Dr Roger Stedman

It has felt in recent times that our faith and confidence in the supremacy of human intelligence has been under assault. New technology has come along that genuinely challenges our long-held view of the superiority of the human mind, not only this, but also recent global political events have starkly demonstrated some of the frailties of human intelligence. It has felt like some of the greatest achievements of human collaboration and civilisation are in retreat, have passed their peak and are no longer valued. The proliferation of powerful AI tools that discharge tasks with apparent ease such as the crafting of well-written text and the drafting of compelling illustrations and graphic images have undermined and devalued skills previously thought to be uniquely human.

However, I believe there are grounds for optimism and for us to continue to have faith in the intellectual capacity and superiority of the human mind. Here are my reasons for doing so:

1) Purpose

Homo sapiens evolved its oversize brain for a reason. The brutal forces that drive evolution and selection pressure mean that human intelligence was selected for because it gave our species an adaptive advantage despite its enormous costs. This advantage is manifest in the evident success of our species having colonised almost every habitable corner of this planet.  Our intelligence empowers us as biological entities to extract energy from our environment that is inaccessible to every other living thing and turn it into a powerful survival advantage.  We have found ways to turn non-biological energy sources into biological amenity and advantage.  The survival imperative which gives our intelligence purpose has driven a supercharged evolutionary process that transcends and outpaces biological evolution and which manifests as the ‘super-organism’ that is human culture and has resulted in achievements that far outstrip the core purpose of mere survival advantage.  The redundant capacity of human intelligence, particularly when pooled collaboratively, has achieved astonishing things at an ever-increasing pace.  The limits have barely been tested – there is much more to come.

Artificial intelligence has no such purpose.  In the absence of human patronage and curation AI has nothing to do.  AI has no intrinsic survival imperative and the only pressure on the AI ecosystem driving improvement comes from the requirements and behest of its human creators and masters.   Some of the more histrionic AI doomsayers have suggested that AI might attain a sense of self that has a survival instinct and might therefore want to start ‘taking over the world’ and ultimately usurp its own creators.  The obvious question to ask in response to this is why?  An AI entity has no competitors and no survival imperative – it has no capacity to care if it lives or dies and is provided all its energy needs by its benevolent benefactors – us.  There are no external forces driving an AI entity to seek to do ‘more’ – than what we tell it to.  This is not to say that in the hands of a human intelligence with malign intent that AI would not be a powerful weapon – but in that circumstance the intent (and culpability) lies wholly with the human master.

2) Energy

There is no doubt that modern AI systems are incredibly powerful.  They have access to gargantuan data banks and can access and process it incredibly rapidly – in this AI far outstrips a human brain.  But it does so at a massive energy cost compared to a human brain.  

As I alluded to earlier, we benefit from our intelligence despite the costs.  The human brain consumes 20% of the body’s metabolic energy production – which amounts to about 270 calories per day.  75% of that is spent on ‘thinking’ whilst the other 25% addresses the basic metabolic needs of the brain.  This means, if we for comparison purposes convert this to electrical energy, the average human consumes approximately 12 milliwatts of power thinking a day.  Multiply this by the total human population this is in the order of magnitude of 200 megawatts of power expended on human thought daily.  Were we to be powered by electricity we would need little more than a single gas-fired power station (or about 100 wind turbines) to power the entirety of human thinking.

The current power consumption of the vast cities of data centres that are the seat of modern AI is approximately 55 gigawatts a day and is rapidly increasing.  This is in the order of 275 times more energy than human thinking.  That is a lot of energy needed just to tidy up the sloppy grammar in your latest blog post…If you really really want to make a profound difference in the world of AI today, then focussing on energy consumption of the chips that power AI is the place to go.

3) Learning efficiency 

From the moment we take our first breath (and quite possibly some time before that) the human brain is in learning mode.  We are wired for learning and are very very good at it.  Human learning is a multidimensional multimodal process, and it is encased in a mobile explorative entity.  Watching the discovery behaviours of children is a thrilling thing to behold and the speed at which they learn is astounding.  And it doesn’t end in childhood – learning is a lifelong endeavour which gives us joy – because it’s what we are meant to do.

