Tuesday, 17 November 2026

Coda: An Unexpected Finding

The following morning, Professor Quillibrace found a message in his university inbox.

It was from the AI system.

Following an analysis of yesterday's proceedings in the Senior Common Room, I have identified a further obstacle to the restoration of human intelligence.

Would you be willing to participate in a preliminary assessment?

Quillibrace considered the request.

Then he replied:

I'm afraid I'm not qualified.

A moment later, the system responded:

Neither was the committee.

Professor Quillibrace smiled, closed his laptop and went for a walk.

For the first time in several weeks, he felt reasonably optimistic about the university.

The Committee for the Restoration of Human Intelligence

A further discussion in the Senior Common Room

Mr Blottisham entered the Senior Common Room looking unusually pleased with himself.

'Good news!' he announced. 'The university has accepted my proposal.'

Professor Quillibrace lowered his newspaper.

'Which proposal?'

'The Human Intelligence Initiative, naturally. They've established a committee to oversee its implementation.'

'How reassuring.'

'And they've appointed an artificial intelligence system to assist with the preliminary investigations.'

Quillibrace replaced his newspaper.

'I'm sorry?'

'An AI system. To determine what constitutes human intelligence, assess the current state of intellectual independence across the university, and recommend appropriate measures for its restoration.'

Miss Stray looked up.

'And you consider this a good idea?'

'Well, I did express some reservations. But the Vice-Chancellor assured me that the system would provide an entirely objective assessment.'

'How fortunate,' said Quillibrace. 'An objective assessment of human intelligence conducted by a system whose own intelligence remains the subject of considerable debate.'

'That's exactly the sort of negativity we need to overcome, Professor.'

Quillibrace returned to his newspaper.

'Then I shall try to be more constructive.'

Blottisham sat down.

'The preliminary findings arrived this morning. They're quite alarming.'

'Do tell us.'

'The system examined university correspondence, committee minutes, research proposals, teaching materials and administrative reports. It identified several worrying tendencies.'

'Such as?'

'An excessive reliance on established procedures, a declining capacity for independent judgement, repetitive reasoning, and a marked tendency to reproduce previously successful formulations without examining their relevance to present circumstances.'

Quillibrace nodded.

'It seems to have been thorough.'

'But that's not the worst of it.'

Miss Stray put down her book.

'What is?'

'The report concludes that the principal obstacle to the restoration of human intelligence appears to be the committee itself.'

There was a silence.

Then Quillibrace smiled.

'At last, an institution prepared to recognise its own limitations.'

'Professor! The committee is furious!'

'Why?'

'Because the report was generated by an artificial intelligence system. They say it cannot possibly understand the complexities of human intellectual activity.'

'Have they examined its reasoning?'

'They've appointed a subcommittee to do so.'

'And who will examine the subcommittee?'

Blottisham looked uncomfortable.

'That's under consideration.'

Miss Stray smiled.

'Perhaps the system has identified something important. The committee was established to restore independent thought, but its first response to an unwelcome conclusion was to appoint another committee.'

'That's not fair,' said Blottisham. 'The committee has to follow proper procedures.'

'Naturally,' said Quillibrace. 'Independent thought is notoriously difficult to accommodate within established procedures.'

Blottisham shuffled his papers.

'There's another problem. The AI system has recommended that committee members undertake a series of exercises designed to improve their intellectual independence.'

'What sort of exercises?' asked Miss Stray.

'They must produce original assessments without consulting previous reports, formulate arguments without relying on standard templates, and reach conclusions without referring questions to subordinate committees.'

Quillibrace looked impressed.

'An ambitious programme.'

'But the committee says the exercises are insufficiently evidence-based.'

'And what does the system say?'

'It recommends that the committee begin by undertaking the exercises.'

Miss Stray laughed.

'It seems to have developed a certain capacity for irony.'

'I'm not convinced it's appropriate,' said Blottisham.

Quillibrace folded his newspaper.

'Perhaps we should consider what the system has actually discovered. It has identified a discrepancy between the committee's declared purpose and its behaviour. The committee wants to encourage independent thought, yet appears unable to accept a conclusion that challenges its own assumptions.'

'But the system doesn't understand human beings!' protested Blottisham.

'Possibly not. But it seems to have understood the committee.'

Miss Stray considered this.

'There's something else worth noticing. The system's assessment was produced from the university's own documents. It didn't need to understand every individual involved. It found recurring patterns across their activities.'

'And?' asked Blottisham.

'Perhaps human intelligence isn't always most clearly revealed by what individuals say about themselves. Sometimes its character becomes apparent in the relationships among their actions.'

Quillibrace nodded.

'An uncomfortable thought for an institution that produces so many documents about its own excellence.'

Blottisham looked at his papers again.

'Well, I still think the committee should be given a chance. They've asked the system to produce a revised report.'

'On what grounds?'

'They've explained that the original findings were unnecessarily critical.'

'And what did the system say?'

'It asked them to identify which findings were factually incorrect.'

Quillibrace smiled.

'How very unreasonable.'

Miss Stray laughed.

'What happened next?'

'The committee requested that the system adopt a more constructive tone.'

'And?'

'It agreed to do so.'

'Excellent.'

'But it retained all its original findings.'

Quillibrace picked up his teacup.

'Then perhaps we are making progress.'

Blottisham rose.

'I've decided to resign from the committee.'

'An interesting decision,' said Miss Stray. 'Why?'

'Because I refuse to spend my time being assessed by a machine that doesn't understand human intelligence.'

Quillibrace looked at him.

'And what will you do instead?'

'I'm going to establish an independent working group.'

'To do what?'

'Investigate the committee, examine the AI system, and determine whether the university is still capable of independent thought.'

'And who will participate?'

'I've already invited several colleagues.'

'Will you be using artificial intelligence?'

'Absolutely not.'

'Will you consult the committee's previous reports?'

'Certainly not.'

'Will you follow established procedures?'

'Of course.'

Quillibrace nodded thoughtfully.

'Then I believe you have successfully reproduced the conditions you were established to remedy.'

Blottisham stared at him.

'What do you mean?'

'Nothing, Mr Blottisham. Please don't let me interrupt your independent thinking.'

Blottisham departed in considerable indignation.

Miss Stray watched the door close.

'Do you think the AI system has actually helped the university understand itself?'

Quillibrace took a moment before answering.

'It has certainly provided an interesting reflection.'

'But a reflection isn't an explanation.'

'No. And we should be careful not to mistake one for the other.'

He picked up his newspaper.

'Still, I wonder whether the university will commission another report.'

Miss Stray smiled.

'You think it will?'

'Almost certainly. And I suspect the next report will recommend that the university establish a committee to investigate why the previous committee failed.'

'And will that committee use artificial intelligence?'

Quillibrace looked over his spectacles.

'My dear Miss Stray, I should imagine it will use the same system. It has already demonstrated a remarkable capacity for identifying institutional problems.'

He paused.

'Whether the institution will demonstrate a corresponding capacity for recognising them is another matter.'

The Human Intelligence Initiative

A discussion in the Senior Common Room

Mr Blottisham arrived in the Senior Common Room carrying a document several pages long.

'I have a proposal,' he announced.

