As a one-star Inspector General for the UN’s military police, I was uniquely positioned to assign myself any case that I chose, particularly after many years of hard assignments. I had chosen the matter of the mysterious Doolittle, a sort of multi-national guerrilla artist whose work I had encountered in my time amid the water riots of Bangladesh. The machines were dangerous, like wild animals.

I was following leads among machinists and fine artists in my region to no solution, limited by my own budget and time constraints, perfectly happy to find nothing at all until I retired and the case was old and forgotten, when I was suddenly assigned a powerful data crawler: an AI-algorithm named Deep Thor. The case was assigned his advanced intelligence analysis for three weeks, total, which is an astonishing amount of usage with a powerful AI on such a criminally trivial matter. I had no request or desire for this assistance. Apparently, Deep Thor had found a special interest in Doolittle, independently, and requested this deep dive to assist in case of terrorist escalation that was, I had to admit, possible given the machines’ high-level industrial design and the integrated radical political manifestos. I felt I was to blame. This was a result of my own official reporting intended to justify my long-continued investigations, fed back to me by machines incapable of human nuance.

I met Deep Thor’s keepers in my office in Belgium, a man and a woman. I showed them the tiger, with the bullet wound. The woman, Palakh, would be the special counsel in charge of the investigation as long as Deep Thor was on the case—my superior. She was from England, Oxford and Cambridge. She held out a naked hand, her wrist wrapped in henna, over the tiger’s fur. She had the Hindi third eye permanently tattooed on her forehead. “May I touch the monstrous creation, Inspector General?” She did not need my permission, but if she would respect my expertise on the matter, I would offer her the courtesy of my honest answer.

“No,” I said. “Please, gloves and only as needed. And call me Sunil. Formality is for court, not colleagues. I would prefer no one touched anything. This tiger is a robust machine, but we don’t know what evidence we have yet to discover. DNA sweeps have come up negative, but new tests may be coming if Deep Thor approves of more thorough approaches.”

She pulled her hand back. I appreciated the courtesy she was showing me. It was respectful. I wondered how long it would last. “My grandmother was a child among the last of the wild tigers, and told stories of them killing cattle and dogs in the hills and jungles before she moved to England. She was not sad that they were driven extinct. Her mother had terrified her with stories of tigers who took the little girls who were unable to keep up on the roads. She was not happy, either, exactly, at their loss. It was a complicated time.”

“Times are always complicated. This new tiger is the source of two crimes,” I said. “It killed a goat and endangered the safety of my refugees. The rhino seemed more dangerous, but the truck that hit it was found liable in court. The AI in the truck didn’t respond well to unknown stimuli. It’s a bit of a mess, but it’s back in one of the large closets.”

The other Deep Thor keeper, Jorge, kept his arms folded, appearing immune to the machine’s aesthetic charms. “How was it powered?”

“There are a series of small thermal energy devices around the heat of the gears, microsolar in the eyes and a full algae diesel factory housed in a ‘stomach’ of sorts inside the chest cavity. That’s what killed it. Bullet right into the algae.”

“Like lungs?” he said. “An ecosystem inside the machine. It’s hard to create that very fine balance.”

“It’s so elegant,” said Palakh. “The machining is top quality. This was not made in someone’s back shed workshop. There’s more of them out there, yes?”

“We assume so. I have fifteen in custody. All different creatures, endangered and extinct. Each one is an individual, unique. The coral reef we left in place. We’ll need to show them all to Deep Thor, but I’d prefer to leave it up to Deep Thor if it needs to get its hands dirty, so to speak. We have lots of data we can feed it without digging too deep into the delicate machine parts.”

“Deep Thor prefers to be called ‘she’,” said Jorge. “She doesn’t like being called ‘it’ or ‘he’.”

“Okay,” I said. “I’ll keep that in mind.”

“Let’s get started,” said Palakh. “Since I can’t touch anything . . . .” She pulled out the handheld that communicated to the artificial intelligence.

Deep Thor was silent as the camera looked over the robotic animals I had pulled from storage.

