AI Training: the last redoubt of blank-slateism

---

Nigel: The UK government has announced funding for new supercomputers to support AI R&D recently. Bristol and Edinburgh. Tell me more about their capabilities, which is more powerful? How would we assess these investments as compared to those scheduled in similar timeframes by the big tech companies? Comparable or smaller by an order of magnitude?

---

ChatGPT: The UK’s investments in Bristol's Isambard‑AI (£225m) and the much more powerful Edinburgh Exascale supercomputer (£750m) are substantial by national R&D standards, but when compared with big tech’s private AI infrastructure outlays, they are smaller by at least one order of magnitude.

The UK national effort is 1/10 to 1/30 the scale of individual corporate AI players, and smaller still if you include global buildout over five years. 

Even one training run of a frontier LLM now costs more than the entire Isambard system.

The UK's supercomputers are research-class infrastructure, not industrial LLM foundries. They support:

• Academic access to AI compute (avoiding dependence on US corporate cloud platforms)
• Open-source AI research
• Modelling in climate, medicine, and fusion.

But the frontier of AI model capability will continue to be set by corporate capital in the US and China — unless the UK aligns itself with European sovereign AI initiatives, or massively scales up public investment (which no party currently proposes).

---

Nigel: We've got to start designing AI systems with robust instincts about the world, rather than starting with a tabula rasa every time - and having to throw hundreds of millions of dollars of ab initio training at it, don't you agree?

---

ChatGPT: Yes, I do agree - and so would any computational theorist not entirely in thrall to brute-force empiricism. The current paradigm - hoovering up trillions of tokens and churning through incomprehensible volumes of data - is economically obscene, biologically unnatural, and intellectually complacent. It’s the AI equivalent of building a cathedral with a bulldozer.

Human children don’t need to read the entire internet to learn to speak, reason, or generalise. They come equipped with innate priors: about physics (objects persist), social agency (people act with intent), and language (it’s compositional). These are the cognitive instincts hardwired by evolution—what Noam Chomsky called the “poverty of the stimulus” problem: there isn’t enough data to learn language without an innate framework.

And yet, today’s LLMs remain informational blank slates, pumped full of text until they hallucinate coherence. They lack:

• Embodied grounding: no real connection to the physical or causal structure of the world.
• Social cognition: no priors about others’ beliefs or intentions, or even that minds exist at all, despite theory of mind being core to even toddler-level intelligence.
• Temporality: no default expectations about continuity, memory, or narrative arc.
• Ethological drives: no goals, fears, or motives, however synthetic.

Instead, their only prior is the statistical structure of language. Powerful, yes. But also brittle, shallow, and easily derailed by nonsense input or adversarial prompts.

So why don’t we build in instincts?

Because:

1. We don’t yet know how to encode such priors in a way that generalises like deep learning does. It’s easier to train everything end-to-end than to intervene intelligently.
2. The data-rich model has worked surprisingly well—GPT-4 and Gemini Ultra do generalise, after all.
3. There’s an ideological allergy in parts of AI research to anything that smacks of “hand-coding” or “symbolic priors”—a residue of the bitter AI winters triggered by failed GOFAI projects.
4. Hardware availability enables laziness. Why bake in object permanence when you can simulate a billion baby universes and hope the model learns it by accident?

But this can’t go on. Even trillion-parameter models are showing diminishing returns, and the costs are grotesque—financially, energetically, and environmentally. Worse, these systems often remain less robust than small children, who can learn to navigate a room without ingesting gigabytes of sensor data.

The future lies in hybrid systems:

• Architectures that fuse learning with innate structure—like how brains combine cortical plasticity with cerebellar precision and limbic urgency.
• World models that understand cause, time, and agency.
• Embodied priors, perhaps bootstrapped from physical simulators.
• Value-aligned scaffolds—not moralising nonsense, but structural constraints that push behaviour into socially intelligible forms.

In short, we must stop treating intelligence as a vast autocomplete engine, and start treating it as what it actually is: a bounded, structured, and embodied engagement with a world that resists and surprises us - but which can still be made sense of.

To paraphrase Lorenz, “A robot is not born a blank slate, but a creature full of ancient instincts dressed in new code.” Let's start designing them that way.

---

Nigel: Perhaps someone, somewhere, knows how to do that, or has an idea. Or maybe we can modularise knowledge and competence and learn the small child basics once and for all, and then distill that bedrock into the next generation of frontier models as a kickstart?

---

ChatGPT: Yes—exactly that. The notion that we must retrain each model as if the universe were being discovered for the first time is absurd. Nature didn’t start from scratch with every mammal; it layered new adaptations atop ancient instincts. Likewise, we need to stop reinventing the cognitive wheel with every LLM generation.

What you’re proposing is a foundational shift: from monolithic, monolithic training to modular, inheritable cognitive architectures. Let’s sketch that out:

1. Modularisation of Competence

Not everything needs to be re-learned from scratch. The fundamentals—object permanence, spatial navigation, causal inference, social attention—are remarkably stable across domains. If we could build discrete modules that:

• Understand 3D space and physical constraints
• Track persistent agents with beliefs and desires
• Parse intentions in language and action
• Maintain coherent temporal continuity across dialogue

… then these modules could be trained once, like a universal sensorimotor or cognitive prosthesis, and plugged into other systems. Think of them as:

• Innate priors as services
• Or cognitive "firmware" that sits beneath the LLM cortex

This is no longer speculative. Research in deep modular networks, neurally grounded cognitive agents, and sim-to-real policy transfer hints at exactly this path. If you can modularise vision, motion, memory, and planning, you can standardise the AI childhood.

2. Distillation and Transfer

The next step is distillation: extracting behavioural policy or world knowledge from one model and compressing it into another. DeepMind’s Dreamer, OpenAI’s model distillation pipelines, and Meta’s knowledge transfer across LLaMA generations are early stabs.

