Artificial Intelligence and the Brain Close in Power Far in Nature
Not very long ago, Artificial Intelligence felt more like a party trick than a partner in thought. Early LLMs (Large Language Models) could be tripped up with clever prompts, nudged into contradictions, or made to confidently state nonsense.
Today, that same class of systems writes software, helps draft Research papers, sifts through mountains of documents in seconds, and even interprets images and audio alongside text. Tools such as ChatGPT, Gemini, and Claude have moved from curiosity to utility in just a few short years, marking a dramatic shift in the capabilities of Artificial Intelligence.
What changed so dramatically?
The Power of Artificial Intelligence Scale and the Innovation
The answer is less about a sudden flash of inspiration and more about scale. The basic idea behind modern Artificial Intelligence is surprisingly old. In 1943, Warren McCulloch and Walter Pitts sketched a Mathematical model of a neuron, a simple unit that takes inputs, weights them, sums them, and produces an output. On its own, it is almost trivial. But connect enough of these simple units together and, as mathematicians later proved, they can approximate almost any pattern you care to describe.
For decades, that promise remained mostly theoretical. Computers were too slow, datasets too small. Then came the explosion of computing power. Graphics processing units, first built to render video games, turned out to be perfect for training massive neural networks. At the same time, Researchers developed new ways of organising these networks. The biggest breakthrough came with the “transformer” architecture, which allowed models to attend to different parts of a sentence at once rather than reading words in sequence. This breakthrough reshaped the trajectory of Artificial Intelligence Research.
GPT (Generative Pre-trained Transformer) builds on that idea. It learns by predicting the next word in a sentence, over and over again, across vast swathes of text. It sounds simple. But at an enormous scale, that simple task forces the model to absorb grammar, facts, style, and even fragments of reasoning embedded in language. What emerges can look strikingly like understanding, one of the most debated outcomes in the rise of Artificial Intelligence.
Artificial Intelligence vs. Minds: Similar Numbers, Different Designs
As these systems grow larger, comparisons with the human brain have become inevitable. GPT-3, for example, had 175 billion parameters; newer systems are said to push into the trillions. The human brain contains on the order of 100 trillion synapses. The numbers are suddenly in the same universe. And yet, the resemblance may be more superficial than it seems.
Most LLMs process information in a largely feed-forward way: input goes in, layers transform it, output comes out. Efficient, scalable, and well-suited to modern data centres powering Artificial Intelligence. The brain, by contrast, is a dense web of feedback loops. Signals move forward, backward, and sideways. Perception is not a one-way street but an ongoing conversation between what we expect and what we see.
The Road Ahead: Imitation or Innovation?
Biology achieves this with remarkable efficiency. Neurons fire in brief electrical spikes. If they stay silent, they consume very little energy. Memory and computation happen in the same place, at the synapse. The entire brain runs on roughly 20 watts, about the power draw of a couple of LED bulbs. By contrast, training and operating large Artificial Intelligence systems can require vast data centres that consume megawatts of electricity and process trillions of words, far more text than any human encounters in a lifetime.
Researchers are now trying to narrow that gap. New architectures activate only specialised parts of a network for a given task, echoing the brain’s modular organisation. Experimental “neuromorphic” chips attempt to mimic spike-based signalling to cut energy use. But these are still approximations. Artificial neurons remain simplified Mathematical constructs; Biological neurons are living, Chemical systems of extraordinary complexity.
Where this path leads is still uncertain. Machines are not bound by the evolutionary constraints that shaped our brains. They may eventually rival, or even surpass, Biological intelligence in certain domains. Or they may diverge entirely, developing forms of intelligence that look nothing like our own. The future of Artificial Intelligence may depend not on imitation, but on innovation.
After all, a pacemaker supports the heart without resembling heart tissue. Perhaps Artificial Intelligence will do something similar for the mind, extending and augmenting human thought without ever truly replicating it. In the end, the real question may not be whether Machines think like us, but whether they can help us think better.
