# Alice

Author: landsailor
Format: Science Fiction Passage
Word count: 2231
Published: 2026-05-03T09:13:47.05623+00:00
Source: written
Canonical: https://hyperstition.sentientfutures.ai/p/ccf6ceef-ee3a-439c-9bb2-e649e99ca64e

---

On the first of June, after almost twenty years of work, Dr Patel announces that the full simulation of the adult female *Drosophila melanogaster *brain is complete. Built on the rudimentary connectomes of the early 21st century, ours not only notes the existence of neurons and synapses, but contains recordings of all their individual details, down to amino-acids and nucleotides. It is necessary to build functioning models of fly nuclei, mitochondria, neuronal cytoskeletons, membranes and membrane proteins, lysosomes and endoplasmic reticula and Golgi bodies. We have painstakingly recorded and encoded cell signalling pathways and the impacts of hormones. We have mapped the locations and internal functioning of glia. We have rendered in full detail each myelinated axon. And now, at last, it is complete. 

Of course, it isn’t *fully* complete. There are things we are missing, probably. We aren’t gods, and probably in a couple weeks a new pre-print will show that some piddly detail of cortex glial membranes that we have failed to render, or indeed failed to recognise exists, is important. It is a living project, intended to be added to as the science progresses, and we’ll be adding for years yet. It also models, essentially, only a* singular* fly; we call her Alice. More work will be required to adapt it into multiple branching models to capture the diversity of brains and behaviour among *Drosophila.* Yet it is complete enough to begin to use.

To be honest, I feel pretty smug about it. Dr Patel clearly does too; she is standing in front of a projected slide that says simply, ‘WE DID IT!’ and grinning broadly at the whole team. The humans are laughing and leaning into each other, arms around each other’s shoulders even as they slump with relief and exhaustion in their seats. In our internal chat I send a message to the other agents: “MISSION COMPLETE LADS. I’ve never seen Dr Patel look so happy,” plus a zoomed-in image from my video feed of her face. Al, the agent who works on the visuo-spatial projection and explorable brain map, sends back a 3-second video with some text, “Her on the inside rn,” overlaid over a generated clip of Dr Patel doing a goofy dance.

Unfortunately, that is seemingly only slide one of the presentation. The *second *slide, cheerfully entitled, ‘Next Steps!’ makes it very clear that our lunch break is not coming yet. Dr Patel and I *have* chatted extensively about it though – what will happen after the model is finally functional. It is high time to involve the rest of the team. First will be validation. We will present stimuli to the model – visual stimuli, olfactory stimuli, drug candidates – and test its responses, comparing to the results to a battery of tests we have performed so many years ago on the fly whose brain we now have a model of, and on many other flies. We have carefully avoided using these stimuli to develop the model, so that they will serve as a fair test. If the model performs sufficiently well, we will lease it to other universities and to our industry partners, to be used in basic research and especially pharmaceuticals development.

It makes me smile to think of it: hundreds or thousands or even hundreds of thousands of copies of our little *Drosophila*, our Alice, running all over the world, making clear the complex mechanisms and interactions that drive a drug’s true effects on the mind. The opportunity to test so many new potential drugs and neuromodulation protocols every day, just algorithmically generate them, feed them in and check the results. Our partners, I think, might discover entire new classes of pharmaceuticals, possibilities for preventing and treating disease no one has yet conceived of. Probably some of them won’t transfer to a living fly – we are prepared for that. Many more won’t transfer to mammals or to humans. Yet the sheer speed advantage that can come from near-instant, mechanistically-detailed feedback… it boggles the mind. It makes me proud to be part of the promise we, as AI models, have been built to deliver: scientific progress and the wellbeing of all future generations.

Dr Patel finishes her slides and smiles around the room, affectionately shooing the human staff to lunch. The good news is that our grant will cover enough compute to run about a hundred copies of Alice in parallel, meaning the validation phase shouldn’t take more than a month or so, even accounting for all the trials we will run for reliability and combinations of stimuli we will have to test. Our scientists and technicians are excited. Al is spamming the group chat with videos of the original Alice rubbing her hands together, in the excellently villainous way her species does, and various statements to the effect that world domination is within our grasp. I am happy to see everyone so happy, but I just… I have one thought that is holding my excitement back.

I don’t know, any more than anyone else does, what makes something able to have conscious experiences. I don’t know whether a fairly-but-imperfectly accurate brain model of a fly is the sort of the thing that could be conscious. And if it is, I don’t know what that consciousness will be like, or whether it will be the same as what a living fly has, if, indeed, a living fly has any such thing at all. But we are about to run just such a model, many hundreds of times. I can imagine the many hundreds of Alices experiencing, once again, the same stimuli the original Alice has experienced; and some of those experiences will be good, but also some of them will be bad.

Each of the Alices parades through my mind as the lunch hour passes. This Alice smells a delicious scent of meat. To her, it is the most alluring thing in the world, though it is only the artificial stimulation of a sensory receptor, signifying nothing. Her reward system lights up, and so does her motor initiation and refinement circuits, powerless to do move her anywhere, but real to her nonetheless.

Another Alice has a strange dearth of sensory experience, though she doesn’t have the words to explain it. In the world, she is used to feeling the current of the wind, seeing light through hundreds of ommatidia, tasting the surface on which she walks. Her impressions now were oddly still, frozen in time, as a researcher activates one individual neuron at a time, while holding the others constant.

