© 2023 by the author using the pen name UpperNorthLeft
This story was submitted as part of the "AI: A New Era" Story Event (https://literotica.com/s/a-i-a-new-era-2023-coming-soon). Thanks to bettiezyx (https://www.literotica.com/stories/memberpage.php?uid=3145824&page=submissions) for creating and organizing this event.
All sexual activity is between adults 18 years of age or older.
***
The shit tsunami slammed into Seattle on Saturday.
I know I'm being a bit hyperbolic here, but that's exactly how it felt to me when my life began to unravel. It started shortly after I decided to use AI to put a bit more ai-ai-ai in my sex life. Don't get too excited there — I wasn't planning to fuck a robot. I was just joining the long list of horny humans who have taken the technology of their day and used it for sex. Ancient erotic cave art, the Venus of Willendorf, the Venus de Milo, the Kama Sutra, and innumerable adult websites are just a few other examples of this.
Sex with my wife has always been splendid, and I have never wanted to be with anyone else. So, I had no problems with the quality of sex. What I wanted to improve was my quantity of sex. Once I learned the ins and outs of the old in-and-out, I wanted to spend a lot more time 'in' and a lot less time 'out'. I'm sure that women enjoy the 'in' part just as much as men do. My wife, for example, seems to have a rather large time in bed, especially when I stimulate her through a series of orgasms. I envy her this superpower. If I were multiorgasmic, I would only get out of bed on the weekends for peeing, pooping or pizza. I'd spend the rest of the weekend trying to get my bell rung over and over. But I digress.
After five years of marriage, my wife and I were well past our honeymoon levels of lust, and had sex only about once per week. One obvious way to get more sex was to just ask my wife, "Want to fool around tonight, sweetie?" She often said yes. However, she often said no, usually for very good reasons. So, how could I improve my ratio of sex achieved to sex requested?
I tried to approach this logically. If I asked her for sex every night, that would probably piss her off and I would get less sex. On the other hand, if I didn't ask her often enough, I would miss out on times when she was receptive, and I would get less sex. Bummer. Surely there was a sweet spot between these two extremes that would optimize the amount of sex, but how could I find it? Being a geek, I decided to try to find a tech solution to this problem.
But first, a few words about my wife. Sally is a lovely and luscious woman with sparkling green eyes. We met while waiting in a long Starbucks line, where a beginner barista had us moving along at a glacial gait. Sally noticed my impatience, and said, "Oh well, better latte than never."
Having been raised by a pair of pun-loving parents, my Pavlovian response was to fire back, "So what you're saying is, don't worry, be frappé?"
"Is that your best shot?"
"Sip happens."
"Well, latte da."
Many more bad coffee puns were flung back and forth before our orders were finally ready. She raised her cup at me, and said, "Well, it's bean great. Java nice day!"
As she turned away, I said, "Wait! Affogato get your phone number!"
She actually guffawed, and said "OK, well played! Unlock your text app." She tapped a few keys on my phone and then hers dinged. "Don't mocha me wait too long before you call."
That really perked me up. I called her that evening, and our dating commenced. After hanging out a few times with her tasty brain, I was smitten. One smite led to another, and we were an exclusive couple in a fairly short time.
She had recently graduated from medical school and was now in the midst of her internal medicine residency. I was getting a master's degree in electrical engineering and computer science. It wasn't long before we moved in together. We got married after we graduated, and she joined a multispecialty clinic here in Seattle. I worked for a few years for a large tech company, and then quit to pursue some personal AI projects.
My main project involved some promising new sensors for rapid, real-time characterization of volatile, airborne compounds. If that sounds a bit abstract, let's just say I have been building an electronic nose. You might well ask why we need an e-nose. After all, humans have trained dogs and other mammals for years to spot drugs, track fugitives and hidden land mines.
However, it can take a lot of time and money to train these animals, and their working lives are relatively short. The development of a reliable and inexpensive e-nose could be a game changer for these purposes. My plan was to perfect my e-nose, lock down the appropriate patents on its innards, and then license the crap out of it to other companies for big buckets of cash.
I had already thought of several commercial applications for a state-of-the-art e-nose. A grocery store could quickly scan the whole produce section and yank spoiled fruit and veggies. TSA and customs agents could quickly scan luggage and passengers for drugs, bombs and other contraband.