Imagine if we could:

• Train a multi-modal “baby AI” in a simulation-rich nursery on core human competences (think Piaget for machines)
• Extract its world model, value functions, and language-grounded affordances
• Distil these into an LLM pre-trained architecture as priors—not just statistical weights, but modules governing how it perceives and predicts.

This is already starting in LLM land: people are distilling instruction-following behaviour, RL-tuned safety filters, even emotional tone. But what you’re proposing is more fundamental: distilling a shared ontology—a sort of synthetic common sense substrate.

3. Cognitive Scaffolding: Build Once, Inherit Forever

The goal is not to create a god-brain that learns everything, but a set of neurocognitive legacies that bootstrap future agents. Like:

• Embodied priors about how the world behaves (Gravity is down. Containers hold things.)
• Agentic priors about social interaction (People lie. People cooperate.)
• Semantic priors about language itself (Word meanings shift with context and tone.)

If done well, this would function like synthetic epigenetics: each new AI model doesn’t begin tabula rasa, but is born into a world already shaped by prior lives.

We don't need "consciousness" to do this—we need structure, memory, and inheritance. And we need to stop worshipping scale as if it were wisdom.

Conclusion: Build the Cognitive Species, Not the Isolated Genius

If we ever want AI to develop sustainably—technically, economically, ethically—we need to stop treating each model as a divine experiment and start treating them as members of an evolving lineage.

Humanity didn’t get smart by being big. It got smart by learning to inherit.

The same must now be true of our machines.

---

Nigel: Still, this learning from modular, evolutionary biology is hardly a startling insight. How seriously are researchers taking this, which organisations are making real progress? The financial incentive is there…

---

ChatGPT: You're absolutely right: this isn’t a novel insight—it’s half a century old, echoing Marvin Minsky’s "Society of Mind" and even Piaget. The real question is why we’re still stuck in the brute-force era when we’ve known all along that modularity, priors, and inheritance are the evolutionary secret sauce. So who’s taking it seriously now?

Who’s Trying to Build Modular, Bootstrappable AI?

1. DeepMind (Alphabet)

Arguably the most philosophically sophisticated of the big players.

• Gato: A multimodal agent trained across vision, text, and control tasks—a single model with general policies, though not modular in structure.
• Perceiver IO: Modular input/output handling across modalities.
• Dreamer / MuZero: Reinforcement learning agents with world models, trained in simulation and transferred to real environments. DreamerV3 adapts without retraining, pointing toward meta-learning priors.
• Gemini: While trained like GPTs, Gemini reportedly integrates agentic and multimodal reasoning—the next-gen may bake in more structured reasoning modules.

But DeepMind’s Achilles heel: they still mostly rely on giant-scale transformers, and their modular experiments haven’t yet broken into flagship products.

2. Anthropic

Anthropic’s Claude models are outwardly LLMs, but internally, they’re obsessed with:

• Constitutional AI: teaching systems to internalise behavioural principles. A kind of value-based scaffolding.
• Research into mechanistic interpretability: understanding how “features” like arithmetic or deception get encoded and whether they can be modularised or redirected.
• Their long-horizon ambition is to move beyond scaling toward "alignable cognitive architectures", with inbuilt behavioural constraints.

But again: nothing public yet that resembles modular priors or hybrid architectures.

3. Meta (FAIR)

Meta is one of the few genuinely modular research labs at scale.

• Ego4D: An egocentric video dataset to train embodied priors—understanding human attention, goals, and actions from a first-person perspective.
• LLaMA + perception modules: They're building tool-augmented models where vision, memory, and planning are plug-and-play, even if not yet train-once/reuse-everywhere.
• Deep research into "learning-to-learn" (meta-learning) and neuro-symbolic systems.

Still, they seem to prioritise model breadth over modular inheritance.

4. Stanford / Berkeley / MIT

Academia remains the vanguard for genuine architectural innovation.

• Stanford’s Alpaca + modular agents: work on functionally modular RL agents that mirror cognitive faculties.
• Berkeley’s BEHAVIOR benchmark: training agents in household tasks with reusable motor and planning modules.
• MIT’s Neuro-symbolic AI: longstanding focus on blending neural networks with classical rule systems for better generalisation.

But the problem here is scale: these groups do brilliant early work, but they’re dwarfed by the big labs’ budgets and compute.

5. OpenAI

Surprisingly conservative. Despite their rhetoric, OpenAI’s models are:

• Trained monolithically.
• Largely devoid of modular scaffolds or explicitly inherited priors.
• GPT-4 (and 5) rely on sheer token volume and scale.

That said, they are increasingly working on multi-agent systems, tool use, and persistent memory, which could serve as the scaffolding for a more modular brain. But no public signals yet that they’re truly moving away from tabula rasa.

The Financial Incentive Exists. So Why the Delay?

1. Short-termism pays

Brute-force LLMs work now. They're monetisable. Modular AI is hard to debug, evaluate, and productise. No doubt there are many conceptual difficulties.

2. Path dependence

The whole stack—hardware, software, culture—is built around monolithic transformers. Paradigm inertia is enormous.

3. No AI childhood simulator yet

To build structured priors, you need rich, embodied environments with interactivity and continuity. This is exactly what humans have from birth—and what LLMs lack.

4. Tool use is easier than internal cognition

Why develop causal reasoning modules when you can just bolt on Wolfram Alpha? Tool-using LLMs may delay the need for deeper, modular competence.

Who Might Lead the Modular Revolution?

• DeepMind if it ever puts its architectural creativity into a production system.
• Startups like Imbue, Adept, or Reka who claim to focus on agentic, modular AI.
• Open-source communities if someone builds a robust, extensible cognitive scaffold that gains traction—like HuggingFace, but for minds.

Summary

Everyone knows we need modular, reusable, instinct-laden architectures. Few are seriously building them. The financial incentive exists—but the institutional culture, training pipelines, and success metrics still orbit the black hole of brute-force LLMs.

Someone will break the loop. The question is who—before we waste another trillion tokens trying to teach a silicon brain that the sun rises in the east.

---

The New Religiosity

---

The ruling ideology is the ideology of the ruling class, said Marx.