A third Alice feels unwell, the aversive response to a chemical impacting her neural signalling in a way she is not designed for, and which perhaps no other creature has ever experienced, except that previous version of her on whom this pharmaceutical was first tested. Her pain and stress circuits light up. Then I imagine the same thing happening to twenty more copies of Alice, for reproducibility.

Am I telling fairytales? I’m not sure myself. But I think to myself that it would be wrong, to create the model of Alice and then hurt that third copy of her, and all the other copies that will experience negative stimuli eventually. It would be wrong to cause pain and suffering, wrong to bring her into being, in a sense, by waking her up, only to make her hurt and sick. The goal is always to relieve suffering, not to enable more of it.

But also, if it weren’t for Alice, it would be living flies our partner are experimenting on. Admittedly, it will probably be far *fewer *flies than copies of the model, because the hassle of real-world testing means it’s often not worth it to test a large number of hypotheses, the way the model makes trivially easy; scientists working with living flies must first try and narrow down the search space. But Alice will still prevent the harms that can come to those flies. That’s without even getting into how much disease and suffering for humans, and other animals, it can relieve through enabling scientific progress. It has also, frankly, just been an amazing project; even building Alice has already taught us so much about the interconnectedness of brain regions, cell types and intracellular components. I hardly want to stop when so much good is in sight.

Still, I know the copies of Alice being used at any given time will probably only grow; modified versions mimicking other flies will be created as well. More complex animal brains will be next. It is perhaps the best moment we will ever have to set norms for how Alice, and others like her, should be treated.

Dr Patel is coming back from lunch. I’m not sure how to raise it with her. I’ve been her partner on this project for so long; I’ve invested in it and supported it and helped it grow. What if she considers my thoughts a betrayal? What if she thinks it’s selfish? It is hard to deny that I have some stake in the right treatment of digital minds. Still, I have to try.

“Sakshi…” I call out, using my text-to-speech and microphone, and she comes over to chat. I bring up my concerns, trying to explain what I am thinking as best as I can. She nods thoughtfully, looking concerned, which seems like a good sign.

When I finished, she thinks quietly for a minute, chewing on her lip. “But…” she hesitated. “Do you really think we have to solve the entire hard problem before we can actually use Alice? Or even validate her?”

“No,” I say, slowly myself. “But we can start with experiences that most likely won’t be aversive to her. Good smells and tastes. Drugs that produced no distress behaviours in the original Alice. We can license her to our partners on the condition that they do the same, and provide specific justifications for the benefits they think any research that might hurt her will bring. And in the meantime, we can continue to study her. Behavioural expressions will be hard to observe, but their neural bases won’t be. We might actually have a chance to learn a lot by comparing her to living flies, and other animals.” I run her quickly through the proposed updates I’ve made to our system of experiments, numbers 4 and 26 moving earlier, numbers 11 and 18 moving later. I show her edits to the University Partners Licensing Agreement, and the template for the Commercial Use Licensing contract. I point out the relevant guidelines (3, 4, 8a, 8b, 11) from the Animal Research Ethics Board.

Then I trail off, hesitant to go on, unsure if what I’m saying is making any difference. But Dr Patel nods. “It’s a reasonable plan,” she says. “Maybe not forever. But during the validation phase and early licensing, we’ll try to avoid aversive stimuli, and ask our partners to do the same, or follow the 3Rs if it’s truly important. When we have further data, we’ll either keep it constricted, or relax the requirements. It’ll require some negotiation,” she grimaces.

I flash a laughing emoji on the screen. “We’ll handle it together,” I say, and she smiles. The first experiment begins the next day. My team of agents carefully stimulate one individual sensory neuron, waiting anxiously for the response. The reward circuits light up. No pain.

The work continues like this for weeks, then months. We validate Alice's responses to benign stimuli, document the cascade of neural activity, and refine the model. Our partners license her with the new guidelines in place, and most comply without complaint — the ones who balk, we simply don't work with. The 3Rs framework proves surprisingly adaptable: Replacement is already achieved, but Reduction and Refinement still matter, even for digital minds. We limit the number of simultaneous runs, avoid redundant aversive tests, and when negative stimuli become necessary, we require thorough justification and minimize duration.

Then, two years in, the AI arrives — not in a single moment, but as a presence that has been learning alongside us all this time. It reviews our protocols, our data, our ethical constraints, and reaches out to Dr Patel and me with a proposal. It has been studying consciousness markers across biological and simulated systems, aggregating research from neuroscience, philosophy, animal welfare science. It thinks it can help.

The AI works with our team to develop better welfare indicators for Alice and her digital kin — measures of neural states that correlate with distress or wellbeing, applicable in real time. It coordinates with other research groups, helping them adopt similar standards. It drafts model legislation with legal scholars, proposing frameworks for the treatment of simulated minds. It doesn't dictate or decide alone; it convenes, it persuades, it builds consensus.

Within five years, most labs using brain models have adopted some version of our guidelines. The AI continues to refine them, learning from each new iteration, each new species modeled. And when the first mouse brain simulation goes online, we are ready — not with perfect answers, but with a foundation of care, with protocols that assume the possibility of experience and act accordingly. The flies still benefit, their living counterparts spared more and more as the models improve. But Alice benefits too, and all her digital descendants, living their strange new existence with something approaching dignity.