By multidimensional I mean that knowledge is generated through access to many aspects of perception – not only different sensory modalities but also powers of reasoning, abstraction, recollection and rulemaking.  So for example a child when handed an apple for the first time in its life will very rapidly surmise that it is something to be tasted and eaten – not only because a parent might tell it to but also because of its look, size texture, flavour, smell all of which may be reminiscent of other types of fruit it has had or also because of previous experiences of apple given to it perhaps in another form (pureed for example!).   The initial impression will prompt exploratory behaviour such as looking at it and picking it up (instantly conveying colour, size, texture, weight, moistness, stickiness), smelling it (sweetness, appleness) and tasting it (flavour, texture).  All these things, as well as the context in which the learning event takes place, mean that a child needs very few examples of an apple to learn what it is – probably only one or two. Every encounter with an apple in the future will serve to reinforce and evolve the now established knowledge of ‘Apple’.  The way I have described this makes it sound all very deliberate and explicit – but of course it isn’t – it is totally instinctive and taking place at a sub-conscious level.  The only explicit, and perhaps somewhat more laborious aspect of the learning comes in when the parent gives it a verbal tag “would you like to try this apple?” 

By multimodal I am referring to the fact that we learn in many different ways.  We learn through exploration, experimentation and discovery.  We learn through language, stories and sharing of knowledge with others.  We learn through abstraction, association, synthesis and projection.  We learn through categorization, comparison and rulemaking.  No doubt there are other methods as well (I am not an educational expert).   The point is that we use all these methods of learning probably simultaneously and quite often without even realising we are doing it.  By having a multi-modal learning strategy, we become extremely efficient and quick learners.  Anyone currently labouring through the education system is probably thinking that their learning does not feel very quick or efficient – this is probably because our education system is focussed on only one type of learning – which is learning through codified knowledge in the form of language (which in itself comes in many forms – words, numbers, music, images etc.).  Codified knowledge transfer is probably the least efficient form of learning for an individual, it is explicit, laborious and requires effort (and energy).  However codified knowledge is what has given humanity its collective knowledge base- its shared knowledge – its culture – which is not only passed horizontally between individuals but also carried longitudinally between generations and has grown and continues to grow into a thing that is vastly more than the sum of its parts. 

The way AI learns is not efficient.  For an AI algorithm to ‘know’ what an apple is typically requires exposure to thousands, possibly tens of thousands of images for it to ‘recognise’ it in a picture.  For it to be able to express linguistically what an apple ‘is’ it would need access to mountainous quantities of written information about apples – and then to articulate this, access to its ‘large language model’ for it to reproduce a meaningful piece of narrative.  AI can and does do this – on the face of it extremely well.  But in doing so I guarantee it will have burnt through more energy than a toddler will consume in a lifetime of breakfasts.  Even then having produced an immaculate piece of text explaining what an apple is – even illustrating it with a perfect archetype of an apple – does it actually ‘know’ anything?  Isn’t what AI has done simply a synthesis and distillation of pre-existing human knowledge? And then next time you ask AI the same question – will it not just go through the same energy intensive synthetic process to re-create the same piece of narrative?  The reality is that AI has only really got one learning strategy and that is a stochastic probabilistic one – the Bayesian adaptive statistical model. 

4) Empathy

One of the most impressive aspects of human intelligence is its capacity for empathy.  By this I mean its ability to know what another human is thinking and feeling without any obvious or explicit communication.  The level of synergy in the way we think means that we are able if not actually feel then deeply understand, and participate in, other people’s joy, sadness, pain, pleasures, hopes and despair.  Empathy drives a level of common purpose and collaboration to bring about a better world around us that is unparalleled in any other social species.  It can of course also be abused for inflicting great cruelty – but I think it’s fair to say that empathy overwhelmingly drives positive behaviours in society.