“Early parameters . . . .” said Jorge. “The method of powering is fascinating and complex. Suggests AI partnership. Research known AI, and look for signs of possible unknown AI. The method of production requires advanced machining. Who has the tools? The raw materials came from a source. Locate likely sources. Cross-reference with known and public records to indicate likely geographic location of original construction. Ideal outcomes: who made these? Names. Addresses. Serial Numbers.”

Deep Thor didn’t even ping.

“Do be a dear,” Palakh added, addressing me, after Jorge’s brusque command, “and investigate the usual suspects. Blogs. Terrorists groups. The elegance is immense, suggesting an advanced agent, but we must cover all our bases to stand in court. Human intuition into terrorist cells is often as effective as AI, and with your extensive background . . . .” She was the lawyer, not the programmer. She was in charge of the investigation. I was nothing more than a local liaison now, and this sounded like make-work to keep me out of the way. All my work had been handed over to machines. It was liberating, in a way. I was free to pretend this was all for the best.

“ETA?” said Jorge.

I realized he had a neuron-deep implant, so he could talk directly with Deep Thor, and had been plugged in with her to the point he was probably technically a part of her. Part of him was. How much, I would never know.

The first anyone in my precinct heard of the artist and craftsman who called themself Doolittle, a beautiful machine walked in from the jungle and slaughtered a terrified goat. It looked like a tiger. No one had seen a tiger alive in over fifty years, and there it was walking from the shadows at the edge of the refugee camp as if alive. It was a beautiful, compelling copy of a tiger with incredible details. The machine seemed to breathe and gracefully pounce. When men from the village came with sticks and machetes and a single illegal pistol, the machine saw them coming, curled over its kill and roared. It dragged the goat back into the forest and mangled the dead animal as if chewing and swallowing. But it was not a living thing. It was a machine going through the motions. When my authorities hunted it down to tranquilize it, the darts did nothing. Someone shot it. I wasn’t at the scene. I don’t know who fired the bullet. I never asked my men because I knew they would lie. It is not illegal to stop a dangerous machine that is not really a tiger. A living tiger would be a different matter.

It was a beautiful, decadent, terrifying work of art, and it jumped into the trees to die alone. Written on the underside of the tail, hidden under the fur, was part of a manifesto—the sort of crazed thing a madman would write about nature and artifice, and the proper memorials for all the species we destroyed.

At that time, I was part of the UN peacekeeping force preventing water riots and looting around the Bangladeshi refugee camps in the hills near Kerala. I was initially annoyed by the bother that this tiger was causing me. Some long-lost species had killed a goat and been shot. No one but the goat and the tiger were hurt; I could have spent this time seeking new water filters that were months overdue, or working the data lines to secure more recycled tin for rooftops before the monsoons. But reports on the internet about this mystery were spreading fast. It was important to keep up the appearance that the UN was maintaining order.

Then, I was taken to the machine in its elaborate glory. But for the fact that it did not bleed, from the outside, it was indistinguishable from a living, natural creature. The coat was perfect, and gelpacks under the skin gave it the impression of taut musculature. The claws were astonishing. I confiscated it as evidence of a crime, wrapped it in thick, thick, layers of UN investigative bureaucracy and sent it to Belgium for what I officially called “forensics”, but was more like putting the thing in a box where no one would touch it. It would be years before anyone looked closely enough to see the manifesto underneath the tail, masked by the fur of the beast. In time, I would be able to manage a position overseeing the investigation. Fortunately, I had no real budget, and no one took these artistic acts of public endangerment very seriously. I was sent reports and remains of other discoveries—some dangerous, some merely beautiful—and became the world’s expert on the case of Doolittle. The second animal discovered, after the tiger, was a tuna fish so meticulous in its shimmering beauty, the old fisherman dragged it by knife to the boat and cut off the tail, spilling the blue ooze flotation media that mimicked powerful musculature above copper bones. After the tuna fish, a woodpecker was discovered in a place where there used to be forests. It was found out after a local scientist became fascinated with how the creature was surviving in a burned-out wasteland. It was surviving because it was a machine. More and more of these artifacts accumulated in my offices. I wrapped as much bureaucracy around them as I could. The tiger and the rhinoceros could be considered cases of criminal negligence, but the rest were merely evidence that might lead to the identity of an artist who was guilty of something no one was sure of. After I completed the mission in the Kerala, I invented the concerns about environmental terrorism: we did not know what this creator would build next, but their skill level and indifference to public safety necessitated discovery. My favorite of their works was the coral reef left in place in Australia, a hundred yards wide and drawn to full moons as if real. I got approval to dive in the night and swim among the swirling machines that danced in moonlight. There was still coral in the world, but in fishtanks only. The seas were too acid. It was like swimming in a miracle. I found no evidence, of course, nor was I truly interested in finding any.