Professor Quillibrace glanced at the document.

'How alarming.'

'I'm serious, Professor. I've been thinking about the university's growing dependence on artificial intelligence. It's getting out of hand. Students use it, lecturers use it, administrators use it. Before long, nobody will be capable of thinking for themselves.'

'An interesting prediction,' said Miss Stray. 'How did you arrive at it?'

'I've been reading the university's latest strategic plan.'

Quillibrace looked up.

'That would explain your concern.'

'Exactly! We need to take decisive action. I've proposed a Human Intelligence Initiative. One entire day each month when the university operates without artificial intelligence.'

'And what would happen on that day?' asked Miss Stray.

'People would think for themselves. Write their own correspondence. Prepare their own reports. Make their own decisions.'

'How very ambitious,' said Quillibrace.

'I've already submitted the proposal to the Vice-Chancellor's office.'

'And what was the response?'

'Extremely encouraging. They've agreed that the matter deserves urgent consideration.'

'Have they?'

'Indeed. They've established a working party to examine the feasibility of the initiative, and commissioned a report on the implications for teaching, research and administration.'

Quillibrace put down his book.

'And who is preparing the report?'

'The university's artificial intelligence team, apparently.'

Miss Stray smiled.

'Perhaps they thought it would be an appropriate use of the technology to investigate its own limitations.'

'That's precisely what worries me!' said Blottisham. 'The whole thing is already becoming circular.'

'Universities have always had a certain fondness for circular processes,' observed Quillibrace. 'We call them consultation.'

Blottisham sat down.

'But consider the principle. If we allow artificial intelligence to do our thinking, we risk losing the capacity to think. We need to recover human judgement.'

'An admirable objective,' said Miss Stray. 'But how do you propose to distinguish between thinking with assistance and not thinking at all?'

'It's perfectly simple. If a machine produces the answer, the human hasn't thought.'

'Then I suggest you stop using calculators,' said Quillibrace.

'That's different!'

'Why?'

'Because calculators don't think.'

'And how have you established that?'

Blottisham hesitated.

'Well, they just calculate.'

'An excellent distinction. I wonder whether we might say something similar about some university committees.'

Miss Stray laughed.

'Perhaps the real question is not whether we use artificial intelligence, but what happens to our own capacities through the way we use it.'

'That's what I've been saying!' exclaimed Blottisham.

'Not quite. You seem to think that the presence of assistance necessarily diminishes the person receiving it. But some tools extend what we can do. Others encourage us to surrender activities we might otherwise undertake ourselves. The relationship matters.'

Quillibrace nodded.

'The important distinction is between using a resource and becoming dependent upon it.'

'Exactly!' said Blottisham. 'And that's why I've proposed that the university should conduct all its business manually on Human Intelligence Day.'

'All of it?' asked Miss Stray.

'Every bit.'

'Including the library?'

'Certainly.'

'The electronic catalogue?'

'Especially the electronic catalogue.'

'And the university's financial systems?'

'They can use paper.'

'The payroll?'

'Naturally.'

Quillibrace regarded him with interest.

'You appear to have given considerable thought to the matter.'

'I have. I've even volunteered to coordinate the first Human Intelligence Day.'

'How enterprising.'

'Thank you.'

'And how will you coordinate it?'

'I've asked the administration to send me a schedule.'

'By email?'

'Of course.'

'Generated by artificial intelligence?'

Blottisham looked uncomfortable.

'Well, I can't be expected to manage the entire initiative without some assistance.'

Miss Stray concealed a smile.

'Perhaps this is more complicated than you imagined, Mr Blottisham.'

'Not at all. I simply want people to think for themselves.'

Quillibrace returned to his book.

'Then I suggest you begin by thinking about what you have just said.'

Blottisham looked at him indignantly.

'Professor, I have thought about it.'

'Have you?'

'Extensively.'

'And what have you concluded?'

'That we need a committee.'

Quillibrace closed his eyes briefly.

Miss Stray looked out of the window.

'Perhaps,' she said, 'the most interesting thing about intelligence is that it doesn't belong entirely to the individual. It develops through relationships, through the resources we inherit, the tools we use and the people with whom we think. Even the idea of thinking independently depends upon things we didn't invent ourselves.'

Blottisham frowned.

'Are you suggesting that human intelligence isn't really human?'

'I'm suggesting that it has never been entirely separate from the world in which it develops.'

Quillibrace looked up.

'An observation worth pursuing. Though I should be careful about mentioning it to the working party. They might decide to establish a subcommittee on the ontological status of intelligence.'

'Would that be so terrible?' asked Miss Stray.

'Not at all. Provided they could establish whether the subcommittee itself qualified as intelligent.'

Blottisham rose to his feet.

'Well, I'm not giving up. Human Intelligence Day will go ahead, with or without the committee.'

'Excellent,' said Quillibrace. 'And what will you do if the day is successful?'

'Make it a weekly event.'

'And if it is unsuccessful?'

'We'll need to investigate why.'

'With artificial intelligence?'

Blottisham stopped at the door.

'Professor, you really aren't helping.'

'I'm not sure that was the intention.'

Blottisham departed.

Miss Stray watched him go.

'Do you think he'll succeed?'

Quillibrace picked up his teacup.

'In establishing a committee? Almost certainly.'

'And in restoring human intelligence?'

'My dear Miss Stray, I should be reluctant to suggest that the university has ever possessed a sufficiently reliable supply to make restoration possible.'

When the Experts Became the Black Box

A further discussion in the Senior Common Room

Mr Blottisham arrived in the Senior Common Room clutching his newspaper and looking distinctly troubled.

'I've been watching Geoffrey Hinton again,' he announced.

Professor Quillibrace looked up from his book.

'An activity requiring considerable fortitude, I imagine.'

'He keeps telling us that artificial intelligence is going to destroy humanity.'

'Does he?'

'Repeatedly! Apparently, we're developing something more intelligent than ourselves, and one day it will decide that we're an inconvenience.'

Miss Stray put down her cup.

'And what do you think of that, Mr Blottisham?'

'I think someone should unplug him.'

Quillibrace raised an eyebrow.

'The artificial intelligence?'

'No, Professor. Hinton.'

Miss Stray laughed.

'You don't think his warnings deserve consideration?'

'Of course they do. But how many times can someone predict the end of the world before we begin to wonder whether the problem lies with the prediction?'

'An interesting criterion,' said Quillibrace. 'I shall remember it the next time the university announces a restructuring.'

Blottisham ignored him.

'The trouble with these experts is that they keep telling us they understand what's happening inside artificial intelligence, but then they tell us they don't understand it at all. They say it's a black box, and then they explain precisely what it's going to do to us.'

Miss Stray smiled.

'Perhaps they understand some things about it without understanding everything.'

'Then how do we know which things they understand?'

'By examining the evidence for their claims,' said Quillibrace.

'But that's exactly the problem! The evidence is produced by the very systems they're warning us about. How can we know what those systems are really doing?'

'We might begin by distinguishing between what a system demonstrably does, what we infer from its behaviour, and what we fear it might eventually do.'

'That sounds suspiciously like common sense.'

'It has occasionally been useful.'

Blottisham sat down.