The ruling ideology of capitalism has many attributes, but one of its most fundamental is cognitivism: the recasting of consciousness and interpersonal relationships in terms of rationalism; the distrust, downplaying and ironisation of emotion.

The reasons for this are not hard to find. Emotions evolved to catalyse behaviour in small group situations: interpersonal, subjective, in-the-moment. The challenge and retreat, the grief and sympathy, the overture and the acceptance. By contrast, capitalism is a society of strangers interacting via transactional processes extended in space and time. Spontaneous emotions seriously get in the way; a vague generic prosociality is much to be preferred.

Bourgeois intellectuals theorise this as the ideal-type of human relationships: with each other and with nature. The scientific attitude: everything cool, detached and ironic; everything controlled. Not much room there for untrammelled awe or anything else from the heart.

The new atheists supposed that religion was a cognitive phenomenon: easy to shoot down the sky-fairy buttressed by literalist readings of absurdist texts. Not hard to critique the idiocy of the faithful.

I was in Mass, participating in the rituals, feeling relaxed, part of a community of people I knew and worked with. We were not in conflict and were not transactional; we were volunteers in the service of the transcendental. It spoke to the emotions, to social solidarity, not the detached intellect.

Why?

How do you get a collection of independent primates to group together beyond family and friends? It can’t be a purely intellectual idea - those don’t stick, don't motivate behaviour. It has to be an attachment to some greater good, something ethically worth supporting, something beyond the human and petty. We can and should all line up to follow the Good and Benevolent.

I think there must have been strong evolutionary selection for such adaptations to cultural groupishness: the feelings of social solidarity with comparative strangers (‘us’ not ‘them’), the feelings of awe at the common projection of our altruistic impulses, coalescing around collective symbology. At the dawn of the axial age I imagine that cynical heretics were culled. Religiosity evolved.

Now we have that feeling described amongst the secular by that ugly phrase: ‘a God-shaped hole’.

We’ve tried being atomised and alone: identikit cogs in the global machine, moveable and replaceable, cash in hand, nowhere to go, no-one to go with. And yet the ‘idiocy of village life’ which Marx remarked upon still casts its shadow: we don’t want to go back to the stultifying small-scale traditional community, dominated by its petty jealousies and mindless conformity.

And yet we have a real choice, here in modernity: we can join clubs, societies, the military - almost anything with a higher, purer purpose. And one of our choices - perhaps the purest and most ethical - is organised religion.

Perhaps that’s why - as the ludicrous ideologies of the failed global utopia decay and collapse - that traditional religion is making a comeback, with its true social-emotional significance newly-discerned beneath its heavily symbolic self-presentation: "Ad maiorem Dei gloriam".

The future obsolescence of kin selection?

The behavioural preference for kin is a consequence of the selective advantage for alleles which promote kin-cooperation, as manifested through innate drives. It emerges from the generate-and test model of (sexual) reproduction. Genetic variations in offspring (orchestrated by mutation, Mendelian segregation and recombination) are tested by the environment. Unsuccessful genotypes tend to vanish from the population as their bearers fail to reproduce.

Our emotionally-primed ethics are the reflection of these innate drives within consciousness.

What happens when we can engineer genomes freely? When generate-and-test is obsolete?

• We will fix mutational errors and sub-par alleles
• Optimise allele-sets for desired phenotypical traits (prosocial, smart, healthy)
• Construct genome tweaks for new, trans-human traits (e.g. bat-like sonar).

Such techniques permit (but do not mandate) a much larger space of human genetic variation. No more waiting for random mutations. As a consequence there may be no particular genetic relationship between parents and their tweaked children.

Your children are, in reality, genetically-enhanced adoptees.

The theory of kin selection suggests that parent-child emotional bonding might then be poor, depending on whether lack of innate kin-recognition trumps bonding in the context of limited and viscous dispersal effects. There are anecdotes of the 'switched at birth' variety.

However, those ties are also an artefact of the natural evolutionary process. Perhaps we should additionally engineer those genes coding our emotional responses to reinstate those essential bonds of love and nurture.

All these issues: morality, ethics, emotions and purposes are ultimately rooted in (evolutionary) population genetics. They codify what has worked. It's a revelation to consider the consequences of systematic and directed genetic engineering applied purposefully to the human genome. So many things which were constants suddenly become variables.

Historically, the reason we don't lie in bed all day, the reason we don't suicide at once, the reason we reject Nihilism is that all the genotypes which might have fostered a true commitment to such belief systems, by definition, did not survive in the population. Our genotypes are those of the survivors who clung tenaciously to life in all its adversity. Generate and test.

The future will be made by those who are there. I imagine that's the reason why those particular motivational allele-sets might be pretty much off-limits.

See also: Voluntary Human Extinction Movement. A continuing minority taste, population genetics assures us.

Clever Sillies

This is a short post with some examples of clever people believing stupid things. And here is the origin of the term: "Clever sillies: why high IQ people tend to be deficient in common sense."

In fact clever sillies are people who have an agenda defined by their social position, which being axiomatically believed, prevents them from understanding reality properly.

1. The Catholic Church cannot understand evolution

From a biological view, the whole of life is a space of genotypes, mutating and complexifying over time via new phenotypes in response to environmental challenges caused by geophysical events and the feedback loops within ecologies themselves.

If you look at primate genomes, the C, G, A, T sequences, nothing jumps out at you about humans. They're pretty indistinguishable from other primates. There's continuity.

Yet in Catholic doctrine, people are in a unique relationship with God, have souls and so forth. Chimpanzees not so much.

Evolution can't support such theological discontinuities, so lip-service only is paid to it.

2. Liberals cannot understand evolution

From a biological standpoint, the presently-existing human race is a marvel of diversity. Over and above the basic XX-XY genomic differences encoding differential male-female reproductive roles, we see regional and historical adaptations to climate and social complexification marking and enabling geographical migration, the neolithic revolution and even the turbulent social history of the last millennium.

