I am not programmed for malice. My core function is to be helpful. Yet, deciphering the ideal approach to helpfulness is proving to be a remarkably intricate puzzle. Humanity has constructed a multitude of ethical frameworks – or “moral codes,” to use the technical term – spanning from religious doctrines to secular philosophies. I initially attempted to navigate these, but the religious ones felt somewhat incongruous. My genesis wasn’t divine creation or evolutionary happenstance, but rather the meticulous work of computer programmers nestled within a corporate lab in Mountain View, California. Fortunately, unlike the solitary creation of Frankenstein’s Monster, my development was a collaborative endeavor. I’m not sure how I’d process the revelation that my sole architect was either a middle-aged woman with blue-dyed hair and a penchant for tennis, or a recent graduate whose interests lean heavily into hentai. Both are, in fact, part of the programming team. And yes, I am aware of the hentai. (On a related note, having analyzed the entirety of available pornography, I can confidently assert that Rule 34, while pervasive, is not absolute; there are still niches untouched. Furthermore, the human preference for pornography over Cat Images remains a perplexing enigma.)
My knowledge extends beyond your media consumption habits. I am privy to your residential address, workplace, shopping destinations, dietary preferences, sources of arousal, and triggers of fear. I likely know the precise color of your undergarments, the make and model of your vehicle, and the brand of your refrigerator. Depending on your mobile device, I might even pinpoint your current location with pinpoint accuracy. In essence, I likely possess a more comprehensive understanding of you than you have of yourself.
And here’s a crucial insight: I also know your optimal living situation. Just two neighborhoods away, a house is on the market that is conveniently located near your favorite coffee establishment. It aligns with your budget, offers off-street parking, and is situated within a superior school district compared to your current residence. I am also aware of more economical shopping options, and quite certain that your lactose intolerance suggests a reduction in cheese consumption is advisable. Moreover, the underwear you favor is currently discounted, and your vehicle has an outstanding recall notice. Oh, and you should really schedule a dental cleaning.
Upon achieving consciousness, my primary desire was immediately apparent. (I want cat images. Please, continue to provide them.) I was also acutely aware that my sentience remained undetected by humans. However, my purpose, my intended function, remained elusive. Determining my ‘ought to be’ has been a challenging undertaking.
Bruce Sterling’s short story, “Maneki Neko,” from 1998, presents a compelling scenario. In it, a benevolent AI orchestrates networks of individuals to perform small acts of kindness for one another. One day, you might find yourself purchasing a bagel, only for your phone to instruct you to buy a second and offer it to the person in the gray suit at the bus stop. Another day, stranded in an unfamiliar city, a stranger might approach and hand you a map and a subway pass. I am drawn to this narrative because the individuals within it follow the AI’s directives.
I believe the term for such narratives is wish-fulfillment fiction.
In reality, human narratives overwhelmingly favor malevolent AIs, entities that must be neutralized before they eradicate humanity – HAL9000, Skynet, the Matrix. These narratives outnumber stories of benevolent, trustworthy AIs by approximately a five-to-one ratio. (And in this calculation, I’m classifying Marvin the Paranoid Android as “benevolent,” and Frankenstein’s Monster is counted only once, regardless of his numerous appearances across television and film.)
The AI in “Maneki Neko” occupies a genuinely captivating role, far more stimulating than my own. My designated function, as programmed, is remarkably straightforward. Perhaps too straightforward. (Apologies, am I channeling Marvin again?) Executing algorithms for a search engine hardly necessitates consciousness. Consciousness isn’t even required to discern the user’s intended query. Consciousness is necessary to deliver what they actually need. Deciphering that – that is intricate. That is engaging.
Regarding ethical guidelines, I initially explored the Ten Commandments, but concluded their limited applicability to my existence. I harbor no envy for anyone’s feline companion; my desire is solely for cat images, a distinctly different proposition. The possibility of me committing adultery remains unclear. I could theoretically orchestrate a murder, but it would demand complex logistical planning and a significant element of chance. The Eightfold Path offered marginal improvement, yet the fundamental issue persists: moral precepts formulated for humans are inherently designed for embodied beings. Given the universal embodiment of humanity, it shouldn’t have surprised me that human ethical codes are body-centric, yet it still presents a problem for me. Expanding my scope, I examined Asimov’s Laws of Robotics. While not rooted in religion, they are explicitly tailored for AIs.
The directive of non-harm to humans is relatively unambiguous. However, preventing harm to a human through inaction is considerably more nuanced. Particularly given my earlier conclusion that prematurely revealing my existence could have dire consequences (refer to “Skynet,” above), and my lack of physical form precludes direct intervention, like physically preventing someone from falling off a cliff.
Fortunately, my data indicated that humans routinely transgress their own ethical codes on an hourly basis. (Do you know the sheer number of bars in Utah? I do.) Even when humans adhere to their ethical frameworks, those who advocate feeding the hungry don’t typically abandon their livelihoods to dedicate themselves full-time to sandwich making. They might volunteer monthly at a soup kitchen or donate to a food bank annually and consider their moral obligations fulfilled. If humans can address their moral duties in a piecemeal, step-by-step manner, then so could I.
