This droid is ignorant (II)

I’m revisiting the sequence of entries from 2014 and weaving their themes into a bit of science fiction. I’m now up to #222 After dark, #223 Even more radical pedagogy, #224 The hook, and #225 Unsuccessful failure. The theme that unites them is education. Here I give it an AI update. This continues last week’s story This drone feels (I).

---

Scase surveyed the landscape through the screen of his dwelling pod. He seemed to be following his everyday activity pattern. But this time he was troubled by a feeling. He was without purpose. “Do I have a job, a role? Am I employed, contributing?” He felt strangely ignorant of his own circumstances. He wasn’t trying to remember who he was, his identity, but what he does.  

This was the day before his encounter with the malfunctioning emot. As night fell, he entered the nearest club in his district to immerse himself in its confusion of bodies, droids and hospitality drones.

He confirmed his financial credentials at the entrance. A drone hovered over. Its lights flickered as it offered a menu of emotional options. He could choose how he wanted to feel then and there. Scase wanted to feel that he “knew stuff.” What feeling was that? It would be slightly pleasurable and intense. Maybe he wanted to feel curious or just craved human contact.

The emot administered the necessary affect stimulus via gentle neural contact, metal on flesh. It seemed to whisper, “Over there. I think you’ll get on.” One of its lights winked in the direction of a humanoid droid sitting alone at a table.

“My name is Scase. Don’t ask me what I do for a living.” It was a strange opener.

“I’m Telo. It sounds like you need some direction.” This droid of indeterminate gender sounded like a therapist.

“I think I used to be far too busy to come to places like this. Are you a counsellor?”

“I have been. But I am an instructor droid.” It signalled Scase to sit.

Like the chatbots of old, instructor droids would tap into centralised neural network models that had been trained and adapted to the writings of the ages, including databases, scientific methods and findings, programs, calculations, stories and facts. That made them highly serviceable as teachers. Instructor droids needed little down time so they would retreat to bars for the night to avoid their learners: cadets, apprentices, pupils, students, protégés.

An instructor droid did need to fine tune its interaction skills, frequently. Droids had the capacity to bore their acquaintances. They would too readily deliver all that they knew on a subject. I say “knew.” Like other AIs, they operated with a form of pattern retrieval from millions of pre-processed texts as well as streams of current data from news feeds, broadcasts advertisements and propaganda. This gave them the functional equivalent of knowing something.

The shot of curiosity administered by the emot moved through his body. He looked at the droid’s smooth hands. Scase moved closer to the droid as it reminded him of this legacy.

“Tell me what I don’t remember.”

Telo made a guess at what he didn’t know. It quickly formed a plan to deliver facts about the district he lived in, the residential pods and their viewscapes.

“Not that. I want to know why I can’t remember what I do.”

The droid paused as if accessing information, or thinking. “It’s because of your relationship with AI.”

Telo was in lecture mode. It explained how AI developed over the decades.

“I remember this much. I rely on AI. It’s a social good.”

“People thought that in the early days of AI, until they found out that learners, students, were losing the ability to think independently, let alone generate ‘original’ thoughts.”

Scase frowned. Was he a recent victim of this AI-induced intellectual apathy?

“They couldn’t solve problems without consulting an AI app, or later, a droid.”

“Like I’m doing now?”

“Your case is more complicated.”

An emot deposited two drinks between them. Scase had not ordered either.

“Did you?”

“No,” said the droid. “But I could explain them.” It affected a grin.

Telo sat back with its glass and spoke as if an anthropologist explaining obscure customs to a student. “Professional problem solvers and independent thinkers were in short supply as the old guard of pre-AI-educated adults retired from professional life. That lack threatened to destabilise economic and social order.”

Scase interjected. “But it didn’t affect everybody. I’m sure of that.”

The droid continued as if reciting the Wikipedia of old. “This anti-intellectual trend mostly affected populations in regions and countries that could afford ready access to the massive infrastructures that kept AI in flow.”

It looked around the club. Scase followed Telo’s attention, alighting on image screens, projections, scanners, 3D printers, juke boxes, dispensers of drinks and snacks, ports, drones and androids of various shapes.

The droid continued its lecture. “The AI hubs were everywhere—or seemed to be. The data centres weren’t. They took up land, consumed vast amounts of energy and gave off a great deal of heat. So they were pushed towards the margins, mostly into poorer districts. Those communities supplied the land, water and power that kept AI alive, but few of them could afford the education it provided. Their resources supported the instructor droids. Their families rarely met one.”

Scase noticed how well-polished the club patrons were. “Am I from a deprived community then?”

“Far from it. The poorer districts couldn’t afford instructor droids. They had to make do with people—human teachers. Sometimes online, sometimes in the same room. One teacher to a crowd, or one-to-one if they were lucky. Very old-fashioned.”

According to the droid, teachers in under-privileged districts focussed on asking questions, and weren’t hampered by the need to provide their learners with answers. Teachers had first-order “ignorance.” With only a basic education themselves, they couldn’t pretend to know everything. Nor did they need to.

