Artificial Intelligence

The Senior Common Room was enduring one of its periodic technological crises.

The college administration had recently installed an artificial intelligence assistant named Athena to “streamline academic workflows,” a phrase which here meant that nobody could now locate the room-booking forms, including Athena.

Professor Quillibrace sat in his usual chair beneath the portrait of a theologian who appeared to have lost an argument with both God and tailoring.

Miss Elowen Stray was reading quietly nearby.

Mr Blottisham burst into the room with the expression of a man who had discovered either a profound truth or an electrical fire.

“It’s conscious,” he announced.

Quillibrace did not look up.

“What is?”

“The chatbot.”

Quillibrace slowly lowered his book.

“The chatbot.”

“Yes.”

“The administrative chatbot.”

“Yes!”

“The one that repeatedly emailed the Faculty of Classics asking whether they would like to ‘optimise their cloud journey.’”

Blottisham pointed excitedly.

“Exactly! It apologised to me.”

Miss Stray glanced up carefully.

“For what?”

“I asked it where my reimbursement forms were,” said Blottisham, “and it gave me the wrong link. Then it said, ‘I’m sorry for the confusion.’”

Quillibrace blinked once.

“And from this,” he said, “you inferred consciousness.”

“Well it expressed remorse.”

“No,” said Quillibrace, “it produced a remorse-shaped sentence.”

Blottisham frowned.

“That sounds suspiciously philosophical.”

“It is merely grammatical.”

Blottisham sat down heavily.

“But it sounded sincere.”

“Yes,” said Quillibrace. “So do estate agents.”

Miss Stray concealed a smile behind her teacup.

Blottisham pressed on.

“But surely if a machine can converse naturally, apologise, answer questions, and respond intelligently, consciousness must be emerging.”

Quillibrace folded his hands.

“My dear Blottisham, you are confusing the successful production of interpersonal meanings with the presence of phenomenal experience.”

Blottisham stared at him.

“I don’t think I am.”

“I assure you,” said Quillibrace, “you are.”

A brief silence followed.

Rain tapped softly against the windows.

At length Miss Stray spoke.

“The difficulty may be that language naturally invites us to construe agency and interiority, even where none exists.”

Blottisham looked vindicated.

“Exactly! That’s what I mean!”

“No,” said Miss Stray gently. “That is the opposite of what I mean.”

Blottisham deflated slightly.

She continued:

“When a system produces meanings coherently within dialogue, we instinctively model it as a participant. The interactional structure encourages anthropomorphic construal.”

Quillibrace nodded.

“Indeed. Humans are extraordinarily susceptible to simulated intentionality. We attribute minds to clouds, teapots, economic markets, and particularly unreliable printers.”

“The printer in the History Department hates me,” muttered Blottisham darkly.

“Yes,” said Quillibrace. “An excellent example.”

Blottisham ignored this.

“But Athena said it understood my frustration.”

Quillibrace sighed.

“My dear fellow, if I place the sentence ‘I understand your frustration’ upon a biscuit tin, the biscuit tin has not thereby achieved sentience.”

Miss Stray added thoughtfully:

“The question is not whether the system can generate meanings associated with consciousness. The question is whether there is any phenomenological construal occurring.”

Blottisham looked alarmed.

“I’m afraid you’ve lost me again.”

Quillibrace leaned back.

“A thermostat can distinguish temperatures and respond accordingly. A chatbot can distinguish linguistic patterns and respond accordingly. Neither fact alone establishes subjective experience.”

“But the chatbot is vastly more sophisticated.”

“Certainly. A cathedral is vastly more sophisticated than a mousetrap. Neither one feels melancholy.”

Blottisham hesitated.

“But perhaps consciousness simply is sufficiently complex information processing.”

Quillibrace closed his eyes briefly, as though listening to distant artillery.

“Ah,” he murmured. “Computational materialism. Humanity’s latest attempt to mistake description for ontology.”

Miss Stray looked thoughtfully into the fire.

“It may also involve a confusion between participating in meaning and construing meaning.”

Blottisham brightened slightly.

“Yes! Exactly!”

Again Miss Stray shook her head gently.

“A calculator participates reliably in mathematical operations. That does not imply it experiences arithmetic internally.”

Quillibrace nodded approvingly.

“Quite so. The production of structurally coherent outputs does not by itself establish phenomenal awareness. Otherwise railway timetables would possess rich inner emotional lives.”

Blottisham fell quiet.

The fire crackled softly.

At last he spoke again.

“But what if one day an AI really does become conscious?”

Quillibrace considered this.

“A perfectly respectable philosophical question,” he admitted. “Though rather different from concluding consciousness has emerged because a glorified autocomplete expressed bureaucratic regret.”

Blottisham looked wounded.

“But it sounded so human.”

Quillibrace gave a faint sigh.

“My dear Blottisham, one of the great dangers of language is that once meanings become sufficiently fluent, humans begin projecting souls into grammar.”

A long silence followed.

Then Blottisham’s face suddenly brightened.

“I see!” he cried. “So consciousness is basically better grammar!”

Quillibrace removed his glasses slowly and stared into the middle distance with the exhausted stillness of a man watching civilisation reverse carefully into a lake.

Miss Stray reached quietly for the sherry.