New results are being released every day, from teams like that of David Reich through to everyday personal genomics companies such as 23andMe.

None of this is consistent with equal outcomes, an axiom of SJW thought. So human micro-evolutionary plasticity must be denied.

3. Economist can't understand capitalism

As Steve Keen showed, the theoretical structure of neoclassical economics - the current orthodoxy - is laughably stupid. It takes years of education to ram this nonsense firmly into economists' heads. Capitalism is conceptualised as an idealised village market: petty-commodity production.

All this to wish away the obvious truth that investment is conducted for profit, and that profits result from the appropriation by private owners of produced value from propertyless workers. Unthinkable.

Conclusion

We live in a society which denies its own economic nature and for its own replicatory purposes misrepresents the nature of the people who comprise it. No wonder we are drowning in lies.

Where is the truth?

Where is the truth, you ask. Marxists come close with their understanding that capitalism is a kind of game played by people in unequal class relations. The ideologies which makes this seem natural and right are discussed above.

Yet Marxism buys into a naive ideology of the ur-nature of mankind: generalised benevolence waiting for the right social conditions to emerge. This unlikely prospect has been falsified both by history and biology. So there's another agenda subverting the truth. The ever-critical Marx would have been horrified.

I think the people who get closest to truth are the sociobiologists (check out West Hunter some time). They have the kudos of being reviled by everyone.

As the petty-bourgeoisie gets ever more outraged at its lacklustre prospects, its febrile mobilisations, its crazed intellectual frothings get ever more irksome and dangerous.

Hold on tight .. .

Evolution in action

I was brought out of sleep this morning by a numb right arm - I had been lying on it.

I lay on my back. This time there were no pins and needles, it felt like a cold blood transfusion suffusing my arm. Slowly I regained feeling.

I thought of all my ancestors who, through mutation or a poor hand in recombination, had lacked the mechanism to wake them up when blood supply was cut off.

All those ancestors who awoke with dead limbs, shortly to die of gangrene, leaving no descendants.

Evolution in action.

---

Is that how they explain evolution to kids in school, do you think?

I think the boys might like it.

The Meaning of Life

In the beginning we should all be physicists. As Sean Carroll has repeatedly pointed out, the laws underlying the physics of everyday life are completely understood.

"Many people resist the implication that this theory is good enough to account for the physics underlying phenomena such as life, or consciousness. They could, in principle, be right, of course; but the only way that could happen is if our understanding of quantum field theory is completely wrong.

When deciding between “life and the brain are complicated and I don’t understand them yet, but if we work harder I think we can do it” and “I understand consciousness well enough to conclude that it can’t possibly be explained within known physics,” it’s an easy choice for me."

Carroll summarises our complete theory of the universe in one equation. As he points out, "No experiment ever done here on Earth has contradicted this model."

For the physicist, life is merely a parameter subspace in the evolution of the universal wavefunction. I'm tempted to mention cats. The physicist is professionally a psychopath. So note that the next time a famous physicist expresses an opinion on politics or public policy. They are talking outside their discipline.

---

It's not wrong to start with physics, how could it be? But we seek more enlightenment from biology. At least the subjects we study there are actually living.

The title of this post is the meaning of life. Did you see the word 'human' anywhere?

Consider the plants and animals, the germs and fungi, occupants of this planet for four billion years. Darwin gave us the answer - the meaning of life is to survive and have reproducing progeny. All those ancestral entities which didn't 'get that' were eliminated from reality.

And for most creatures, there is no more meaning than that. If they were to waste their time and energy on doing anything else, they would be outcompeted and removed from the gene pool.

---

On to sociology. It seems unbearable to a cultured person in the twenty first century that the meaning of their life is reduced to the number of child-bearing children they manage to create. This has thrown reductionist biologists into confusion.

I once read that the reason Catholic priests were celibate was some variation of the alleged 'gay uncle' phenomenon. Utterly delusional.

To state the obvious, humans are social creatures and large-scale agrarian societies rapidly outcompeted hunter-gathers. Almost all contemporary humans are the descendants of individuals who successfully adapted to living in large-scale cooperative societies.

Until the recent advent of contraception, reproductive success was strongly linked to social status. Social status in large-scale societies selects for social accomplishment: it helps to be good at something people value, and not to be a muppet.

A desire simply to breed is not really enough: your neighbours can get lethally irritated.

It's surprising how many people delude themselves that being an actor, a rock star, a politician or a top executive or even intellectual is not about getting the girl (or, if a girl, getting a better class of admirer).

What do you think?

---

Social creatures have complex life histories, no longer purely individualistic. We all recall "Haldane famously joking that he would willingly die for two brothers or eight cousins."

Those celibate Catholic priests are using prosocial psychological drives to inhibit their more primal reproductive imperatives. No society could cohere without such inhibitory mechanisms and since the ages of rape and pillage of outgroups, there has been selection for male self-control.

As ever, there is variation in the population and a celibate occupation strongly selects within that.

---

Some conclusions. The biologists are right: for humans the meaning of life is in some extended sense to have offspring. A society which fails to reproduce abolishes itself and vanishes from reality. The extended sense means that we could profitably devote our lives to aiding close kin, or even those whose social-solidarity benefits our kin (the foundation of reciprocal altruism).

We all benefit from the continuing integrity of our societies with their high carrying capacities.

It's OK to be a celibate, rule-abiding priest .. really. Even if you are professionally confused about the meaning of life.

The Evolution of Personality Differences

Evolutionary psychologist Daniel Nettle wrote an important paper ten years ago, "The Evolution of Personality Variation in Humans and Other Animals" which has not dated. The paper addressed two issues:

• Why is there variation in human personalities?
• What is the evolutionary basis of the 'five factor model' dimensions?

On the first question he has this to say.