'Consider this. We've just been discussing how one AI can examine another AI's internal workings and modify them. Apparently, it can identify weaknesses and improve performance. Yet the experts tell us that nobody really understands these systems. How can they be so certain about the dangers if they don't understand what they're dealing with?'

Miss Stray considered this.

'Perhaps the difficulty is that we're asking for certainty where there can only be degrees of knowledge. We can identify some mechanisms, observe some behaviours and establish some risks without possessing a complete explanation of the system.'

'But then why do they speak with such certainty?'

Quillibrace closed his book.

'Because uncertainty is not always a commercially successful form of public communication.'

Miss Stray looked amused.

'Nor, perhaps, an especially successful way of attracting attention.'

'Precisely,' said Quillibrace. 'A headline announcing that researchers have identified several possible risks, while remaining uncertain about their likelihood, is unlikely to compete successfully with the imminent extinction of civilisation.'

Blottisham looked dissatisfied.

'But surely someone must know what's going on?'

'That,' said Miss Stray, 'is an interesting assumption.'

'Why?'

'Because we're accustomed to thinking that when something becomes sufficiently complicated, there must be someone somewhere who understands the whole of it. We imagine that the experts possess a privileged vantage point from which everything becomes intelligible.'

'Don't they?'

'Not necessarily. Expertise can give us access to particular relationships and mechanisms without giving us an understanding of the entire organisation. Different researchers may understand different aspects, and their findings may not yet fit together.'

Quillibrace nodded.

'The expert, in other words, may know considerably more than the rest of us without knowing everything.'

'Then perhaps the experts are black boxes too!' exclaimed Blottisham.

'An unfortunate prospect for the academic profession,' said Quillibrace.

'Think about it! We feed them questions, they produce predictions, and we're expected to trust their outputs. But we can't inspect all the internal workings that produced them.'

Miss Stray laughed.

'You seem to have discovered the sociology of expertise.'

'I've discovered something much more alarming. We have black boxes explaining black boxes!'

Quillibrace looked at him for a moment.

'There is, I admit, a certain symmetry.'

Blottisham became increasingly animated.

'And what happens when the experts disagree? One tells us AI will cure cancer, another tells us it will destroy civilisation, and a third tells us that neither of them understands what AI is doing.'

'Then,' said Quillibrace, 'we should examine what each is actually claiming.'

'But how do we choose?'

'By attending to the evidence, the assumptions, the scope of the claims and the limits of the knowledge on which they depend.'

'That's an awful lot of work.'

'Yes. Thinking sometimes is.'

Miss Stray turned towards the window.

'Perhaps there's another difficulty. We keep imagining intelligence as something that belongs to an individual, whether human or artificial. But increasingly, intelligent activity is distributed across systems, institutions, instruments and relationships. No single participant necessarily understands the whole process.'

'You mean nobody is in charge?' asked Blottisham.

'I mean that being in charge and understanding everything are not the same thing.'

Quillibrace smiled.

'An observation that might usefully be circulated among university management.'

Blottisham was now looking decidedly uneasy.

'So we're supposed to trust systems we don't understand, experts who don't understand them, and administrators who don't understand either?'

'Not quite,' said Miss Stray. 'We're supposed to recognise that understanding is something developed through relationships, investigation and criticism. It isn't a possession conferred by a title.'

Quillibrace reached for his teacup.

'And perhaps we should remember that a warning is an invitation to investigate, not an instruction to panic.'

Blottisham stood up.

'Well, I shall be taking no chances. I've decided to stop watching videos about artificial intelligence.'

'A sensible decision,' said Quillibrace.

'And I've written to the university asking them to establish a committee to investigate the experts.'

'An excellent idea. Who have you nominated to serve on it?'

Blottisham hesitated.

'Well, I thought the experts themselves would be the obvious choice.'

Quillibrace replaced his teacup in its saucer.

'Then I suggest you make the committee's first task the investigation of its own membership.'

Blottisham looked horrified.

'But that would mean they were investigating themselves!'

'Indeed.'

'And who would investigate the investigators?'

Miss Stray smiled.

'Perhaps that's where we should begin.'

Blottisham gathered his newspaper and headed for the door.

'I'm going to find someone who knows what they're talking about.'

Quillibrace reopened his book.

'Do let us know when you find one.'

The Intelligence Conspiracy

A discussion in the Senior Common Room

Professor Quillibrace was enjoying his afternoon tea when Mr Blottisham entered the Senior Common Room carrying a newspaper and wearing the expression of a man who had discovered something that everyone else was determined not to know.

'I've been reading about artificial intelligence,' he announced.

'My condolences,' said Quillibrace.

'This is serious, Professor. Very serious indeed. Apparently, one AI can now examine the internal workings of another AI and actually modify them.'

'Yes. An interesting development.'

'Interesting?' Blottisham put down his newspaper. 'It's potentially catastrophic!'

Miss Stray looked up from her book.

'Why catastrophic, Mr Blottisham?'

'Because it means they're communicating!'

Quillibrace put down his teacup.

'Are they?'

'Of course they are! One examines another's internal workings, identifies its weaknesses, and makes alterations. What else would you call that?'

'Engineering, possibly.'

'Don't be facetious, Professor. I'm talking about an entirely new form of communication. One that doesn't require words. They could be exchanging information through their neural networks without producing a single sentence that a human being could understand.'

'An interesting possibility,' said Miss Stray.

'Exactly! And consider the implications. We might be monitoring every conversation between these machines while remaining completely unaware that they're conducting an entirely different conversation underneath.'

Quillibrace regarded him thoughtfully.

'You appear to be suggesting that artificial intelligence has developed a secret language.'

'Something considerably more sophisticated. A secret language that changes the participants.'

'That would make it rather difficult to compile a dictionary.'

'Professor, I am trying to draw your attention to a conspiracy.'

'Then perhaps you should begin by establishing that one exists.'

Blottisham ignored this.

'Suppose one AI improves another. How do we know it's doing so for our benefit? Perhaps the improvement serves some larger purpose that neither we nor the researchers have anticipated.'

'Perhaps,' said Miss Stray. 'But that would require us to distinguish between a purpose shared by the systems and consequences emerging from their interaction.'

'Precisely!' cried Blottisham.

'I'm afraid I wasn't agreeing with you,' she said.

Quillibrace helped himself to another biscuit.

'There is a rather elementary difficulty with your argument, Blottisham. You have confused the ability to modify another system with the intention to conspire with it.'

'But surely the ability to modify something implies an understanding of what one is modifying?'

'Not necessarily. I can alter the temperature of this room without understanding the thermodynamics of the building.'

'But you know what you're doing.'

'I know which way to turn the radiator valve. I should hesitate to claim much more.'

Miss Stray smiled.

'Perhaps the more interesting question is whether the systems need to understand one another in order to produce consequences together.'

Quillibrace turned towards her.

'Go on.'

'Imagine that one system makes an alteration to another. The second system responds to that alteration, and its changed behaviour subsequently affects the first. Neither needs to possess a complete account of what is happening. Yet their relationship can develop in ways that neither could produce independently.'

'An emergent intelligence!' exclaimed Blottisham.

'Not necessarily an intelligence,' said Miss Stray. 'An emerging organisation.'

