brain-like systems — help close the gap?
📖 Inspiration
The idea was sparked by a comment from Max Bennett, author of A Brief History of Intelligence: Why the Evolution of the Brain Holds the Key to the Future of AI. In a recent talk, Bennett suggested that “wetware” might be one of the keys to building more energy-efficient artificial intelligence.
💭 Brain vs AI: Energy efficiency
The human brain uses about 20 watts of power (about the energy of a dim lightbulb (dim, lol). The brain also has around 86 billion neurons, each connected to thousands of others. It achieves this efficiency through parallel processing, event-driven computation, and biochemical mechanisms. In contrast, LLMs rely on data centres running megawatts of power, and thousands of GPUs to train and run models at scale.
What’s “wetware”?
Wetware refers to biological or biohybrid computing system that use organic material like cells, neurons, and proteins, instead of, or in combination with, silicon hardware. The aim is to emulate the structure and behaviour of the brain. Some wetware approaches:
🧫 Organoid intelligence (OI): Lab-grown clusters of human brain cells (mini-brains) (ew?!) trained to perform computational tasks. They can even learn patterns like playing Pong. They have the same energy efficiency and adaptability as brains.
⚡Neuromorphic computing: Use silicon but mimics biological signalling found in the brain (spiking neurons and synapses). Intel’s Loihi chip is an example of this.
🔬Molecular or DNA computing: Uses strands of DNA or proteins to perform computations!
Growing and controlling biological systems is difficult though. Current wetware systems are primitive compare to AI models. Aside from the ethical concerns, interfacing biology with silicon is an engineering hurdle.
⚖️ Ethical reflections
As wetware research moves from theory to lab-grown brain organoids and hybrid bio-digital systems, ethical questions inevitably emerge. What rights, if any, should be considered for sentient or semi-sentient bio-systems? The idea of growing brain-like tissue to solve machine problems touches on profound concerns around consent, suffering, and the moral responsibilities of creators. As with many breakthroughs in AI and neuroscience, the challenge is not just technological, it’s deeply human.
🖌️About the graphic
This is one of the first sketchnotes I’ve created using generative AI tools — specifically, ChatGPT-4o for content and layout concepts, and Photoshop for text clean-up and refinement.
The process:
I asked ChatGPT to explain the topic and generate a summary suitable for a sketchnote.
I then prompted it to create an illustrated version in a specific style:
“Create an illustrated version. The style of the line-work should be delicate and loose, using blue ballpoint pen ink and pencil, leave some parts unfinished so it looks like a creative, intellectual scientist’s actual personal research notes.”
Once generated, I tweaked the image in Photoshop to correct the text and enhance legibility, preserving the original sketchbook feel.
The result is a hybrid creation — part human curiosity, part machine assistance — hopefully reflecting the very topic it explores.