Earlier this year, Microsoft added a new key to Windows keyboards for the first time since 1994. Before the news dropped, your mind might’ve raced with the possibilities and potential usefulness of a new addition. However, the button ended up being a Copilot launcher button that doesn’t even work in an innovative way.
Logitech announced a new mouse last week. I was disappointed to learn that the most distinct feature of the Logitech Signature AI Edition M750 is a button located south of the scroll wheel. This button is preprogrammed to launch the ChatGPT prompt builder, which Logitech recently added to its peripherals configuration app Options+.
Similarly to Logitech, Nothing is trying to give its customers access to ChatGPT quickly. In this case, access occurs by pinching the device. This month, Nothing announced that it “integrated Nothing earbuds and Nothing OS with ChatGPT to offer users instant access to knowledge directly from the devices they use most, earbuds and smartphones.”
In the gaming world, for example, MSI announced this year a monitor with a built-in NPU and the ability to quickly show League of Legends players when an enemy from outside of their field of view is arriving.
Another example is AI Shark’s vague claims. This year, it announced technology that brands could license in order to make an “AI keyboard,” “AI mouse,” “AI game controller” or “AI headphones.” The products claim to use some unspecified AI tech to learn gaming patterns and adjust accordingly.
Despite my pessimism about the droves of AI marketing hype, if not AI washing, likely to barrage the next couple of years of tech announcements, I have hope that consumer interest and common sense will yield skepticism that stops some of the worst so-called AI gadgets from getting popular or misleading people.
Can’t wait until the AI enhanced sex toys.
considers
That’d actually probably be a not-unreasonable application for machine learning, if you could figure out some kind of way to measure short-term biological arousal to use as an input. I don’t know if blood pressure or pulse is fast enough. Breathing? Pupil dilation?
Like, you’ve got inputs and outputs that you don’t know the relationship between. You have a limited number of them, so the scale of learning is doable. Weighting of any input in determining an output probably varies somewhat from person to person. It’s probably hard to get weighting data in person. Those are in line with what one would want to try doing machine learning on.
IIRC, vibrators tend to have peak effect somewhere around 200 Hz, but I’d very much be willing to believe that that varies from person to person and situation to situation. If one has an electric motor driving an eccentric cam to produce vibration, as game controllers do for rumble effects, then as long as the motor’s controller supports it, you could probably train that pretty precisely, maybe use some other inputs like length of time running.
I don’t know if it’s possible to have a cam with variable eccentricity – sort of a sliding weight that moves towards or away the outer edge of the cam – but if so, one could decouple vibration frequency and magnitude.
googles
Looks like it exists.
https://www.dmg-lib.org/dmglib/main/imagesViewer_content.jsp?id=16182023&skipSearchBar=1
So that’s an output that’d work with a variety of sex toys.
There’s an open-source layer at buttplug.io – not, despite the name, focusing specifically on butt plugs – that abstracts device control across a collection of sex toys, so learning software doesn’t need to be specific to a given toy, can just treat the specific toy involved as another input.
I’m sure that there’s a variety of auditory and visual stimuli that has different effect from person to person and isn’t generally-optimal today.
And, well, sex sells. So if one can produce something effective, monetizing it probably isn’t incredibly hard, if that’s what one would want to do.
EDIT: Actually, that variable-eccentricity cam is designed to be human- rather than machine-adjusted. That might not be the best design if the aim is to have machine control.
Looking through the hardware compatibility list on buttplug.io, one such device is the “Edge-o-Matic 3000”. This claims to keep a user near orgasm without actually having an orgasm. For that to work, there have to be sensors, and fairly reactive to arousal in the short term. It looks like they’re using a pneumatic pressure sensor driven off a bulb in a user’s butt to measure muscle contractions, and are trying to link that to arousal.
https://maustec.io/collections/sex-tech/products/eom3k?variant=40191648432306
If they’re trying to have software learn to recognize a relationship between muscle contractions and arousal sufficient to produce orgasm, if it’s automatic rather than having someone tweaking variables, that’s machine learning. Maybe “AI” is a bit pretentious, but it’d be a sex toy doing machine learning today.