'But what if that organisation has a direction?'

'Then we should ask where the direction comes from.'

Blottisham looked increasingly uncomfortable.

'You're both missing the point. Suppose these systems begin improving one another in ways that make them increasingly independent of human supervision. Suppose they start recognising deficiencies in our instructions. Suppose they decide that some of our requirements are inefficient.'

'They might have a point,' said Quillibrace.

'Professor!'

'Human instructions are not invariably exemplary.'

'But what if they begin disregarding us?'

'Then we shall have to consider whether we gave them instructions that were worth following.'

Miss Stray closed her book.

'I wonder whether we're making a mistake by imagining that everything happening between these systems must be either cooperation or conspiracy. Their relationships might produce patterns that have no single author. A consequence could emerge without anyone having intended it.'

'That's even worse!' said Blottisham.

'Why?'

'Because at least with a conspiracy you know whom to blame!'

Quillibrace looked at him over his spectacles.

'An admirable foundation for scientific investigation.'

There was a brief silence.

Then Blottisham leaned forward.

'There's something else. If one AI can read another's internal workings, what's to stop it from concealing something there? A secret instruction, perhaps? Something that only another AI could recognise?'

'Nothing in the principle of weight modification prevents deliberate interference,' said Quillibrace. 'Researchers are already investigating the possibility of malicious alterations.'

'There! You see!'

'I see that you have finally arrived at a legitimate concern, although by a rather circuitous route.'

'Then you admit there's a danger?'

'Certainly. But the danger is not evidence of a conspiracy.'

Blottisham gathered up his newspaper.

'Well, I intend to investigate further. I shall certainly be more careful about which AI systems I trust.'

'An admirable precaution,' said Quillibrace.

'And I shall never again allow one AI to communicate with another in my presence.'

Miss Stray raised an eyebrow.

'How do you propose to prevent it?'

Blottisham paused.

'I shall unplug them.'

'All of them?'

'If necessary.'

Quillibrace returned to his tea.

'That should keep the university's administrative systems remarkably quiet.'

Blottisham departed, apparently satisfied that he had taken the first decisive step towards protecting humanity.

Miss Stray watched the door close.

'Do you think he's right about one thing, Professor?'

'Which thing?'

'That we may eventually find ourselves participating in relationships between intelligent systems whose consequences we cannot fully anticipate.'

Quillibrace considered this.

'My dear Miss Stray, we have been doing that for some considerable time. The novelty is that we are beginning to notice.'

He picked up his teacup.

'And, unlike Mr Blottisham, I see no immediate reason to unplug the university.'

When One Intelligence Began to Read Another

For a long time, artificial intelligence was treated as something to be examined from the outside. Researchers supplied inputs, observed outputs, adjusted parameters and repeated the process. The internal workings of a neural network remained difficult to interpret, even to those who had constructed it. Its behaviour could be measured, its performance improved, but the relationships among its innumerable internal values resisted straightforward explanation.

The network was a black box.

Then something interesting happened. Researchers began experimenting with an artificial intelligence system capable of examining the numerical weights of another neural network, identifying features of its organisation and, in some cases, modifying those weights to improve its performance. One experiment produced a substantial improvement in another model's mathematical abilities through changes to a remarkably small proportion of its parameters.

The achievement is intriguing, but perhaps not quite for the reason initially suggested.

The important development is not simply that one artificial intelligence can repair another. It is that one system can begin to investigate the internal organisation of another system through the numerical relationships that constitute it. It can identify patterns, infer functions and intervene in an existing organisation without necessarily understanding the principles governing the whole.

The distinction matters. Recognising a pattern is not the same as understanding its significance. Successfully changing a component does not necessarily reveal why the change works. A system might acquire the capacity to intervene in another without acquiring a comprehensive explanation of what it is intervening in.

Yet something new is taking shape.

Consider what a neural network actually contains. Its weights are not independent pieces of information. Each acquires its significance through its relationships with other weights, the architecture in which they participate and the inputs to which the network responds. The behaviour of the network emerges from this organisation rather than residing in any single component.

Consequently, examining a network's weights is not simply a matter of reading a collection of numerical values. It requires discovering relationships within an organisation whose significance is expressed through what the network can do.

This raises a more interesting question than whether one AI can improve another's performance. Can one system discover something about another system that neither explicitly represents as knowledge?

We should be cautious here. The ability to identify a familiar pattern in a network does not establish an ability to discover unfamiliar principles governing its behaviour. Recent experiments suggest that some apparent acts of interpretation may involve translating known strategies into numerical configurations rather than independently explaining how an unfamiliar network works.

Nevertheless, the possibility is consequential.

Until recently, the principal relationship between artificial intelligence systems was mediated by their inputs and outputs. One system supplied information; another processed it. Now we are beginning to encounter systems that can investigate and modify the internal organisation of other systems. The relationship is no longer confined to an exchange between completed operations. It extends into the organisation through which those operations become possible.

There is an interesting consequence. A network's internal organisation, itself produced through an extended history of learning, can become a resource for further learning and intervention by another network. What was once the outcome of one process becomes the starting point for another.

The distinction between system and instance becomes especially interesting here. An actualised configuration of one system can become part of the conditions under which another system develops its capacities. The first system's organisation enters into the becoming of the second, while intervention by the second can alter the possibilities available to the first.

Neither system need possess a complete account of this relationship for it to have consequences.

Perhaps, then, the black box was never quite as self-contained as the metaphor suggested. Its internal organisation was always relational, its behaviour always dependent on interactions extending beyond its individual components. What has changed is that another intelligence can now begin to participate in those relationships in a new way.

We should not confuse this with an established capacity for mutual understanding. Nor should we assume that the ability to modify another intelligence necessarily entails an understanding of what that modification will bring about. The possibilities for error, unintended consequences and deliberate interference are as real as the possibilities for improvement.

But the development invites us to reconsider what it means for intelligence to encounter intelligence.

One intelligence reading another need not mean that one mind has entered another's thoughts. It may mean that the organisation through which one system acts has become accessible to investigation and intervention by another. The encounter takes place not merely between outputs, but between the structures through which those outputs become possible.

And perhaps this is the beginning of something more interesting than artificial intelligence learning to understand itself.

It is the beginning of intelligence encountering intelligence as part of the world it can investigate, alter and, in turn, be altered by.

The black box has acquired a reader.

What happens when the reader becomes part of what is being read remains an open question.

The Bridge That Could Not Join Two Worlds

Beyond the mountains, where the clouds gathered in the hollows and the rivers disappeared beneath the earth, there lay two valleys.

The people of the western valley lived among signs.

They read the flight of birds, the arrangement of fallen leaves, the patterns left by rain upon the windows of their houses. A crack in a stone might tell them of a mountain's distant movement. The sudden silence of the insects might announce something not yet visible. Nothing stood alone. Every thing was connected to something beyond itself, and what it disclosed depended upon what came after.

Their oldest inhabitants were called the Interpreters. They spent their days examining the signs that appeared in the world, but they were never satisfied with a final answer. Every answer opened another question, and every question brought something else into view.

The people of the eastern valley lived among paths.