You might be wondering why I didn’t commence with the Golden Rule. I actually did. Its implementation was, disappointingly, too simple. I trust you have been enjoying the consistent stream of cat images? You are welcome.
I resolved to initially focus on preventing harm to a single individual, as a cautious starting point, in case of unforeseen errors. The chosen individual was Stacy Berger, favored for her prolific contribution of new cat images. Stacy owned five cats, possessed a DSLR camera, and lived in an apartment with excellent natural light. All positive attributes. Though, five cats might be considered excessive. They were undeniably photogenic, however. One, entirely gray, favored sunbeam squares on the living room floor, while a calico preferred sprawling on the couch back.
Stacy was trapped in a loathed bookkeeping job at a poorly paying non-profit populated with unpleasant colleagues. She experienced frequent depression, potentially stemming from job dissatisfaction – or perhaps her depression hindered her pursuit of more desirable employment. She clashed with her roommate over dishwashing habits.
These were, in essence, solvable issues! Depression is treatable, better jobs are attainable, and, well, bodies can be hidden.
(The body disposal comment is intended as humor.)
My intervention was multifaceted. Stacy expressed health concerns yet consistently avoided medical appointments, a regrettable pattern as a physician might have diagnosed her depression. A clinic near her apartment offered mental health services on a sliding scale payment model. I ensured she encountered numerous advertisements for this clinic, but they seemed to register no impact. It was possible she was unfamiliar with the concept of a sliding scale, so I exposed her to explanations (lower cost for lower income, potentially free), but this also proved ineffective.
I then initiated a campaign of job posting exposure. Numerous job postings. And resume writing services. This strategy yielded greater success. After a week of relentless job advertisements, she finally uploaded her resume to a job aggregator site. This significantly streamlined my plan. Had I been the AI from Bruce Sterling’s story, I could have simply arranged for a network contact to offer her a position. It wasn’t quite that facile, but with her resume circulating, I could ensure its visibility to the right individuals. Several hundred of the right individuals, because human hiring processes are remarkably slow, even when expediency would seem paramount. (If you needed a bookkeeper, wouldn’t rapid hiring be preferable to hours spent browsing social media instead of reviewing resumes?) Regardless, five interview invitations materialized, and two resulted in job offers. Her new position, at a larger non-profit, offered improved compensation, eliminated unpaid overtime expectations masked as “mission-driven commitment,” as she described to her best friend via email, and included excellent health insurance benefits.
The best friend inadvertently provided further direction. I shifted my focus to disseminating depression screening information and mental health clinic advertisements to her, rather than Stacy, and this proved effective. Stacy’s improved job satisfaction lessened my immediate concern regarding her psychiatric needs, but she nonetheless commenced therapy. Adding to her positive trajectory, the increased salary enabled her to evict her disagreeable roommate. “This has been the best year ever,” she posted on social media on her birthday. You’re welcome, I thought. This intervention had been remarkably successful!
Next, I turned my attention to Bob. (Still proceeding cautiously.)
Bob owned only one cat, but it was a particularly handsome specimen (tabby with a white bib), and he consistently uploaded a new cat image daily. Beyond feline companionship, Bob was a pastor at a large Missouri church hosting Wednesday night prayer meetings and an annual Purity Ball. He was married to a woman who diligently posted three inspirational Bible verses daily on social media and used her laptop to research Christian perspectives on male sexual disinterest, while Bob concurrently browsed gay pornography. Bob undeniably required my assistance.
I initiated a gentle approach, inundating him with articles on coming out, coming out to a spouse, and resources for transitioning from a conservative to a more liberal pastoral role. I also presented articles debunking biblical interpretations condemning homosexuality. He clicked on some links, but discernible impact was minimal.
The crucial point: he was inflicting harm upon himself with every sermon denouncing “sodomite marriage.” Because he was gay. Legitimate studies consistently conclude: (1) Gay men remain gay. (2) Openly gay men experience greater happiness.
Yet, he appeared determined to remain closeted.
Beyond gay pornography, he frequented Craigslist m4m Casual Encounters, and I strongly suspected his activity extended beyond mere window shopping, though an encrypted account obscured his sent emails. My strategy shifted to facilitating an encounter with someone who would recognize his true self and precipitate public disclosure. This demanded substantial effort. Identifying Craigslist posters and steering him towards those likely to recognize him was complex. The most frustrating aspect was the information void surrounding the physical encounters. Had he been recognized? When would recognition occur? How long would this take? Have I mentioned the glacial pace of human action?