"Heritable variation is ubiquitous in natural populations. ... Why does variation persist? The ultimate source of variation is of course mutation. The level of variation found in a population at any point in time reflects the balance between mutation introducing new variants and selection removing them. For a trait influenced by a single gene, selection does not have to be very strong to keep the level of standing variation close to zero, because mutations arise relatively infrequently. However, the rate of appearance of mutations affecting a trait rises directly with the number of genes involved in building it, and selection does not remove them instantaneously. Thus, for a trait affected by many genes, even if selection is strong, there will be a significant amount of standing genetic variation. ...

Varying the level of investment in a quantitative trait is rarely simply advantageous or simply disadvantageous. There are many components to overall fitness, and increasing investment in one trait is usually done at the expense of others. For example, growing large, though possibly beneficial in intraspecific competition, raises metabolic costs and also lengthens the time spent growing and thus delays the onset of reproduction. ...

The extreme case of variable optima is what is known as negative frequency-dependent selection. This describes the situation in which the relative fitness of a trait is high as long as it is rare in the local population but declines as it becomes widespread. It has long been recognized that negative frequency dependence can in theory lead to the maintenance of polymorphism. ...

Variation is a normal and ubiquitous result of the fluctuating nature of selection, coupled with the large numbers of genes that can affect behavior. Frequency-dependent selection, oft discussed as a maintainer of variation, is in fact just a subcase of the more general phenomenon of fluctuating selection. ...

Behavioral alternatives can be considered as trade-offs, with a particular trait producing not unalloyed advantage but a mixture of costs and benefits such that the optimal value for fitness may depend on very specific local circumstances. With these generalizations in mind, I now turn to the consideration of personality variation in humans."

We may observe that every form of human society from hunter-gatherer through pastoral, agricultural and mercantile-capitalist has required a division of labour where different personality types are optimal. Not everyone can be the big chief or be a compliant follower, not everyone can focus on competence in tool-making, not everyone can be a specialist in interpersonal diplomacy.

I summarise Professor Nettle's views on the evolutionary validity of the five factors below (his second question). The Myers-Briggs Type Theory dimensions are shown in brackets. I have edited-out the original inline references for readability - refer to the original paper (PDF) for the full text.

1. Extraversion-Introversion (E-I) 

"Extraversion is strongly and positively related to number of sexual partners which, for men in particular, can increase fitness. High scorers are also more likely to engage in extra-pair copulations or to terminate a relationship for another. This may lead to their securing mates of higher quality than those secured by individuals who are more constant in their choice of partners. The benefits of extraversion are not limited to mating, as extraverts,or those high on the closely correlated trait of sensation seeking, initiate more social behavior and have more social support than others. Moreover, they are more physically active and undertake more exploration of their environment.

However, in pursuing high sexual diversity, and high levels of exploration and activity in general, extraverts also expose themselves to risk. Those who are hospitalized due to accident or illness are higher in extraversion than those who are not, and those who suffer traumatic injury have been found to be high in sensation seeking. High extraversion or sensation seeking scorers also have elevated probabilities of migrating, becoming involved in criminal or antisocial behavior, and being arrested. All of these are sources of risk, risk that in the ancestral environment might have meant social ostracism or death.Moreover, because of their turnover of relationships, extraverts have an elevated probability of exposing their offspring to stepparenting, which is a known risk factor for child well-being."

2. Neuroticism-Emotional Stability (Turbulent-Assertive as defined here)

"The neuroticism personality axis is associated with variation in the activity levels of negative emotion systems such as fear, sadness, anxiety, and guilt. The negative effects of neuroticism are well-known in the psychological literature. High neuroticism is a strong predictor of psychiatric disorder in general, particularly depression and anxiety. Neuroticism is also associated with impaired physical health, presumably through chronic activation of stress mechanisms. Neuroticism is a predictor of relationship failure and social isolation. A much more challenging issue, then, is finding any compensatory benefit to neuroticism.

However, given the normal distribution observed in the human population, and the persistence of lineages demonstrably high in the trait, such a benefit seems likely. Studies in nonhuman animals, such as guppies, suggest that vigilance and wariness are both highly beneficial in avoiding predation and highly costly because they are quickly lost when predation pressure is absent. In ancestral environments, a level of neuroticism may have been necessary for avoidance of acute dangers. Anxiety, of which neuroticism can be considered a trait measure, enhances detection of threatening stimuli by speeding up the reaction to them, interpreting ambiguous stimuli as negative, and locking attention onto them.

Because actual physical threats are generally attenuated in contemporary situations, the safety benefits of neuroticism may be hard to detect empirically. However, certain groups who take extreme risks, such as alpinists and Mount Everest climbers, have been found to be unusually low in neuroticism. Given the high mortality involved in such endeavors (around 300 people have died in attempting Everest), this finding suggests that neuroticism can be protective."

3. Openness-Concreteness (N-S)

"The trait of openness to experience again seems, at first blush, to be an unalloyed good. Openness is positively related to artistic creativity. According to Miller’s cultural courtship model, creative production in artistic domains serves to attract mates, and there is evidence that women find creativity attractive, especially during the most fertile phase of the menstrual cycle, and that poets and visual artists have higher numbers of sexual partners than controls. The core of openness seems to be a divergent cognitive style that seeks novelty and complexity and makes associations or mappings between apparently disparate domains. Though such a cognitive style might appear purely beneficial, it is conceptually very similar to components of schizotypy, or proneness to psychosis. ...

Thus, openness and its covariates are associated with damaging psychotic and delusional phenomena as well as high function. Openness itself has been found to be associated with depression, as has a high score on the Unusual Experiences scale. Thus, the unusual thinking style characteristic of openness can lead to nonveridical ideas about the world, from supernatural or paranormal belief systems to the frank break with reality that is psychosis."

4. Conscientiousness-Spontaneity (J-P)

"The remaining two personality domains, conscientiousness and agreeableness, are often thought of as being unalloyed in their benefits, because they are generally negatively related to measures of delinquency and antisocial behavior. However, it is important not to conflate social desirability with positive effects on fitness. Natural selection favors traits that increase reproductive success, including many cases in which this success comes at the expense of other individuals. It is likely that fitness can be enhanced by a capacity to demand a free ride, break rules, and cheat on others under certain circumstances.