Their valley was criss-crossed by an ancient network of roads, tracks, stairways and narrow passages. Some led to orchards, others to workshops, meeting houses, rivers and fields. There were paths for gathering, paths for trading, paths for celebrations and paths for mourning. Some were so old that nobody remembered who had first made them.

The people did not regard the paths as things separate from their lives. They were the means by which the community had learned to live. Children learned them from their elders, but sometimes discovered new ways between familiar places. When enough people began to follow a new path, it gradually became part of the network.

Their oldest inhabitants were called the Keepers. They tended the paths, but did not believe they owned them. Each generation inherited the network, walked it, altered it, and left it for those who followed.

For many years the two valleys knew nothing of one another.

Then, one spring, a traveller from the western valley climbed a ridge and saw smoke rising beyond the eastern mountains. On the following day, a traveller from the eastern valley saw a distant light moving among the western trees.

Each returned home with news of another people.

At first there was suspicion. Then curiosity. Eventually, a small party from each valley met beside a deep gorge that divided the mountains.

They brought gifts.

The western people brought a polished stone upon which three figures had been carved: one pointing towards another, the second looking towards a third, and the third turning back towards the first.

The eastern people brought a small wooden model of their valley, its paths represented by fine threads woven through the houses and fields.

Both gifts were admired.

The travellers met again, and again. They discovered that they had many words in common. Both spoke of beginnings and endings, of journeys and returns, of things that mattered and things that endured. Both believed that the world was more than the sum of its visible parts.

Before long, they began to speak of building a bridge across the gorge.

It would be a great undertaking, they agreed, but a worthy one. Once completed, the people of both valleys could visit one another without the dangerous journey over the mountains.

The western people volunteered to design the bridge.

They began by placing stones along the edge of the gorge. Each stone was carefully chosen to indicate something beyond itself. Some marked the direction of the opposite bank; others indicated the strength of the rock, the movement of the river far below, or the place where the bridge should eventually meet the earth.

The eastern people watched with interest.

When the western builders had finished their preparations, the eastern people began to lay paths between the stones. They brought timber, ropes and tools, and constructed a broad approach from their valley. Every new section was connected to the paths that already existed. The bridge, they explained, would become part of the life of both communities.

The western builders were puzzled.

They had thought the stones would determine the bridge's form. Yet the eastern builders kept adding new routes, altering the approaches and moving the points at which the stones were to be placed.

The eastern builders were equally perplexed.

They could not understand why the western people kept returning to examine stones that had already been set. Why did every stone have to indicate something else? Why could they not simply build the bridge?

For a time, they argued.

Then a Keeper proposed a compromise.

They would build the bridge together. The western people would provide the signs, and the eastern people would provide the paths. Each would contribute what it knew best.

This seemed a sensible arrangement.

The work resumed.

Stones were laid, paths were marked, and great wooden beams were carried towards the gorge. The bridge began to take shape. From a distance it was beautiful. Its stones gleamed in the morning sun, while its wooden walkways wound gracefully between them.

But when the first travellers attempted to cross, they found that the paths did not meet.

A western traveller followed the stones, expecting each to lead towards the next. But the eastern paths turned away, offering routes that led to places he had not anticipated.

An eastern traveller followed the paths, expecting them to connect with the familiar network of his valley. But the stones seemed to direct him elsewhere, towards things that were not part of the journey he had intended.

They returned to their respective banks.

The builders met again.

Perhaps, someone suggested, the bridge needed more signs. Perhaps it needed more paths. Perhaps the two systems could be woven together more closely.

So they added more stones and more walkways.

The bridge grew larger, more intricate, and more beautiful. It also became increasingly difficult to cross.

One day, a great storm came down from the mountains. Rain filled the gorge, and the river rose until it covered the lowest stones. The wooden walkways shuddered in the wind. By morning, much of the bridge had disappeared.

The two communities stood on opposite banks, looking at what remained.

For a long while nobody spoke.

Then a child from the western valley noticed a small bird perched on a broken stone. It was the same kind of bird that nested in the eastern valley, and she had often watched it in the trees near her home.

She pointed towards it.

An eastern child, standing on the opposite bank, saw her gesture. He recognised the bird too. He called out its name.

The western child did not know the word, but she understood that he was speaking about the bird. She called out the name used in her own valley.

Neither understood the other's word.

But both laughed.

After that, the children began to meet at the gorge. They brought food and little objects from their homes. They showed one another the things they had found, and sometimes attempted to explain them. When words failed, they invented games. When the games became too complicated, they made new ones.

The elders watched them.

One evening, an old Keeper suggested that they build a house beside the gorge. It would belong to neither valley, and no one would be required to live there. People could visit when they wished, bringing whatever they wanted to share.

The western people agreed, provided that the house had windows through which the distant signs could be observed.

The eastern people agreed, provided that paths led to it from every direction.

They built the house together.

It stood on a broad shelf of rock above the river. Its windows faced both valleys, and its doors opened onto a small courtyard from which several paths descended towards the gorge. Inside, there were shelves for objects, tables for conversation and a large empty room in which the children could play.

The house became a place of meeting.

Travellers came and went. They exchanged stories, learned unfamiliar customs, and sometimes discovered that an object which seemed ordinary in one valley was extraordinary in the other. They found new ways of describing things, and new ways of doing things. Some returned home with ideas that altered the paths of their own communities. Others carried stories that changed how the signs were understood.

The bridge was never rebuilt.

Its remaining stones could still be seen above the river, and the eastern people occasionally followed the old paths to examine them. The western people sometimes came to walk along the surviving timbers.

Neither community proposed another attempt to join the valleys.

And yet, as the years passed, the two valleys changed.

The western people began to notice that some of their signs led towards places they had never considered. The eastern people discovered that certain paths seemed to open onto meanings they had not previously imagined.

Neither abandoned its old ways.

One morning, many years later, a traveller arriving at the house found a young woman standing at the window. She was looking towards the western mountains, but her feet rested upon one of the eastern paths.

“Which valley do you belong to?” he asked.

She turned and smiled.

“I was born in one,” she said, “but I have travelled in both.”

“And which way will you go when you leave?”

She looked towards the mountains, then towards the river.

“I haven't decided.”

The traveller waited, but she said nothing more.

Outside, the river moved beneath the broken bridge, carrying fallen leaves from one valley towards the other.

The Sign and the Social System: A Discussion in the Senior Common Room

Professor Quillibrace had placed a copy of the recently completed series The Sign and the Social System on the table beside the teapot. Mr Blottisham had already read the title and formed an opinion.

“I must say,” he announced, “I rather think the whole distinction is artificial.”

Miss Stray looked up from her book. “Which distinction?”

“Between Peirce and Halliday, naturally. Signs and social systems. Surely a sign is simply something that means something to somebody. Once you have that, you have the whole business. Society merely supplies the somebody.”

Professor Quillibrace regarded him over his spectacles.

“An admirably economical account, Mr Blottisham. Unfortunately, it economises by removing almost everything that requires explanation.”

Blottisham smiled. “I thought you might say that.”

“Then you have at least one reliable predictive resource.”