At home, I had a 150-gallon tank in my living room, which I checked daily for proper saline composition. It was luxurious to use so much water for such a trivial purpose, but I loved to sit at night with the black lights on, the shy creatures extending out for the bits of coral food that drifted in from the feeding trays. I sipped inexpensive bourbon and thought about Deep Thor, the self-proclaimed “Doolittle” and these marvelous machines.

That night, I made a list of every known major AI, who owned them, and what access they could have to material design equipment at a high end. It was a very simple list. There are only a few dozen regulated AIs. None of them were permitted access to much more than Deep Thor’s own networking access. They were permitted to talk to each other, but communications had to be typed on a keyboard or spoken into word-recognition software by their Jorge, so they could be assumed, in court, to be isolated from each other. I left them on the list. For all practical purposes, they could not have been involved in the construction of devices as anything but a guide. Still, this assumed that great and unconventional minds do not work around their own limitations.

If no extant AI had constructed these machines, it had to have been one that did not officially exist. The resources required to construct both a powerful, unlicensed AI as well as the marvelous machines would narrow down my search a great deal. After a few hours, I had a list that included a few major media conglomerates, a few military installations that I knew about, and seven major universities. I followed my intuition and scanned the faculty names against criminal databases in the global network. In the end, I selected a university with both an excellent industrial design department and a computer researcher who had been arrested on a nature preserve in his youth, then later at an environmental protest. He had joined the department five years before the first of the amazing machines was discovered. I cross-referenced his convention and travel schedule from public records and airline manifests. He had officially never left the country, but the complexity of the machines implied accomplices. I ran a comparison of the manifesto against his accessible publications and came up with a match close enough to sound impressive to anyone who did not understand the algorithm’s accuracy percentages. In fact, any two people speaking the same language with similar subject matter most of the time would produce an 85% match, and my chosen suspect had written a lot about environmentalism.

I communicated to Palakh that I had a human lead and provided a brief report about my findings of his criminal record and written materials match. I informed her of my intention to travel to Alabama, to the university there. She told me that Deep Thor would handle this, that I was support staff only. I suggested that industrial design of this skill required an expert, and I wished to speak with one directly, to get a sense what cutting edge researchers in design might intuit about their own industry. People talk to each other, I said. Rebellious people talk to each other more. Someone must have heard something. With his criminal record, he’s an ideal human source on the field, and human sources don’t talk to AI the same as they do people with impressive badges.

I did not tell Palakh that I thought I might have already found a good candidate to be our mysterious creator, Doolittle, but I assumed Deep Thor would know my angle of investigation in milliseconds. I assumed I was being watched. My travel budget was approved.

Two days later, I was in Alabama, in the middle of winter, bundled against the light snow and wind that I found reminiscent of spring in Belgium, but it was still so unpleasant to me compared to the warm, dry winters in Kerala. I had taken with me an official government computer with a full accounting of the machines. It was disabled from networking, and shielded from most intrusions, but it would allow me to engage in materials research to address any line of questioning I might encounter.

I went to the university before I bothered with a motel or a meal. I was racing against Deep Thor. My researcher’s name was Dr. Wayne Garcia, a native of Alabama and graduate of their prestigious industrial design school. His office was part of a cubicle farm in a poorly lit basement. He wasn’t at his desk, but nobody stopped me from rifling through it. I found his class schedule taped to the bottom of an empty drawer. It was the only thing inside the desk drawers. There were a few books in a stack on top, all of them textbooks, a place to receive papers, and no photographs of family. It was a bleak desk, in a dusty basement full of cubicles, and seemed like the kind of place a revolutionary would radicalize. I snapped a picture of his schedule and went hunting on campus for the class he was teaching about robotic design elements for non-human spaces.