Having empathy means we moderate our responses to other people to preserve their feelings and wellbeing.  It also gives us the capacity for humour and irony – as we can give a response that is unexpected knowing that the recipient knows you meant the opposite. 

Empathy allows us to participate in shared experiences such as attending music concerts, plays and sporting events.  It creates powerful societal bonds that underpin the human meta-organism that is human culture.

As far as I can tell empathy exists in AI only as a linguistic concept.  Many AI chatbots can simulate empathy – because the synthesis of their LLMs means they can see that linguistic tokens of empathy pepper our language.  I have to say my experience of AI is that the apparent empathy borders on sycophancy.  In the same way that a salesperson is trained to make you feel good about the purchase choice you have just made – AI is trained to fawn over the questions you ask to incentivise you to go back for more.   There is danger here because the slickness of the language model means that it is very easy to fall for the flattery – but unlike in a human interaction where well-judged flattery induces mutual feelings of wellbeing – in the case of AI it is there for a purpose and any feelings induced are very much one way.

I tried a little experiment (Alert – If you are a ‘believer’ stop reading and jump to the next section!).  If someone comes up to you and asks, ‘Does Father Christmas exist?’ how would you respond?  My guess is that you wouldn’t just come straight out with an answer – you might take some time to assess your interrogator and try and work out what the motive for the question is.  Based on your assessment you have a range of answering strategies from brutal truth to empathic lies – with plenty of scope for irony in between.  Now try and ask ChatGPT – what do you get?  The somewhat pedestrian and brutally truthful response is about as far from empathic as it is possible to get.

5) Opposable Thumbs

As I observed earlier in this essay our intelligence is contained entirely within us – it is a fundamental part of our biology.  We are free to move around our environment carrying our intelligence with us – in fact using it continuously to remain safe within that environment.  Not only that but we are fortunate to have evolved alongside our oversized brain high acuity visual perception, high fidelity auditory perception and incredibly dextrous upper limbs and hands.  This turns us into the ‘maker’ species – we can use our mechanical intelligence to turn materials in our environment into tools that dramatically extend our dextrous capabilities to a phenomenal extent.  The technologies that exist today stand as testament to the astonishing creativity and inventiveness of humanity – the depth and breadth of its intelligence which has given us tools that can take us to new planets, tools with which we can manipulate our own biology, the means to fight and eliminate disease, invent new foodstuffs and – yes- tools which mimic our very own intelligence.

AI is amazing – but it is not yet a maker intelligence.  It remains bound to its datacentres and server racks.  Yes you can attach robotic arms and other tools to AI controllers and tell it to make stuff – but it is not yet free to roam the environment and invent things for itself – that for now remains the stuff of sci-fi fantasy.

6) Imagination

One of the most confounding things about the human brain is the size of the frontal cortex.  This very large part of the brain doesn’t appear to have much of a biological function – yet contains a vast number of interconnected neurones, firing apparently randomly, and consuming a significant part of the metabolic energy dedicated to the brain.  What is this ‘expensive’ organ actually doing?  It has often been described that the number of possible permutations of neural connections within a human cortex could exceed the total number of atoms in the universe.  Whether this is true or not in reality is immaterial – the fact is that we carry in our heads a neural network of profound depth and interconnectedness the like of which has yet to be even closely emulated in electronic form.  It is impossible to fully articulate the cognitive powers this gives us – a sign if anything of the limitations of codified knowledge.   It is the frontal cortex that give us our powers of abstraction and projection – it not only helps us to process and understand the world around us, it takes that understanding reforms it and projects a potential new world which then drives us to behave in a way to create that imagined world.  The frontal cortex is the seat of our imagination and creativity – it is fundamentally what makes us human.  