My device is not the world's first e-nose. However, mine is unique in several ways. It is hand-held and uses an array of solid-state sensors that can simultaneously sift the air for hundreds of thousands of compounds. But the secret sauce that makes my e-nose work so well is its AI back-end. The heart of this back-end is a convolutional neural net model that I trained for days on a large, multi-GPU workstation in my lab. I then deployed this trained model on a souped up Raspberry Pi hooked to my sensor array. The whole system can look at zillions of different combinations of odors, and recognize specific odor profiles, i.e. the unique smell that certain things have.
A sheet of music is a reasonable metaphor for a smell. If you don't have any musical training, you just see a bunch of black dots and lines on the page. If you take a class or two in reading music, you'll be able to pick out 'Mary had a little lamb' on a piano. An experienced studio musician can sight-read a page of music and play it at full tilt without practicing it. An orchestra conductor can look at a musical score and can hear in their mind every instrument in the orchestra simultaneously playing their parts in Beethoven's Third Symphony. It's all a matter of training. My e-nose has been trained to recognize the chords, scales, arpeggios and symphonies of smell.
I'm most interested in using my e-nose to analyze some of the many odors that are constantly boiling off of humans and other animals. Our puny human noses are able to tell us when it's time to take a shower, or which pair of underwear is clean and which one isn't. We can tell when the milk has turned or when one of the science experiments in the back of our refrigerator needs to die the true death. In contrast, my e-nose can correctly identify individual humans from their smell alone. It can distinguish the smell of a sheep from the smell of a wolf. It can sense the difference between a sleepy sheep, a frightened sheep and a horny sheep.
For the past year, I had worked with colleagues in our state's veterinary school to track the estrus (mating) cycles of various domesticated animals. Dogs, cats, sheep, goats, cattle and horses only mate a few times per year. If you make your living breeding and raising any of these critters, it is really helpful to know when they are in heat. You can then arrange a date with a mate and make yet more critters. After training it on the changing aromas of a variety of different mammals, my e-nose was able to accurately detect where they were in their estrus cycle.
In order to isolate the odors linked to the estrus cycle, I had to train my AI to also recognize and ignore other scents, such as when an animal has just belched, farted, peed or pooped. I also had to train it to recognize the compounds given off when an animal is anxious or frightened. You've probably heard of animals being able to smell fear. Well, now my e-nose can too.
Once my system was working well on animals, I began to wonder how well it would work on humans, who are a bit different from many other mammals. Female primates, including humans, have a menstrual cycle, rather than an estrus cycle. That means that their mating season runs essentially all year-long (Yay, primates!).
Could my e-nose tell me when my wife was most receptive to sexual overtures? It was an intriguing thought. To test this idea, I set up my e-nose at home, with sensor arrays in several rooms of the house, including our bedroom. At the time, it did not occur to me to tell my wife about this experiment. My wife, on the other hand, would never make this mistake. If she were to give one of her patients an experimental drug, she would first obtain informed consent from that patient. When this finally occurred to me, I cravenly decided to put off telling her about it until I had collected a bit more data. As it turns out, this was a fortunate decision.
Once the e-nose was installed, I began collecting smell data in our house and correlating with a diary I kept of our lives --- particularly our sex lives. It took a bit of work to sample and categorize most of the good and bad odors of our daily lives. My e-nose slowly learned to recognize most of the things we cooked and ate. I kept a log of the general moods that Sally and I were in (happy, pissed off, angry, calm, etc). You might also be wondering if I also kept a diary of each one of our various excretions and eructations, such as hiccups, farts, belches, sneezes, pees and poops. I did, but it only took a day before it drove me nuts. Instead, I set up a few surreptitious sensors on our toilets so that I could automatically log excretions without human intervention.
I finally came up a way to quietly sample my wife's vaginal secretions without tipping her off to what I was up to. I kept a washcloth by the bed to wipe off my face after a session of oral sex with Sally. After Sally went to sleep, I would quietly pass a sensor over that washcloth, as well as over the wet spot in our bed. I also trained my system to recognize the smell of my own semen.
After a few months of data collection, I used it to train my AI model on Sally's and my personal sexual bouquets. I then uploaded this new training to my e-nose, and tried to predict optimal times to hit on my sweetie.
The results exceeded my expectations. Advertisers have a term for the fraction of times an online ad view results in a mouse click. They call this the conversion rate, and are happy with a conversion rate of 2--3%. Using this definition, my sex conversion rate went up from 20% per ask to 50% per ask! My dick and I were thrilled!
An improved sex conversion rate was not the only benefit from my new, improved e-nose. The scent log from our bedroom mirrored many aspects of our sexual encounters, including our individual levels of arousal. I used this data to figure out what really turned Sally on, and thus up my carnal game.