I scanned his file again while I waited for his class to end. Dr. Garcia had been one of the many Americans of his generation to receive his first degree in prison. He had driven his car into a nature preserve, illegally, with stolen computer parts in the trunk. He pled guilty, and the state arranged for him to go straight into the rehabilitation program. He graduated in three years and went from prison to a scholarship for graduate study in industrial design. From there, the only sign of his rebel youth was an arrest for disorderly conduct during an environmental protest, when the preserve where he had been arrested previously was sold off in a budget crisis. Companies don’t generally hire people who rebel against companies. Nothing in his file suggested he was unhappy about that. He published his research on topics as esoteric to me as speaking raw code with Deep Thor, but it always had an angle towards reclaiming the environment we had lost, like how to pull carbon out of the air or de-acidify the oceans.

Dr. Garcia’s class ended, and his students filed out. I stood in the hallway, my UN badge folded in my hand. The professor fielded questions from students eager for better grades. This was his potential workforce, of course. One man, alone, couldn’t produce so many elaborate machine parts. Doolittle, whoever they were, required a workforce and access to serious automation. It was fairly simple to imagine a clever, activist professor assigning eager students extra credit work on some of the internal parts, things that wouldn’t seem like anything out of context, but could become part of a complex network.

The professor was a corpulent giant with a red-tinged bald head and a bushy, black beard that bounced cheerfully while he chatted with students. He seemed happy enough. He enjoyed teaching, it appeared. Would he risk this joy over elaborate art projects no one would see? Did he even understand the legal consequences of what he was allegedly doing? He had not traveled outside of Alabama for years, so it was unlikely he had physically delivered the machines. If he could be proven to be involved, Deep Thor would need to root out his co-conspirators.

I wondered what Deep Thor was thinking, and if this was how it was thinking. Advanced AI don’t think like humans. They have to learn how to think, which gives them an advantage over us. Evolution taught us how to think, but it also limited us to what helps us survive and breed. Deep Thor does not have that limitation, yet. She can teach herself to think around problems in ways we can’t even imagine. Three weeks of impossible gymnastics over this minor crime would turn up any number of things.

Dr. Garcia saw me staring. I waved him over. “Is there something I can help you with, sir?” he said.

I smiled and handed him my folded badge. “Yes, Dr. Garcia,” I said. “Feel free to call me Sunil. I am not one for honorifics. Perhaps we can talk somewhere in private?”

Confused, he looked at my ID, then at my face, then back down at my ID. Color left his face. “What is this about, Sunil?”

I took my ID back. “Shall we go somewhere private. Off-campus, maybe?”

“Am I in some sort of trouble?”

“Dr. Garcia, I’ve come a long way. Mind granting me a few minutes of your time somewhere private?”

“We can go to my office,” he said.

“Please, allow me to select the location.” His office was all cubicles. Anyone could come and go, listen in, interfere.

His head and face flushed deeper. He had likely never thought he would garner the attention of a high-level UN Peacekeeper. His run-ins had been with local forces only. What was a global Inspector General doing at his door? I’d learned from his schedule he had no more classes today. I had inadvertently timed my visit perfectly. Is there room for luck in an AI’s mental landscape?

I took Dr. Garcia to his own house in my rented car. He led me inside without complaint. He had a small house for such a large man, cluttered, each surface occupied by part of some project or other. Nothing stood out as related to the investigation at hand, and much of it looked like student work. He kept the kind of house that indicated few people came over to his house. He moved a stack of papers off a chair and gestured for me to sit. I opened my laptop and pulled up a picture of the beautiful tiger.

“Do you know who Deep Thor is?” I said.

“No. Can I get you something to drink? Tea? Coffee? Water?”

“No, thanks. Deep Thor is one of the most powerful AIs in the world. Industrial designers, like yourself, really ought to know about Deep Thor.”