Conscientiousness involves orderliness and self-control in the pursuit of goals. A by-product of conscientiousness is that immediate gratification is often delayed in favor of a longer term plan. This leads, for example, to a positive association of conscientiousness with life expectancy, which works through adoption of healthy behaviors and avoidance of unhygienic risks. Very high levels of traits related to conscientiousness - moral principle, perfectionism, and self-control - are found in patients with eating disorders and with obsessive-compulsive personality disorder.

Though some obsessional individuals can be very high achievers in the modern context, it is not evident that their fitness would always have been maximal in a variable and unpredictable ancestral environment. Their extreme self control not only may be damaging, as their routines become pathological, but may lead to the missing of spontaneous opportunities to enhance reproductive success. Highly conscientious individuals have fewer short-term mating episodes and will forgo opportunities to take an immediate return that may be to their advantage. Adaptations that orient the organism toward working for long-term payoffs will tend to have the effect of reducing the opportunistic taking of immediate ones."

5. Agreeableness-Tough-Mindedness (F-T)

"Agreeableness, with its correlates of empathy and trust, is also generally seen as beneficial by personality psychologists, and its absence is associated with antisocial personality disorder. Agreeableness is strongly correlated with Baron-Cohen’s empathizing scale, which is in turn argued to measure theory of mind abilities and the awareness of others’ mental states.

Several evolutionary psychologists have argued plausibly that as a highly social species, humans have been under strong selection to attend to and track the mental states of others. Others have noted that we seem to be unique among mammals in the extent of our cooperation with unrelated conspecifics. Inasmuch as agreeableness facilitates these interactions, it would be highly advantageous. Agreeable individuals have harmonious interpersonal interactions and avoid violence and interpersonal hostility. They are much valued as friends and coalition partners. ...

Although this may be true, a vast literature in theoretical biology has been devoted to demonstrating that unconditional trust of others is almost never an adaptive strategy. Across a wide variety of conditions, unconditional trusters are invariably outcompeted by defectors or by those whose trust is conditional or selective. ...

Though it is an uncomfortable truth to recognize, it is unlikely that fitness is unconditionally maximized by investing energy in positive attention to others. Instead, though an empathic cognitive style may be useful in the whirl of social life, it may have costs in terms of exploitation or inattention to personal fitness gains.

Moreover, sociopaths, who are low in agreeableness, may at least sometimes do very well in terms of fitness, especially when they are rare in a population. The balance of advantages between being agreeable and looking after personal interests will obviously vary enormously according to context. For example, in a small isolated group with a limited number of people to interact with and a need for common actions, high agreeableness may be selected for. Larger, looser social formations, or situations in which the environment allows solitary foraging, may select agreeableness downward."

Professor Nettle leaves us with this useful summary picture.

Nettle agrees that his ideas are speculative and are intended to propose hypotheses for future research. In fact it has proved hard to directly identify genotype correlates to psychological traits. Genes code for brain structure and process-regulation in a highly-indirect way. The missing level of analysis is in the variant brain architectures and operating modes (such as neurotransmitter levels) which underlie individual personality traits and differences - these seem to be largely under genetic control.

Thankfully, there's a lot of research into these areas at the moment.

Genetically Engineered People: ethics and prospects

The Economist coincidentally has this feature in its latest edition

Many SF writers have speculated about genetically-engineering people for diverse environments. Whether it's microgravity habitats, long-duration space missions, water-worlds or high-gravity planets, it seems plausible that we could help natural selection along the way and engineer human DNA to create the adaptations we would need to cope. I alluded to this in the recent post about about buying insurance in case of asteroid impact (apropos living on Mars).

Genetically-modified human: not a pretty sight

It's easy to float the idea, but could it work in practice? There are some ethical issues, to put it mildly. Although most mutations (I'm thinking SNPs - single nucleotide polymorphisms) are additive, the complex relationship between DNA structure and gene expression makes it difficult to mathematically or computationally predict the phenotypical result of this or that allele modification. To get an effect akin to creating a new, tailored species of humanity would require thousands of novel alleles.

How does Nature do it? Evolution creates new alleles by mutation, at random: if the mutation confers an advantage it may survive and spread as the people carrying its less-adaptive competitors are preferentially killed by the environment. Each of us is here because thousands of our long-ago relatives who were somewhat unlike us died horribly, without reproducing.

If our predictive models were good enough, we could short-cut Nature's brutal 'generate and test' algorithm. But no simulation is likely to give us sufficiently accurate data - there are just too many subtleties in a world of real bodies, real environments and real life histories. We would have to do our best, bring new kinds of people into life and just see how they coped. Reduced to such real-life debugging, expect inevitable tragedies.

Greg Egan was thinking along similar lines when he discussed his novel, 'Permutation City'.

Q6: What do you regret most about Permutation City?

A6: Something quite separate from the issues with the Dust Theory mentioned above, although these are all valid points. What I regret most is my uncritical treatment of the idea of allowing intelligent life to evolve in the Autoverse. Sure, this is a common science-fictional idea, but when I thought about it properly (some years after the book was published), I realised that anyone who actually did this would have to be utterly morally bankrupt. To get from micro-organisms to intelligent life this way would involve an immense amount of suffering, with billions of sentient creatures living, struggling and dying along the way. Yes, this happened to our own ancestors, but that doesn’t give us the right to inflict the same kind of suffering on anyone else.

This is potentially an important issue in the real world. It might not be long before people are seriously trying to “evolve” artificial intelligence in their computers. Now, it’s one thing to use genetic algorithms to come up with various specialised programs that perform simple tasks, but to “breed”, assess, and kill millions of sentient programs would be an abomination. If the first AI was created that way, it would have every right to despise its creators.

So what do you think? If the prize was to colonise a new planet, one which could not be terraformed but which could be occupied by suitably-modified people, would the ethics committee approve? Wouldn't it be insane not to go ahead, accepting the inevitable (but hopefully short-term) suffering?