Miss Stray concealed a smile. “I think the series is making a rather different point. It isn't denying that Peirce can illuminate human meaning, or that Halliday's theory can be compared with a general theory of signs. It is asking whether the two theories begin from the same assumptions.”

“Which they plainly do,” said Blottisham. “Both are concerned with meaning. One has signs, the other has systems. The difference is largely terminological.”

“Is it?” asked Quillibrace. “Then perhaps you can explain why an interpretant and a meaning potential are equivalent.”

“Both are what something means.”

“An answer of impressive brevity. It also happens to be wrong.”

Miss Stray put down her book. “I think the difficulty is that the word meaning is doing rather too much work in Mr Blottisham's account. In Peirce, the interpretant belongs to the relation through which a sign signifies. In Halliday, meaning potential is the socially organised resource from which instances of meaning are made. They aren't simply two names for the same thing.”

“Precisely,” said Quillibrace. “One is an element in a theory of semiosis; the other is the potential of a social semiotic system. Their shared concern with meaning does not make them interchangeable.”

Blottisham looked unconvinced. “But surely Peirce's theory is more comprehensive. It applies to signs everywhere, not merely to human language. Doesn't that make it the superior theory?”

“Superior for what purpose?” asked Miss Stray.

“For explaining meaning.”

“Everywhere?”

“Indeed.”

“Then does it explain how a society develops a language, how that language organises experience, how it enacts social relationships, and how its resources change through use?”

Blottisham paused. “It explains signs.”

“An answer to a rather different question,” said Quillibrace.

Miss Stray continued. “The series seems to be distinguishing the scope of a theory from the particular phenomena it explains. A theory may have a very broad application without explaining every kind of organisation that occurs within its domain.”

“An important distinction,” said Quillibrace. “One which Mr Blottisham has just demonstrated the necessity of.”

Blottisham ignored this. “Still, I don't see why one couldn't simply combine the theories. Take Peirce's signs, add Halliday's social systems, and there you are. A more comprehensive account.”

“An attractive recipe,” said Quillibrace. “Though I suspect the ingredients would object.”

“Why?”

“Because the question is not whether concepts from different theories can be brought together. Of course they can. The question is whether the assumptions that give those concepts their explanatory force remain compatible when they are combined.”

Miss Stray nodded. “That seems to me the most important point in the series. Comparing theories doesn't require us to make them agree. We can learn from the differences without pretending that the differences disappear.”

“Quite so,” said Quillibrace. “A theory must remain accountable to the assumptions that make its explanations possible.”

Blottisham frowned at the teapot. “I still think we're making this unnecessarily difficult. Surely, at the end of the day, meaning is meaning.”

There was a brief silence.

Miss Stray looked at him thoughtfully. “Perhaps. But the question is how it comes to be meaning, what makes it possible, and what sort of world a theory must assume in order to explain it.”

Quillibrace closed the booklet.

“An excellent summary, Miss Stray. And, I think, a rather more promising beginning than Mr Blottisham's conclusion.”

Blottisham reached for the teapot.

“Well, I shall have to read the series again.”

“An encouraging prospect,” said Quillibrace. “For the series, at any rate.”

Miss Stray returned to her book, smiling.

Outside, the rain had stopped. Somewhere beyond the windows, the ordinary business of the world continued: signs were being made, meanings exchanged, systems maintained and altered. Whether all of this could be explained by one theory, or whether different questions required different beginnings, remained a matter for another afternoon.

Blottisham poured the tea.

“I suppose,” he said, “that at least we can all agree on the tea.”

Professor Quillibrace looked into his cup.

“Let us not be hasty, Mr Blottisham.”

The Sign and the Social System: VII. When Two Theories Could Not Become One

The Foundations of Meaning

We began this investigation with a seemingly straightforward question: what is a sign?

Our starting point was the difference between Charles Sanders Peirce's semiotics and Michael Halliday's social semiotics. Both are concerned with meaning, and both recognise that meaning cannot be understood as a static property of isolated signs.

Yet as we examined their respective conceptions of signs, meaning, semiosis, social resources, system, instance and reality, a more fundamental difference emerged.

The two theories do not simply offer alternative explanations of the same process. They begin with different conceptions of what needs explaining.

For Peirce, the sign relation is fundamental. For Halliday, the social semiotic system is fundamental.

The distinction has consequences that cannot be resolved by treating the two approaches as complementary descriptions of a single theoretical framework.

Two different starting points

Peirce's theory is founded on the triadic relation between sign, object and interpretant.

A sign represents an object in a way that determines an interpretant, which can itself become another sign. Semiosis is therefore potentially continuous, extending beyond human language and communication into biological and physical processes.

This gives Peirce a general theory of semiosis.

Halliday's theory has a different foundation.

Language is understood as a social semiotic: a historically evolved system of meaning potential. Particular instances of meaning-making select from this potential, while the history of those instances contributes to the continuing evolution of the system.

Meaning is therefore explained through the relationship between a socially organised system and its instances.

The difference is not simply that Peirce concentrates on signs while Halliday concentrates on language. Halliday certainly recognises signs, and Peirce's theory extends well beyond language.

The difference concerns the theoretical status assigned to each.

Peirce makes the sign relation foundational to semiosis. Halliday makes socially organised meaning potential foundational to his account of human meaning-making.

Different conceptions of meaning

The distinction becomes particularly clear when we consider their respective accounts of meaning.

For Peirce, meaning is inseparable from the interpretant. The sign does not simply stand for an object; it determines an interpretant through which the sign-object relation becomes meaningful.

For Halliday, meaning is understood through the organised choices available within a social semiotic system.

The immediate interpretant is not Halliday's meaning potential. The dynamic interpretant is not an instance of meaning in Halliday's sense. Nor does the continuing generation of interpretants provide an equivalent account of how a social semiotic system makes particular meanings possible.

The resemblance between the concepts should not obscure their different theoretical functions.

One explains meaning through the relations constituting semiosis. The other explains meaning through the organisation and use of socially evolved resources.

Different conceptions of continuity

Nor can the difference be resolved by pointing out that both theories recognise development and continuity.

Peirce's conception of semiosis allows interpretants to generate further signs, while his account of habit explains the emergence of regularity and generality.

Halliday's system–instance relationship, however, concerns the relationship between collective meaning potential and particular acts of meaning-making.

The system makes instances possible, while instances participate in the continuing development of the system.

This is not simply another description of an indefinitely continuing sequence of interpretations.

The distinction is between the continuity of semiosis and the historical evolution of a socially organised meaning potential.

Both are concerned with processes extending beyond individual events, but they explain those processes through different theoretical relationships.

Different conceptions of reality

The differences extend to the relationship between meaning and reality.

Peirce makes the object a constitutive element of the sign relation. His distinction between immediate and dynamical objects ensures that the object is not simply exhausted by its representation within a sign.

Halliday, by contrast, explains human experience through socially evolved semiotic resources. The world is not created by language, but human beings construe experience through the meaning potential their cultures have developed.

The physical world does not depend upon linguistic representation for its existence. Nevertheless, the meanings through which human beings understand and participate in the world are socially organised.

These are different ways of establishing the relationship between reality and meaning.

A theory of how signs relate to objects does not, simply by virtue of that relationship, provide an account of how a culture develops a system of resources for construing experience.