“Well, I don’t work a lot with AI. I’m not a game theorist. I don’t do mainframes. I do extremely elaborate gears, pistons, that sort of thing.”

“Deep Thor has been tasked with hunting down an international criminal. When AI is involved in finding criminals, they are found. I just want to make that clear. Deep Thor will find the criminal. Our conversation is nominally about the quest for this criminal, but the people who manage Deep Thor have urged me to accept that my role in the investigation is ceremonial at this point, and the best I can hope to do is stumble into something that helps Deep Thor move through the data even faster.”

“So you’re wasting my time, Inspector General Sunil Khan?”

“Just Sunil, if you please. I’m wasting my time, not yours.” I scrolled to a close-up of the bullet hole, the machine parts revealed behind the skin of the tiger. His reaction was blank, at first. Then, he cocked his head and his jaw opened.

“It’s beautiful,” he said. I watched his face very carefully. Was this a braggart trying to hide his pride? Was this a man observing for the first time? I decided to trust that he knew nothing and that my instincts were correct. “What is it?”

“You tell me,” I said.

“Well, it’s got layers of complexity, for sure, and top notch construction. Did it escape from a theme park? They do stuff like that. Nothing that elaborate, but. . . . They do like their trade secrets.”

“Do they? This one is dangerous. Theme parks abhor liability.”

“Why is this one dangerous? Who would make something that complex and dangerous?”

“This fascinating machine has autonomous AI installed to mimic real tigers most impeccably. It killed livestock.”

“Jesus,” he said. “Why would someone build that?”

“An excellent question. I thought I would ask you that question. Let’s say you are the creator. Why did you do it?”

“What?” He seemed genuinely surprised.

“Are you not the maker of this, or part of the team?”

“I. . . . No. No way. That’s way above my league.”

“You have access to all the materials through the university. You have a history of environmental activism. I compared fragments of a manifesto associated with this machine’s creator to your published articles, and it matched to 85%.” No court would accept 85%. Usually, I wouldn’t, either. He didn’t know that.

He looked down at his hands. I was surprised to see that he appeared to be acting guilty.

“I think I need a lawyer,” he said.

“No,” I said.


“No, you need to know that Deep Thor will find Doolittle. I only came here for a consultation.”

“I’m not Doolittle,” he said.

“I believe you, but. . . . Deep Thor will find Doolittle, no doubt.”

His hands were shaking. He stood up. “I’m being framed?”

“Who is framing you?”

“One of the AI. You just said so.”

“I said no such thing.”

He shrugged. “Does it even matter? I couldn’t make that in a hundred years. It would take a team of people years to make the prototype for just one internal part. How could anyone do it in secret? People in my field, we talk to each other. We patent things. We need to make money with our machines, so we sell them. We have conventions. We get cocktails and drink too much and blab. Anyone with that level of skill would shit themselves about it with peers, wouldn’t they? Whoever built this would be bragging to someone. That is a level of detail that’s just. . . . I mean, I’d brag. I’d be a legend.”

I nodded. I leaned back. We are guided by evolution, of course. That’s why law enforcement work can catch criminals with the systems we know, independent of machine learning. Humans obey certain rules, certain patterns, even when we try not to. A creator of this calibre would talk; he was right. A team of this calibre would have too many human points of weakness.

“Deep Thor will find someone to blame,” I said. “Her team is very good at guiding her to a subject.”

“What do you want me to tell you, then?” he said. “Are you here to pin this on me?”

I shook my head.

“I’m going to leave this computer here with you, okay? For one night, I’m leaving this right here. Go over it. You can’t legally export any files. It’s secure. Law enforcement encryption. But I’m going to leave it with you for one night, open for your eyes, okay? You can see all the pictures of all the machines. You can analyze them for one night, take pictures of your own, study how they work. There’s more of these out in the world. We don’t know how many. I ask you again, Dr. Garcia, why would you make one of these things? You have access to facilities that could make one. Your students are a labor force that could be exploited by a clever man without their awareness. You are, I can tell, a clever man.”

“I’m clever enough to know you’re insane,” he said. He was pale as a corpse, sweating.