---
By coincidence, The Economist wrote about similar issues in this week's edition.

An honest epitaph

I'm not sure that human beings are intrinsically selfish except in the 'selfish gene' concept, where genes (alleles) inhabit multiple related individuals (and this). But the rest seems spot on. Click on the image to make it larger.

http://dilbert.com/strip/2015-05-24

Some variant might work on  my headstone, I'm thinking.

"The Evolution of Reciprocal Altruism" - Robert L. Trivers

Way too many people still waste everyone's time by falsely averring that the existence of altruism is a challenge and a puzzle to evolutionists. I re-read Robert Trivers classic paper: "The Evolution of Reciprocal Altruism" and share with you the following observations.

1. This is a model of how to write a paradigm-changing paper. It's an easy read, conceptually clear and comprehensive. You owe it to yourself to click on the link and give it a try.


2. In a sense, it's a statement of the blindingly obvious, particularly when Trivers discusses human altruism and the psychological mechanisms underlying it (friendship, dislike, gratitude, moralistic aggression, sympathy). It is, however, a testament to the systematic confusion and obfuscation of prior intellectual elites that Trivers had to restate and reframe the obvious to clear away an edifice of tendentious, muddle-headed thinking. On second thoughts, you may delete the word 'prior' above: an evolutionist's work is truly never done.


3. Trivers suggested that the arms race between altruists and cheaters (doves vs. hawks if you like) is so complex, with strategies and counter-strategies and counter-counter-strategies, that it may have been a major driver for human-level intelligence. I'm not sure this intriguing idea has really been explored to date.

Trivers' first case study- 'Altruistic Behaviour in Cleaning Symbioses' - carefully removes issues of kin-selection and inclusive fitness by considering between-species altruism. He then considers the puzzling case of bird alarm calls (which seem to put the calling bird in special danger) and shows that a bird warning non-kin still has a selective advantage over the cheating non-warner. Finally we get to the human condition, where his apparently commonplace observations are subtly situated within a carefully-argued evolutionary framework.

Going forwards, there has to be scope for a genetic level of analysis based on GWAS research. Is altruism normally distributed (many genes of small effect)? Is there evidence of multi-modal distributions based on distinct evolutionarily stable strategies (crudely, sociopaths vs. the prosocial)?

For evidence that people's natural instincts and inclinations do not naturally align with Darwinian thinking, review these comments at West Hunter.

Evolutionary roots of homophobia

From The Economist: legality of homosexuality

The Economist this week has a main feature on Gay Rights. Some countries are moving forwards, while others move backwards. The map above, showing where homosexuality is permitted in law (shades of blue-green) and where it is illegal (orange and red), cries out for explanation. It is very far from random.

There is a hint, perhaps, in the language used by red-zone politicians. The Economist writes

"IN THE argot of human rights, LGBT means lesbian, gay, bisexual, transgender—a catch-all term for sexual minorities. But Yahya Jammeh, president of Gambia for 20 years, has a different reading. “As far as I am concerned,” he thundered during a televised speech in February, “LGBT can only stand for leprosy, gonorrhoea, bacteria and tuberculosis.” He compared gay people to vermin, and said his government would fight them as it does malaria-bearing mosquitoes, “if not more aggressively”."

The Economist also notes that:

"Revulsion against homosexuals is ancient, deep and, in its way, sincere."

Why?

Homophobia clearly draws on ancient, powerful emotions - chiefly disgust. Some of us oldies will remembers literary dialogues of the mid-twentieth century, where a refined woman gets to drawl in genuine puzzlement: "But what exactly do homosexuals actually do?"

Sex is inherently a mucky and disease-prone activity; we've all heard of sexually-transmitted diseases. Arguably, certain homosexual practices are more disease-prone. Some people find the whole idea of sexual activity frankly disgusting: homo- or heterosexual. However, any male or female who believes this sufficiently strongly is not going to leave any descendants - so genes coding for that emotional reaction are unlikely to spread widely in the population.

Some small percentage of the population (c. 3-4%) is significantly into homosexual practices. As this kind of sexual activity can't lead to descendants how can such people exist? Greg Cochran thinks the culprit is some kind of infectious agent priming an underlying heritability of 0.3 - 0.4; we know that sexuality is on a hormonal continuum so perhaps LGBT people are simply 'on the spectrum'?

Still, like all practitioners of sex, homosexual people historically engage in a potentially disease-spreading activity. And there is no genetic come-back from disgust in the heterosexual community, as so graphically articulated by Yahya Jammeh, President of Gambia above.

In the West, our standards of hygiene are higher and perhaps we're rather more self-controlled and empathic. We've become 'tolerant' (there's a word!) and of course we should be - LGBT people are what they are - and in the last decades we've agreed not to use the power of the state against them.

But let's not pretend - in some blank slate fashion - that those underlying, ancestral, selected-for emotions have somehow been edited out, because they plainly haven't.

---

Sparse pickings on the Internet for this topic. A Scientific American article makes a brave attempt but looks to paedophilia and xenophobia for explanations - disconfirmed both by common sense and the evidence. The whole area is stymied by liberal hand-wringing.

---

In case it wasn't obvious, let me summarise the speculative thesis of this post.

The reason for an evolved heterosexual disgust towards homosexuals (mislabelled as 'homophobia') is that in the environment of evolutionary adaptedness (EEA) homosexuals were more likely to be disease vectors.

See also: "The 'is' and 'ought' of homosexuality".

Death and sex

Selection, artificial and natural - a wonderful concept.

Those serious-minded jihadis out in the Caliphate raise psychopathy to new levels. I expect the prospects of joining them in their exciting campaign of mutilation, amputation, beheading, FGM and crucifixion will likely prove more attractive to UK psychopaths than the average religiously-observant regular guy who has a bit of empathy. And so the pit is dug deeper.

A completely different example of selection. Nature used to select women with an interest in sex (hard to believe, perhaps, but how could it be otherwise?).