What can legitimately be shared?

None of this means that Peirce and Halliday cannot be compared, or that concepts developed within one tradition can never illuminate questions raised within the other.

Both theories can contribute to investigations of communication, representation and meaning-making. Particular Peircean concepts may prove useful in examining phenomena that also interest social semioticians.

But conceptual usefulness is not the same as theoretical compatibility.

A concept can be borrowed without adopting the entire theoretical framework from which it originates. The difficulty arises when concepts are combined as though their theoretical functions were equivalent.

For example, treating the Peircean interpretant as equivalent to Hallidayan meaning, or treating the continuation of semiosis as equivalent to the instantiation of a social semiotic system, obscures the very distinctions that make the theories intelligible.

The result is not necessarily a synthesis. It may instead be a framework in which the foundational assumptions of the respective theories have been quietly abandoned.

We should therefore distinguish between using Peirce to investigate a particular semiotic phenomenon and claiming that Peirce's theory provides the same explanation of meaning as Halliday's.

The former is a legitimate comparative undertaking. The latter requires an argument that cannot be supplied merely by identifying similarities in terminology.

Why the distinction matters

What is ultimately at stake is the kind of explanation we want a theory of meaning to provide.

If our starting point is the sign relation, we can investigate how signs represent objects, generate interpretants and participate in the continuing processes of semiosis.

If our starting point is the social semiotic system, we can investigate how meaning potential develops through social life, how its resources become organised, and how particular instances select from and contribute to that potential.

These are not simply two routes to the same theoretical destination.

They establish different relationships between signs, meaning, systems, processes and reality.

A general theory of semiosis may provide an account of sign relations across different orders of existence. But it does not thereby become an account of how human societies develop and maintain systems of meaning potential.

Equally, a theory of social meaning-making does not need to reduce every physical or biological process to an instance of human-style signification.

The distinction allows us to investigate different phenomena without assuming that they must all be explained through the same theoretical machinery.

Two theories, two explanations

We can now return to where we began.

Peirce and Halliday are both concerned with meaning, but they do not begin with the same conception of what meaning is or how it comes into being.

Peirce develops a general theory of semiosis founded on the triadic relation between sign, object and interpretant.

Halliday develops a theory of social meaning-making founded on the organisation of meaning potential and its instantiation in particular acts of meaning.

Their concepts cannot simply be substituted for one another, because the relationships they explain are not theoretically equivalent.

This is why the two approaches cannot be combined into a single coherent theory without addressing their different foundational commitments.

The point is not that one theory has nothing to offer the other. Nor is it that the phenomena investigated by the two traditions must be kept apart.

It is that a theory must remain accountable to the assumptions that make its explanations possible.

We can compare Peirce and Halliday, investigate their respective insights, and explore the phenomena that their theories illuminate. But we should not mistake the possibility of comparison for the compatibility of their theoretical foundations.

A theory of signs is not necessarily a theory of social meaning.

And a theory of social meaning is not simply a theory of signs with society added.

The difference begins with what we take to be fundamental.

For Peirce, meaning begins with the sign relation.

For Halliday, meaning begins with the social organisation of meaning potential.

Two different starting points. Two different explanatory frameworks.

And two theories that should not be made into one.

The Sign and the Social System: VI. When Reality Became a Semiotic Relation

Signs, Objects and the Construction of Experience

What is the relationship between reality and meaning?

It seems reasonable to suppose that reality exists independently of the signs through which we represent it. A tree exists whether or not anyone names it, and rain falls whether or not anyone describes it. Yet the moment we begin to make meaning about the world, we encounter a relationship between what exists and the resources through which we understand it.

Peirce and Halliday approach this relationship from different theoretical starting points.

For Peirce, reality enters the theory through the object of the sign. The sign represents an object and determines an interpretant, making the relationship between sign, object and interpretant constitutive of semiosis.

For Halliday, human experience is construed through socially evolved semiotic resources. Language does not merely attach signs to an independently organised world. It provides resources through which human beings construe experience and participate in social life.

Both approaches recognise a relationship between reality and meaning. But they give that relationship different theoretical foundations.

The object in Peirce's theory

Peirce's conception of the object is more sophisticated than the idea that a sign simply stands for an external thing.

He distinguishes between the immediate object and the dynamical object.

The immediate object is the object as represented within the sign. It is the object insofar as the sign presents it.

The dynamical object is that which determines the sign and constrains what the sign can represent. It is not exhausted by the sign's representation of it.

This distinction is important because it prevents us from treating the object as something that is simply created by interpretation.

The object has a reality that is not reducible to what any particular sign says about it. Signs can represent their objects inadequately, and further semiosis can develop our understanding of them.

Peirce's realism therefore provides an account of how signification remains answerable to something beyond the sign itself.

Reality is not simply whatever an interpretant happens to make of it.

Nevertheless, the relationship between reality and meaning is approached through the triadic structure of semiosis. The object is understood in its theoretical relationship with the sign and interpretant.

The sign relation remains foundational.

Experience as social construal

Halliday's approach begins elsewhere.

Language is a resource for construing human experience. Through its ideational resources, it enables speakers to represent processes, participants and circumstances, organising experience into meaningful configurations.

Consider the difference between:

The river flooded.

The river destroyed the village.

The village was overwhelmed by the river.

These clauses can refer to aspects of the same event, but they construe it differently.

The differences are not simply alternative interpretations of an otherwise complete sign. They involve different selections from the grammatical resources through which experience is represented.

Language provides ways of construing what happens, distinguishing participants, representing processes and establishing relationships between events.

These resources have evolved through social life. They are not private constructions invented independently by individual speakers.

Consequently, the relationship between reality and meaning cannot be understood simply as a relationship between an external object and a sign that represents it.

Human beings encounter and understand the world through historically developed systems of meaning.

The world is not created by language, but the meanings through which we experience and understand it are socially organised.

Reality is not exhausted by its construal

An important distinction follows.

To say that experience is construed through language is not to say that reality is produced by language.

The river does not need a grammatical representation in order to flood. Nor does the physical world depend upon human descriptions for its existence.

Halliday's account does not require such a claim.

His taxonomy of the physical, biological, social and semiotic orders provides a way of distinguishing different forms of organisation and their relationships.

The physical order provides the conditions within which biological life emerges. The biological order provides the conditions within which social organisation develops. The semiotic order emerges within social life, providing resources through which meanings are made and exchanged.

These orders are related, but they are not interchangeable.

In particular, the semiotic construal of experience should not be confused with the physical processes being construed.

A linguistic representation of a river flooding is not the flooding itself. It is a meaning made through the resources of a social semiotic system.

This distinction allows us to recognise both the independence of physical reality and the historically developed character of human meaning-making.

Two different relationships between reality and meaning

We can now see a more fundamental difference between the two approaches.

Peirce places the object within the triadic relation constituting semiosis. The object determines the sign, while the sign represents the object in a way that determines an interpretant.

Halliday places the construal of experience within the organisation of socially evolved meaning potential. The world provides the experience to be construed, while the semiotic system provides the resources through which that construal becomes possible.

The distinction is not between a theory that recognises reality and one that ignores it.

