How to make an AI ‘mind’?

How does the mind-brain problem apply to the domain of artificial intelligence? Satyen K. Bordoloi finds out

Scientists in China have breached another scientific milestone with regard to the physical infrastructure of Artificial Intelligence. A South China Morning Post article reported that Chinese scientists have been able to run an artificial intelligence model as sophisticated as the human brain on their most powerful supercomputer, the Sunway.

The team used the Sunway machine to train an AI model called bangualu (alchemist’s pot) with 174 trillion parameters which are more than the number of synapses believed to be in the brain. This is a first for the world.

It is said to have a speed of a billion-billion operations per second, up to 5.3 floating-point operations per-second i.e., 5.3 exaFLOPS or 5.3 quintillion floating point operations per second. Though there is no clear consensus, it is believed that the human brain can do a billion-billion operations per second. The Sunway system does this exascale computing with the help of over 37 million CPU cores – which was four times more than the US Department of Energy’s Frontier, which was christened the world’s most powerful supercomputer in early June.

The system has a memory of 9 pentabytes which could store over two million DVD-quality movies. Communication between 96,000 semi-independent computer systems called nodes, was at a speed of more than 23 pentabytes per second. This, a researcher said, was parallel computing ability equal to human thinking ‘like eating while watching television.’

The potential uses of such an AI system are obviously immense, from solving unsolved mathematical problems to finding new medicines and protein folds, working on natural language processing, and who knows even in self-driving cars someday and then beyond.

As expected, this piece of news has generated a lot of buzz across the world, the most significant of which is that a silicon human brain has finally been created. However, this myth may be further from the truth than most other myths surrounding AI.

First is the obvious fact that this is indeed a significant development in terms of Artificial Intelligence capabilities. Right now, it is in a lab, a sort of working model prototype, but if made scalable, China would become the nation with the world’s most powerful Artificial Intelligence systems running their projects.

This is tremendous power and complexity that is comparable with the human brain which has over 1014 synapses. And like every other invention, AI can be used both for good and bad. Though there are reasons to believe that China is playing the ‘my-AI-is-bigger-than-your-AI-game’ right now as this announcement comes within weeks of a US computer being labeled the fastest supercomputer in the world, the rest of the world can only hope that China harnesses its power for the good of all humanity.

Secondly, this will create a problem of perception. Already, there is a lot of buzz around AI becoming self-aware, with a google researcher literally thinking it has come alive. To many who do not know, this will seem like humans have finally created the human brain on silicon. From a technical, numeric point of view i.e. the number of synapses in the brain, this would seem to be true. It would also be true from the point of view of how many calculations it can perform.

What is not true, however, is that merely copying the number of the brain’s synapses will somehow automatically lead to a human brain or consciousness.

The first point to remember is that that thing on top of our shoulders has two things: the physical brain and what that brain conjures – the metaphysical mind.

Even if we have created a replica of the human mind on silicon, would that naturally lead to a mind or consciousness? That is the billion dollar question here, whose answer this model will not answer for one simple reason: even neuroscientists do not yet know the full intricacies of how the human brain works, and how its various parts interact.

How little we know about our brains is absurd. Firstly, each node or connection in our brains is unlike any in a computer: they are alive with each node having a proverbial mind of its own. Secondly, the human brain does its 1 exaFLOP calculations per second – unlike the hundreds of thousands of watts of power for any computer system anywhere – on just 20 watts of power. We don’t have the slightest clue how.

The knowledge about what the human brain has and how it operates is still a work in progress that will take decades, if not centuries, to complete. Thus, creating even a near-close replica of the human brain right now is impossible. What the Sunway AI machine has done, and others will no doubt do soon, is replicate the number of calculations that the human brain can do per second. This is brute force and we all know that this is not what the brain is all about.

Take for example the brains of animals. Their brains are similar to humans and must be making a similar number of calculations as the human brain. The brains of some of the smarter animals like elephants, dolphins, whales, octopuses etc perhaps make more calculations than the human brains. Yet, they do not come close to what humans are capable of doing. These animals do have consciousness, but they do not have consciousness similar to humans. In other words, different species have similar brains, but dissimilar minds.

Thus, replicating the human brain is not about copying the brain, but somehow finding the connections and meanings that make up the mind. The mere number of calculations will not cut ice when it comes to creating an artificial mind from a silicon brain.

Human beings, as I have pointed out in previous Sify articles, have an innate habit of anthropomorphizing everything – we attribute human characteristics to things that are not human. The entire animation industry runs on that. What better thing to anthropomorphise than literally a piece of technology that mimics part of the human brain’s operation?

Thus, calling an AI human is part of humanity’s old habits. We will do the same to the new Chinese AI system by Sunway and any other that will come up that matches the human brain’s capabilities. However, those of us who still care about logic must remember this simple fact: we might have duplicated the human brain, but how will we duplicate the human mind?

In case you missed:

• Copy Of A Copy: Content Generated By AI, Threat To AI Itself
• Nuclear Power: Tech Giants’ Desperate Gamble for AI
• The ChatGPT Legacy: How It Changed Our Perception of AI Forever
• How Old Are We: Shocking New Finding Upends History of Our Species
• OpenAI’s Secret Project Strawberry Points to Last AI Hurdle: Reasoning
• 10 Years of ‘Her’: What the Film Got Right about AI, What It Did Not
• The Path to AGI is Through AMIs Connected by APIs
• AI as PM or President? These three AI candidates ignite debate
• Tears of War: Science says women’s crying disarm aggressive men
• PSEUDO AI: Hilarious Ways Humans Pretend to be AI to Fool You