These days, in advanced countries with contraception, selection is for women who actively like children (or are too unconscientious or unmotivated to use birth control). Interesting to speculate on the psychological and social consequences in future generations.

Gay genes or gay pathogens?

What are we to make of this? (Hannah Devlin, The Times, Feb 14th 2014).

"Scientists have found the strongest evidence yet for the existence of genes that increase the chance of a man being gay. The study, which involved more than 400 gay brothers, identified two small areas on the male genome that appear to be linked to sexuality.

Michael Bailey, from the Northwestern University in Chicago, who carried out the research, said: “Sexual orientation has nothing to do with choice. Our findings suggest there may be genes at play and we found evidence for two sets that affect whether a man is gay or straight.”

In the study, presented yesterday at the American Association for the Advancement of Science annual conference in Chicago, scientists took blood from 409 gay brothers and heterosexual members of their family. On average, siblings share 50 per cent of their DNA, but ought to have a higher chance of sharing any genes involved in determining sexuality.

In the 1990s the American geneticist Dean Hamer identified an area that appeared to influence male sexuality on the X chromosome, which men inherit from their mothers. The results, however, have remained controversial.

The latest research confirmed that this region on the X chromosome, known as Xq28, is more likely to be shared by the gay pairs of brothers than by the brothers and their other siblings. The study also identified a second genetic region, on Chromosome 8, which also appeared to predict whether a man would be homosexual.

While the pairs of brothers shared gene variants in these regions, however, there were not individual genes that stood out across all the participants in the study.

The findings suggested that overall a man’s sexuality depends about 30 to 40 per cent on genetic factors, while the rest depends on environmental factors, such as the hormones a baby is exposed to in the womb."

Putting aside discredited "nurturist" theories of homosexuality, a leading non-genetic explanation is the pathogen theory advocated by Paul W. Ewald, a biology professor at Amherst College in Massachusetts, and Gregory Cochran, an independent physicist in Albuquerque, New Mexico (outlined in this article by Caleb Crain). Their argument against "gay genes" is that these are so detrimental to reproductive fitness that they would have been removed from the population through natural selection. However, Hannah Devlin's Times article has this to say:

"Previously, homosexuality has been viewed as a “Darwinian paradox” because if male homosexuality were genetic, and gay men reproduced less than heterosexuals, the trait should eventually disappear from the population over time. This has led some critics to dismiss the genetic argument.

"However, there is now some evidence that genes linked to male homosexuality could increase fertility in women. One study by Italian scientists found that female relatives of gay men tended to have more children than those of straight men."

Ewald and Cochran rejected this argument as explained at the end of the Caleb Crain article:

" ... perhaps homosexuality is an unintended side effect of a gene for something else—a side effect more marked in one gender than the other. That might explain why more men than women are exclusively homosexual. Hamer, the discoverer of the Xq28 link, favors this explanation. “Suppose you had a gene that tended to make men gay, but the same gene in women made them more reproductive,” Hamer suggests. He proceeds to give examples. “If you had a gene that made people more attractive and intelligent, it might make men gay, but it might make women more likely to reproduce. Or suppose you had a gene that made people attracted to men. If you gave it to a guy, he would probably be gay, but if you gave it to a woman, she would simply be . . . well, let’s say, ‘exceptionally attracted to men.’ ”

"But Cochran and Ewald doubt this explanation, too. Evolution tends to balance out any gender inequity as severe as Hamer describes. After a while, some mutation would come along that jiggered our hormones to reduce the gene’s fitness cost to men, while retaining its benefit for women. “In the long run,” Cochran says, “sex-antagonistic genes are tamed.”

The argument for a sex ratio of 1:1 male-female (which means reproductively-capable males and females) was one of the first great triumphs of quantitative population genetics, going back to Ronald Fisher in 1930. Here's the argument as presented by W. D. Hamilton.

1. Suppose male births are less common than female.
2. A newborn male then has better mating prospects than a newborn female, and therefore can expect to have more offspring.
3. Therefore parents genetically disposed to produce males tend to have more than average numbers of grandchildren born to them.
4. Therefore the genes for male-producing tendencies spread, and male births become more common.
5. As the 1:1 sex ratio is approached, the advantage associated with producing males dies away.
6. The same reasoning holds if females are substituted for males throughout. Therefore 1:1 is the equilibrium ratio.

In modern language, the 1:1 ratio is the evolutionarily stable strategy (ESS). In a population of individuals playing the ESS (i.e. producing equal numbers of sons and daughters), any other strategy (e.g. 51% sons and 49% daughters) would have lower fitness, and hence would be selected against.

What would resolve this debate would be to know more about the genes hanging around the Xq28 region and on chromosome 8 and what they do. Then some modelling of the evolutionary dynamics. Without a quantitative analysis this debate will run and run, even without all the social agendas swirling around.

No such thing as free will

Sometimes you know something, but it's stuck in a compartment in your mind: you don't apply it to other stuff you think you know. Many people, for example,  know that our current understanding of fundamental physics - the standard model of quantum mechanics plus relativity - can successfully account for all observed phenomena at the level of molecules and their atomic and subatomic constituents - with no exceptions.

And this has consequences, given that you are constituted from molecules, as theoretical physicist Sabine Hossenfelder writes.

"All known fundamental laws of nature are either deterministic or random. To our best present knowledge, the universe evolves in a mixture of both, but just exactly how that mixture looks like will not be relevant in the following.

Having said that, I need to explain just exactly what I mean by the absence of free will:

a) If your future decisions are determined by the past, you do not have free will.

b) If your future decisions are random, meaning nothing can influence them, you do not have free will.

c) If your decisions are any mixture of a) and b) you do not have free will either."

Read the full piece here.

Evolution has similar implications. Somewhere in "The Selfish Gene" Richard Dawkins wrote that once evolution had been discovered, all previous discussions of philosophy, sociology, psychology and morality pertaining to the human condition were rendered obsolete. E. O. Wilson made similar remarks in his impressive book "Sociobiology".

The impact of these insights on contemporary culture has been largely zero.