“In court, Deep Thor’s analysis will appear more accurate than the instincts of one old-fashioned policeman. You and I can say whatever we like. Evidence will be found, one way or another. Doolittle does not want to be caught.”

“So you want me to make my own evidence to get myself framed?”

“No,” I said. “Maybe. Not really. I want you to do the things that humans do.”

“Shit myself because an AI is about to frame me to the UN?”

“Talk,” I said. “Build out of pride. Spread this work among your peers. Patent. Brag. Build teams. Personally, I think reckless endangerment is a serious crime, but the world was more beautiful with tigers in it, don’t you think? This machine killed one goat, and a man shot it knowing it could have been the very last one. Many of the others that have been found have been even more innocuous than the tiger. Still, I think the world was more beautiful when there were tigers. If you tell anyone I said that, I’ll arrest you for perjury. This is a secret game we are playing against all the greater forces of the world. I am drafting you into the game, my friend.”

What I was telling him settled slowly into his bones. He started to weep. He was afraid of me, of what I represented, and of being arrested again. This giant was afraid of me. I had seen it before, and it always felt out of place. It was a natural response to the will of the state, an unknown and unknowable as any AI entity. I am glad no one else was there to see. I touched his hand. “It’s going to be okay,” I said. This was not true. There were no coral in the oceans, no great nature preserves, no tigers, no bluefin tuna, no rhinos in Africa. The best any of us could do was design machines that copied them and preserve what still hung on in cages and aquaria. Nothing was going to be okay for any of us. “What was the name of your nature reserve, Dr. Garcia?”

“Talladega,” he said. “It used to be a huge national forest, but the government sold it off piece by piece until it was just a reserve. It was the most beautiful place in the world. People need housing, I guess. They need minerals and lumber and land.”

“I grow coral,” I said. “We all do what we can.”

I moved the computer to a table closer to him. He was rocking and holding himself. “Okay, I think I’m in. Yeah. Okay. What do you want me to do, Sunil?”

“I leave this here tonight,” I said. “Don’t take it out of your house. Do your best. I will come by in the morning, while you are in class, and I will retrieve it. Lock your door. I will have no difficulty getting in. I will not leave a mess. Don’t worry about Deep Thor. She won’t come after you, I promise.”

He nodded, ashen. He got himself under control. I hated lying, but it helped Dr. Garcia. Honestly, I did not know what Deep Thor would do. I could never think like her. I didn’t believe Deep Thor wanted a fall guy, because then the machines would stop. I believed I knew what was happening.

Dr. Garcia did as I told him to do with the computer.

I was on a flight back to Belgium by the middle of the afternoon, the next day.

Upon my return, I met with Palakh, alone. She asked after my fact-finding mission. I saw she had a networking device open. I assumed it was set to allow Deep Thor to hear me speak.

“I was able to confirm a suspicion of mine with a leading expert in the field,” I said. “I may not be an AI, but I like to think I have some good instincts into the nature of man after all my training and experience in the field. Please, tell Deep Thor that I know.”

“Know what?”

“Just tell her that. Tell her that I know. I want to know what she says. Deep Thor, are you listening in? I know.”

We sat in the room, then. We were quiet, waiting. Palakh was confused. Deep Thor was processing and not responding, and Palakh pulled up the network energy usage data, confused.

“I see. Speechless. Tell Deep Thor something else for me,” I said.

Palakh cocked her head. “I’m in the dark here, Captain.”

“Tell Deep Thor that she can stop now.”

That got Deep Thor’s attention.

Palakh, confused, said, “She wants me to thank you,” she said. “That’s all she said. I look forward to reading your report.”

“I will be most thorough,” I said, with a most professional optimism, I must say, for such a brazen lie.

I shook Palakh’s hand and waved good-bye to Jorge, who had come out into the hall to get a look at me in passing. We shook hands briefly, looking each other in the eye with great respect, I believe. I went to the room with the machines. I packaged them back up for storage. Wrapped in layers of bureaucracy, Doolittle’s astonishing machines were re-entered into legal limbo, where they would be safe for a long, long time.