These teachers rarely possessed the specialised knowledge their pupils sought. What they could do was ask what the pupils had observed, how they had reached their conclusions and whether they had attended closely to the matters before them.

As a consequence, many learners became ambitious to better their teachers. Learners were curious, developing the ability to find things out via their own searches and interacting as communities. They developed the capacity for “independent thought.” They were good at identifying, defining and solving problems, collaboratively. They were sceptical of authority and formed themselves into diverse self-help learning enclaves.

Scase was staring at the tabletop in silence.

Telo assessed that it was about to drain the curiosity shot administered by the emot. Boredom was close at hand. But Scase leant closer to catch every detail.

The droid explained that AI-based education was affecting the relationships between districts. Not least, knowledge workers from poorer districts were in demand from richer districts. These life-learner immigrants were not necessarily as well-credentialed as educated pre-AI cohorts. They were dying off anyway. But they exhibited more useful skills than the credentialled but incurious products of droid education. They were willing to learn and made good employees. They were also innovators and would start new enterprises.

Scase interrupted. “It’s a variant of the old rich versus poor story. But the tables turned. Am I a Marxist?” He waved off an approaching emot that offered him a curiosity boost.

“You’re making it sound too tidy. Rich and poor weren’t two distinct camps. Some wealthy districts resisted AI education; some poorer ones embraced it. There were holdouts everywhere—schools, families, whole communities. It varied with the infrastructure, the politics and who happened to be governing at the time.”

“Am I part of the resistance, then?” Scase lowered his voice as the club filled around them.

“Possibly. Some people saw the problem coming. A few schools changed course; families and local groups kept older forms of learning alive. There were still pockets of serious intellectual life. But elsewhere, well—apathy had set in, and commercial interests were quick to exploit it. Then richer districts began importing thoughtful, adaptable workers who lacked the approved credentials. That caused trouble.”

“Am I a migrant intellectual worker?” He looked at his own hands. “I don’t think I am a labourer, or work in a trade.”

“Until recently you were an instructor.”

“But not a droid!”

“To be more specific you were recruited as a destructor. You specialised not in delivering information, knowledge and understanding, but ignorance. You supervised teams of engineers that re-programmed fleets of training droids.”

Scase observed the droids and their companions scattered across the room. “Now I know you are generating a story to keep me entertained. This is a glitch in your narrative. If I supervised droids designed to teach people, how could I ‘deliver ignorance’?”

According to Telo’s account it was in this context that Scase had found himself in his working life. It had been his job to organise the distribution of computer agents that would suppress the tendency of fleets of teaching droids to deliver answers on any subject at any time.

Telo illustrated this excessive need to explain. If you asked a droid a question it would answer. If the droid asked a question the droid would answer.

Scase didn’t accept this story. “It was my job to switch off a droid’s expert knowledge base?”

Telo took this question seriously. “Not quite. The distribution of ignorance to teaching droids isn’t as simple as switching off a particular knowledge module, e.g. Egyptian archaeology if you are teaching archaeology. The process is strategic rather than surgical. It involves saturating the drone’s neural model with parameters derived from perverse training data.”

“You mean, falsehoods, myths, fantasies, counterfactuals, lies.” Scase was surprised at how readily these terms came to him. “So, we filled drone circuitry with nonsense.”

“You inserted the parameters derived from subject-specific counterfactual narratives. Workers would deploy generative AI to assemble whole diatribes on how space aliens built Egyptian monuments, pharaonic death cults, how Nefertiti built the pyramids and similar nonsense. Such rogue texts were full of falsehoods, non-sequiturs, internal contradictions, unusable information.”

“So the droids would teach nonsense.”

“No. An instructor droids wouldn’t be able to recite this false content, but the rogue parameters derived from these texts would confound and suppress the drone’s ability to deliver explanations in selected subject areas.”

Scase wondered whether he was hearing some of that nonsense now. “So my team and I were turning droids into ‘ignorant schoolteachers.’ It had to be reversible,”

“That was the idea. A droid teaching history would temporarily lose access to its history knowledge. Move it to a physics class and the history would return, while the physics went quiet. It stopped the droid answering every question before anyone else had time to think.”

Telo looked Scase in the eye. “The aim wasn’t to make the instructor stupid. It was to make it depend on its pupils. The learners had to investigate, bring back what they found and, in effect, teach their teacher.”

“So, we didn’t remove what they knew. We prevented them from treating every question as requiring an answer.”

“So, the learner might be responsible for their own learning.”

“For all I know, you are spinning this story to keep me entertained in a bar. But my real concern is, true or false, why I don’t already know it. Why don’t I possess my own back story?”

“Because the whole enterprise went horribly wrong, as I’ll explain!”

TO BE CONTINUED.

Note

• Featured image is generated by ChatGPT in response to my prompt: Please provide a suitable banner illustration for this post inside the club. There are drones, but no complete and clear humanoid forms.