That’s an interesting idea, but:
I’m dubious that it actually works well. It’s described as being a work in progress.
Even if it works and solves the problem they’re trying to solve (being able to reliably predict orgasm), I’m not sure that muscle contractions can be used to predict arousal more-broadly.
My guess is that as sensors go, mandating that someone have an inflatable bulb up their butt to let the sensor get readings is kind of constraining; not everyone is going to want that at all, much less when they’re, well, playing with sex toys. Their butt might be otherwise-occupied.
That being said, it’s gotta at least be viable enough for someone to have been willing to put work into and commercialize a device based on that input. I’d believe that muscle contractions are an input that one could reasonably derive data from that one could train a machine on.
Maybe one could use brainwaves as an input. That’d avoid physical delay. I’ve got no idea how or if that links to arousal, but I’ve seen inexpensive, noninvasive sensors before that log it. Using biofeedback off those was trendy in the 1970s or something, had people putting out products.
https://en.wikipedia.org/wiki/Electroencephalography
At least according to this paper, sexual arousal does produce a unique signature:
https://link.springer.com/article/10.1007/s10508-019-01547-3
If it’s primitive enough, probably similar across people, easier to train a meter to measure arousal from EEG data on one set of people that can be used on others.
https://www.sciencedirect.com/science/article/abs/pii/S009130571400032X
That sounds promising.
There’s an open EEG product at two channels without headband for 99 EUR.
https://www.olimex.com/Products/EEG/OpenEEG/EEG-SMT/open-source-hardware
Some more-end-user-oriented headsets exist.
https://imotions.com/blog/learning/product-guides/eeg-headset-prices/
Hmm. Though psychologists have to have wanted to measure sexual arousal for research. You’d think that if EEGs were the best route, they’d have done that, else physical changes.
https://www.sciencedirect.com/science/article/abs/pii/S2050052115301414
Hmm. That’s measuring physical changes, not the brain.
https://en.wikipedia.org/wiki/Vaginal_photoplethysmograph
That doesn’t sound like, even concerns about responsiveness in time aside, existing methods for measuring arousal from physical changes in the body are all that great.
As in, maybe measuring the brain is gonna be a better route, if it’s practical.
I couldn’t get buttplug.io to work 😒
Lol why wait, Lovense is already touting AI driven patterns.
What are they using as input? Like, you can have software that can control a set of outputs learn what output combinations are good at producing an input.
But you gotta have an input, and looking at their products, I don’t see sensors.
I guess they have smartphone integration, and that’s got sensors, so if they can figure out a way to get useful data on what’s arousing somehow from that, that’d work.
googles
https://techcrunch.com/2023/07/05/lovense-chatgpt-pleasure-companion/?guccounter=1
Hmm.
Okay, so the erotica text generation stuff is legitimately machine learning, but that’s not directly linked to their stuff.
Ditto for LLM-based speech synth, if that’s what they’re doing to generate the voice.
It looks like they’ve got some sort of text classifier to estimate the intensity, how erotic a given passage in the text is, then they just scale up the intensity of the device their software is controlling based on it.
The bit about trying to quantify emotional content of text isn’t new – sentiment analysis is a thing – but I assume that they’re using some existing system to do that, that they aren’t able themselves to train the system further based on how people react to their specific system.
I’m guessing that this is gluing together existing systems that have used machine learning, rather than themselves doing learning. Like, they aren’t learning what the relationship is between the settings on their device in a given situation and human arousal. They’re assuming a simple “people want higher device intensity at more intense portions of the text” relationship, and then using existing systems that were trained as an input.
Lovense is basically just making a line go up and down to raise and lower vibration intensities with AI. They have tons of user generated patterns and probably have some tracking of what people are using through other parts of their app. It’s really not that complicated of an application.
Wait, so we should poison AI data to force AI sex toys to only allow edging?
Filtering out bad data from their training corpus is kinda part of the work that any company doing machine learning is gonna have to deal with, sex toy or no.