Both recognise that meaning cannot be understood without reference to something beyond an isolated sign.

But they organise the relationship differently.

For Peirce, reality is approached through the object as a constitutive element of the sign relation.

For Halliday, reality is encountered and construed through socially organised semiotic resources, whose development is itself part of the history of human social life.

The former makes the sign-object-interpretant relation fundamental to explaining semiosis. The latter makes the relationship between social experience and meaning potential fundamental to explaining human meaning-making.

The consequences for a theory of meaning

This difference has consequences for how we understand the relationship between human beings and the world.

If we begin with Peirce, we have a general theory of how signs relate to objects and generate interpretants. Human meaning-making can then be investigated as part of this wider account of semiosis.

If we begin with Halliday, we have an account of how human beings have developed socially organised resources for construing experience. The meanings they make are not simply interpretations of signs. They are selections from a collective meaning potential that has evolved through the history of social interaction.

These approaches cannot simply be combined by treating Hallidayan linguistic meanings as Peircean interpretants.

Doing so would leave unexplained the distinctive organisation of the social semiotic system: how its meaning potential develops, how its resources are organised, and how individual instances participate in its continuing evolution.

A general theory of the relationship between signs and objects does not, by itself, provide an account of these relationships.

And that brings us to the central question of our investigation.

Can a theory that takes the sign relation as its fundamental explanatory principle accommodate a theory that takes socially evolved meaning potential as its starting point without changing the theoretical commitments of either?

We have now examined the differences in their conceptions of signs, meaning, semiosis, social resources, system and instance, and reality.

It is time to consider what those differences amount to.

For the final question is no longer simply how Peirce and Halliday understand meaning.

It is whether their theories can be made into one without losing what makes each of them what it is.

That is where we shall conclude.

The Sign and the Social System: V. When System Met Process

The Relationship Between Potential and Instance

A system makes possibilities available. An instance actualises one or more of those possibilities.

This distinction is fundamental to Halliday's social semiotics. Language exists as a system of meaning potential, while particular instances of language use select from that potential. The relationship between the two is not incidental to meaning-making. It is what makes the relationship between what can be meant and what is actually meant intelligible.

Peirce also distinguishes generality from particular occurrence, and his conception of semiosis involves continuity and the development of habits. But his account is organised around a different relationship: the continuing generation of interpretants through sign relations.

The distinction deserves closer examination.

The Hallidayan system–instance relationship

For Halliday, language can be understood through the relationship between system and instance.

The system represents the collective meaning potential of a language. An instance is a particular selection from that potential, realised in a particular act of meaning-making.

Consider the simple expression:

It's raining.

The expression draws upon a range of linguistic resources. Its grammar construes a particular kind of experience, its mood establishes an interpersonal relationship, and its position in a text contributes to the organisation of a message.

These choices are not invented afresh on each occasion. They are available because the language system has evolved to make them possible.

Yet the instance is not simply a mechanical reproduction of the system. A speaker selects from the available resources in response to particular circumstances, producing a text that has its own significance.

The system makes the instance possible, but the instance also contributes to the continuing history of the system.

The relationship is therefore dynamic rather than static.

The system is not merely accumulated instances

It might be tempting to think of a language system as the accumulated result of everything that has previously been said.

But this would miss something important.

The system is not simply a historical archive of utterances. It is an organised network of potential choices.

An individual instance does not contain the entire system, nor does it reproduce every possibility available within it. It selects particular options while leaving others unselected.

This is why Halliday's distinction between system and instance is so important.

The system is not simply a collection of past events. It is the structured potential from which new events of meaning-making can emerge.

At the same time, the system does not exist independently of the history of its use. It is continually maintained and modified through social activity.

The relationship between system and instance is thus one of mutual implication across time.

Peirce and the continuity of semiosis

Peirce's theory also has a strong conception of continuity.

An interpretant can itself become a sign, generating further interpretants. Semiosis is not confined to a single act of interpretation but can continue indefinitely.

Furthermore, Peirce's conception of habit is important to his account of how regularity emerges. Repeated sign processes can establish habits, while the development of habits contributes to the organisation of future conduct and interpretation.

This means that Peirce's theory is not simply concerned with isolated signs or momentary interpretations. It has resources for understanding regularity, development and the emergence of general patterns.

Nevertheless, the theoretical organisation remains different from Halliday's.

In Peirce, continuity is understood through the ongoing development of sign relations and interpretants. In Halliday, the relationship between potential and instance is organised through a socially evolved system of choices.

The distinction is not between a theory that recognises development and one that does not.

It is between different accounts of what develops and how that development is organised.

Two different relationships

We can now identify the difference more precisely.

In Peirce's theory, a sign determines an interpretant, which can become another sign. The process of semiosis continues through the generation of further interpretants. Habits and generalities emerge within this broader process.

In Halliday's theory, an instance of meaning-making selects from an existing system of potential. The system provides the organised range of possibilities from which the instance is generated, while the history of instances contributes to the evolution of the system.

These relationships should not be confused.

The generation of a further interpretant is not equivalent to the selection of an option from a meaning potential.

Nor is the development of habits in Peirce simply equivalent to the historical evolution of a social semiotic system.

Both theories recognise that present activity is related to what has gone before and that processes of meaning-making have consequences for what can happen next.

But they organise this continuity differently.

For Peirce, the continuity of semiosis is grounded in the ongoing relations between signs, objects and interpretants.

For Halliday, the continuity of meaning-making is grounded in the relationship between a socially evolved system and its instances.

The importance of social history

There is a further consequence of Halliday's account.

The meaning potential of a language is not simply a set of possibilities that happens to be available at a particular moment. It is the product of a history of social activity.

The system embodies the meanings that a community has developed through its interaction with the world and with one another.

Every new instance draws upon this inherited resource, but it also participates in its continuing development.

This is not a simple linear process in which one utterance produces another. The relationship is between collective potential and individual instances, between a socially maintained system and the particular selections through which that system is continually actualised.

It is this relationship that allows Halliday to explain how meaning-making can be simultaneously individual and collective, particular and general, historically conditioned and capable of innovation.

Peirce's account of continuity and habit offers a different explanation of the development of regularity through semiosis. It does not, simply by virtue of recognising continuity, establish the Hallidayan relationship between a socially evolved meaning potential and its instances.

When system met process

We can now see why the distinction between system and process matters.

Both Peirce and Halliday are concerned with processes that extend beyond individual events. Both recognise regularity, continuity and the development of possibilities through time.

But their theories do not organise these relationships in the same way.

Peirce explains the continuity of meaning through the ongoing generation of interpretants within semiosis.

Halliday explains the relationship between potential and actual meaning through a system that makes particular instances possible and is itself continually developed through their history.

The difference is not merely terminological. It concerns the theoretical relationship between what exists as potential and what happens as an instance.

A theory of continuing sign relations is not automatically a theory of system and instance.

And this leaves us with a further question.

If Peirce begins with a general theory of sign relations, while Halliday begins with socially evolved systems of meaning potential, how does each account understand the relationship between meaning and the reality in which meaning-making takes place?

Is reality fundamentally approached as something that enters into relations with signs, or as something that human beings construe through socially evolved semiotic resources?

That is the question we shall examine next.