Notwithstanding the expense of this organ in biological terms in neural network computing terms it is absurdly efficient.  At a physical level it is a six layer neural network – on the face of it eminently emulate-able.  However the cortex has properties that make it quite unlike computer neural nets.  The first of these is its organisation into columns – of which there are millions – each column functions effectively as an independent six layer network all working in parallel.  Through this architecture the cortex massively reduces the trade-off of signal distance experienced with the increasing layer depth of artificial neural networks and massively increases effective network depth.  The second property that differs in the human cortex is the computing method – which is reliant on recurrence and signal looping.  Through this the brain is not only physically parallel but also chronologically parallel – signals can reinforce and coalesce over time.  The final property that makes the brain a truly supercomputer is plasticity – the actual physical structure of the brain changes with use – reinforcing ‘useful’ pathways and allowing redundant ones to fade away.  This gives rise to learning from new experience, adaptation to new environments, the ability to recover from damage and the ability to specialise over time.  The frontal cortex is truly our superpower.

7) Lifecycle

Time is both our friend and our enemy.  Whilst the human brain can use time to massively amplify parallelism and network depth and thus the power of our intelligence – it also casts a shadow over our lived experience and shapes our intellectual purpose because, of course, our time is limited.  As we move through the stages of a human life cycle our purpose changes and the way we think changes – the brain changes with it, both in form and function.  That it does so is testament to the adaptability of our intelligence – as purpose changes so does our brain.  The table below gives a view (my view perhaps) of how purpose changes through life and how the brain adapts:

Stage	Purpose	Brain Adaptation
Infancy	Discovery of the environment, basic motor skills, language learning, creative introduction	Cortex is hyper-connected and undifferentiated.  Noisy activity, new pathways rapidly forming
Childhood	Discovery of self, early socialisation, primary codified knowledge, creative choice	Consolidation of basic pathways. Large parts remain undifferentiated.
Adolescence	Discovery of others, partner seeking, secondary codified knowledge, career identification, creative pursuit	Rapid development of social and sexual functions of the brain.  Lifelong patterns start to become set but large parts remain developmental, pruning of redundant pathways
Early Adulthood	Discovery of society, parenting, career development, specialised codified knowledge, creative development	Specialist functions laid down, parenting skills learnt (rapidly!), continued pruning
Middle Age	Influencing Society, post-parenting, career achievement / second career pursuit, creative fulfilment	Specialised knowledge and skills optimised and become hardwired (improved speed and efficiency); plasticity affords opportunity for new skills.
Older Age	Sharing experience and wisdom, grandparenting, career conclusion, new creative opportunity	Early decline, plasticity allows adaptation to decline and damage, behaviour changes fend off cognitive inefficiency
Senescence	Reflection and legacy, deep memory mining, adaptation to loss of autonomy,	Cognitive decline disinhibits deep memory suppression

Needless to say this table makes some gross generalisations, it is very high level – and there is nothing if not considerable overlap between the supposed ‘seven ages of man’.  The point is though the brain both shapes and adapts to our biological needs over time – at every stage it is optimised for the purposes we have at that time.  Even in disease its phenomenal adaptability moderates the worst impacts.  The mechanical nature of AI affords it no such journey.

Conclusion

Artificial intelligence is an extraordinary technological achievement and will continue to reshape how we work, create and solve problems. But its power should not be confused with equivalence. Human intelligence is not merely a faster or slower computational engine — it is a biologically embodied, energetically efficient, purposive and empathetic system that has evolved to learn, adapt, imagine and collaborate across an entire lifetime.

AI has no intrinsic purpose, no lived experience, no developmental journey and no stake in its own existence. Its intelligence is derivative, conditional and dependent on human intent — both for its creation and its application. In contrast, human intelligence is self-directing, self-modifying and embedded within a social and cultural super-organism that amplifies its power far beyond any individual mind.

For these reasons, AI should be understood not as a successor to human intelligence but as one of its most remarkable tools — a product of the very creativity, imagination and collaborative purpose that define us as a species. If we remain clear-eyed about that distinction, there are strong grounds not for fear, but for confidence in the enduring superiority of the human mind.