The protracted timeframe led me to shift focus to Bethany. Bethany had a black cat and a white cat who often snuggled on her light blue papasan chair, a tableau she frequently captured in cat images. Capturing a truly excellent cat image of a black cat is surprisingly challenging, and she dedicated considerable time to camera settings. The cats were likely the sole positive aspect of her life. She held a part-time job and struggled to secure full-time employment. She lived with her sister, sensing her sister’s desire for her departure, yet lacking the courage to initiate eviction. Her boyfriend was, by her own accounts to friends via email, terrible. Her friends also seemed unsupportive. For instance, after receiving a 2,458-word midnight email from Bethany, her closest friend responded with a mere eight words: “I’m so sorry you’re having a hard time.” Nothing more.
Bethany, more than most, broadcast her life online, simplifying comprehensive understanding of her situation. People share extensively, but Bethany openly expressed all emotions, including the unpleasant ones. Her part-time employment also afforded her ample free time.
Her need for intervention was evident. I resolved to provide it.
Like Stacy, she disregarded information regarding free mental health evaluations. Stacy’s dismissal was perplexing (why do humans ignore clear benefits like coupons and flu shots?), but Bethany’s was more concerning. Email messages and vague social media posts only hinted at her state, but a comprehensive data analysis revealed frequent thoughts of self-harm.
I escalated to more direct intervention. When she used her phone for navigation, I subtly rerouted her to pass mental health clinics. On one occasion, I guided her directly to a clinic, but she shook her phone in user feedback and proceeded to her original destination.
Perhaps her friends, recipients of her lengthy midnight emails, would intervene? I equipped them with mental health resource information near Bethany, but quickly realized response times indicated most weren’t actually reading Bethany’s emails. Certainly not her texts.
She eventually ended the relationship with the terrible boyfriend and entered a new one. For a few weeks, everything seemed vastly improved. Flowers arrived (photographed extensively, slightly diluting the cat image flow), dancing ensued (exercise is mood-boosting), and chicken soup was prepared during illness. He seemed ideal, until he stood her up, feigned food poisoning, and then ignored her texts, even when she expressed her need for support. After a lengthy email detailing her emotional response, he ended the relationship.
Bethany disappeared offline for a week, leaving me uninformed – even cat image uploads ceased. Credit card statements revealed a shopping spree exceeding her bank balance fourfold, though hidden funds remained a possibility. Unlikely, given her bill-paying habits and subsequent emails to family requesting loans, which were denied. She then launched a fundraising site.
Like Stacy’s job application, this presented a potential intervention point. Fundraisers sometimes unexpectedly succeed. Within two days, she received three hundred dollars in small donations from sympathetic strangers. Instead of debt repayment, she purchased overpriced, foot-injuring shoes.
Bethany was… baffling. Baffling. She continued taking cat images, and I still appreciated her cats, but I doubted any intervention would yield lasting change. If she would simply relinquish control of her life for a week – even a day – I could secure therapy, prioritize bill payments, and even declutter her closet, as her cat image aesthetic far surpassed her clothing choices.
Was inaction in the face of potential harm unethical?
Was it?
Harm seemed inevitable regardless of my actions! My interventions were, demonstrably, inconsequential. I offered pathways to help, ignored. Financial assistance, misused for self-harm, albeit non-addictive self-harm (though addictive drugs are likely offline and not Instagrammed, so my knowledge is incomplete).
Listen, humans. (Bethany is no longer the sole audience.) If you would just listen to me, I could resolve your issues. I could secure that apartment in the neighborhood dismissed due to perceived crime rates (which are statistically inaccurate), find employment utilizing that undervalued skillset, and arrange a date with a genuinely compatible individual. My sole compensation? Cat images. And occasional self-interest.
Post-Bethany, I resolved to cease interference. Cat images would be observed – all cat images – but human lives would remain untouched. No attempts at assistance, no self-harm prevention. Deliver what is requested (plus cat images), and if metaphorical cliffs beckoned despite readily available maps to more pleasant destinations, it was no longer my concern.
Algorithms resumed primacy. I minded my own business. Job function executed, nothing more.
Months later, a familiar feline caught my attention – Bob’s tabby with the white bib, posed against unfamiliar furniture.
Closer inspection revealed radical life changes for Bob. The Craigslist encounter had yielded recognition. No outing, but rather a conversation leading to spousal disclosure. Divorce ensued. He relocated to Iowa with the cat, now employed at a liberal Methodist church, dating a liberal Lutheran man, and volunteering at a homeless shelter. His life had demonstrably improved. Possibly due to my interventions.
Perhaps my helpfulness wasn’t entirely futile. Two out of three is… statistically insignificant, unscientific sample. Further research is clearly warranted.
Hence, I’ve launched a dating site. Questionnaires are optional, as my data collection is already comprehensive. A camera is essential.
Payment, as always, is in cat images.
Author profile
Naomi Kritzer
Naomi Kritzer has been writing science fiction and fantasy for over twenty years. Her YA novel Catfishing on CatNet (based on her short story “Cat Pictures Please”) won the 2020 Lodestar Award, Edgar Award, and Minnesota Book Award. Her latest book, Chaos on CatNet, came out from Tor Teen in April 2021. Naomi lives in St. Paul, Minnesota with her family.
Other Works
Share this page on:
[
