Machine Learning Bullshit Detector
My invention of the day: a Machine Learning Bullshit Detector. (MLBsD)
Or maybe it will be a meter: Machine Learning Bullshit Meter (MLBsM). When I explained the concept to my friend Q (aka “the Grumpy Luddite) he thought it should have a big analog gauge, like a pressure gauge on a steam locomotive and it could rate the bullshit from 0 (green) on the left for 0% bullshit to 100% (red) on the right. It could also click like a geiger counter or maybe it could hum like a theremin: low pleasing pitches for not bullshit and increasingly high-pitched and irritating for bullshit. Even better it could purr like a cat for not bullshit and yowl like a cat fight for bullshit.
These details of user experience can be finalized later with the assistance of a user experience researcher. The more important question is: how will it work?
Don’t ask me for the minute details, I’m more of a big picture visionary and in fact I’m not involved with the tech industry at all. Last time I wrote any code was in highschool in 1993 when I made programs to do my math homework for me. Unfortunately math class advanced quicker than I could make programs to do my homework and since I spent too much time trying to make a computer do my math homework and not enough time studying math I wasn’t a great student. A year as an exchange student in Sweden distracted me from my fascination with programing. A subsequent B.A. in cultural anthropology and a career as a documentary filmmaker was, in retrospect, maybe not the best financial decision.
The answer to the original question is: machine learning neural network. I’d never heard of a neural network until last summer when I was sitting on my sailboat in Alaska, drinking coffee with my girlfriend, and meditating on the greasy grimy guts of the boats transmission, pieces which where spread around the cabin. For some reason it had suddenly, in mid cruse, decided to work only in reverse — which made getting home on a windless day interesting. My girlfriend is Russian and has a degree in math and another in cognitive science. I was hoping with this background she would be able to figure out what was wrong with the transmission but she’s not a morning person and apparently small Japanese marine transmissions wasn’t a strength at the faculty of Mathematics and Mechanics at Saint Petersburg State University. Instead, I’m not sure exactly why, she decided to explain neural networks to me and I, being just a little A.D.D., forgot entirely about the transmission. To be honest, although she made charts and diagrams on a piece of grease splattered cardboard for me, I still don’t really understand the internal mechanism of a neural network. But I knew, then and there, that I wanted one.
Apparently everybody had the exact same epiphany I did because the same day there where articles about neural networks in my Facebook feed. I quickly forwarded these to Tanya. It occurred to me that with her background she could instantly get a very well paid job teaching machines to take over the world for some evil tech billionaire and in turn could subsidize my sail boat collecting and documentary making habit and compensate for my having not becoming a programmer. I still fail to comprehend why but she showed absolutely zero interest in this opportunity, preferring to comment on cat pictures on obscure Russian social media networks, go mushroom hunting, and complain that we couldn’t go sailing because the transmission was still broken.
I knew I really wanted a neural network but what would it do? Then one day, while riding bicycles with The Grumpy Luddite and listening to him complain about the silly and ill conceived aspects of his job making robotic lettuces, it struck me: A machine learning bullshit detector! I could train it by feeding it big data sets of bullshit and not bullshit and once it could reliably discern one from the other I could let it loose on the world.
What would be a practical use? Well, suppose you are interviewing somebody for a job. You but it on the table, switch it on, and start asking questions. If it purrs like a cat everything is good. If it yowls like a cat fight… Venture capitalists could use it to access pitches. Journalists could use it to to access the words of political candidates. It could be a web browser plugin that color codes the internet for you, highlighting bullshit in brown. The possibilities are virtually limitless.
My girlfriend raised a good question: who identifies what is and what isn’t bullshit? I initially thought she might be a good bullshit detector trainer, after all she is not shy (nor particularly tactful) in her identification of bullshit. But the more I thought about it the more I realized that she might be a little too good at identifying bullshit. If you train your detector that everything is bullshit it’s not very useful is it? Plus the constant sound of cats yowling could get on the nerves. The Grumpy Luddite is one of the smartest people I know. He successfully made the engine on my other sail boat work after I spent months fruitlessly banging on it. But when it comes to training bullshit detectors I fear he suffers from the same shortcoming as my girlfriend. For example, of all the articles I’ve ever forwarded to him the only one he approved of was this very in depth piece which basically argues that the 18th century Chinese wheelbarrow was the absolute apex of human technical development and it’s been all down hill from there. So I think that basically leaves me as the only reasonable bullshit detector trainer.
Alternately, the customer could train their own bullshit detector to their own taste. It could proceed them in the world and drown out any bullshit they might otherwise encounter with the sound of cat yowling. I mean this quite literally: for the more affluent customer it could be an autonomous robot that anticipates their every movement and stays three steps ahead, always on the alert. For the more budget conscious customer it could ride on the end of a selfie stick, like a blind mans cane, to help them navigate around bullshit.
Yet another option, for those too lazy, or too busy, or simply not smart enough to train their own bullshit detector is to buy one pre-trained by a trusted (perhaps celebrity) trainer. Thus The Machine Learning Bullshit Detector will be not only a device but also an online marketplace, similar to the app store, where clients can buy or sell a range of bullshit detector trainings.
If you are a venture capitalist who would like to invest in The Machine Learning Bullshit Detector please contact me.
If you aren’t a venture capitalist but would like to be one of the first to get your hands on a Machine Learning Bullshit Detector please keep your eyes open for our upcoming Indigogo campaign.
I have created a publication on Medium called The Machine Learning Bullshit Detector and this is the first post in it. If you have an idea, and you aren’t sure if it’s bullshit or not, write a short article and submit it to The Machine Learning Bullshit Detector for review. Whether we think it’s bullshit or not, if it’s entertaining we would be happy to publish it and readers can give their opinion in the comments section.
Currently I’m two and a half days into an 8 day train trip across Siberia. I’m on my way to visit Sergey Zimov, a crazy Russian Geophysicist living in a semi abandoned post, post-Soviet, post-apocalyptic ghost town in the most remote corner of North East Siberia who has an improbable plan to save the world from a catastrophic global warming feedback loop caused by melting permafrost by recreating the now vanished, ice age, “Mammoth Steppe” ecosystem, complete with millions of large herbivores roaming the artic like a wooly Siberian Serengeti — so forgive me if I’m a little slow in responding to submissions. You can read more about Zimov and Pleistocine Park here if you’re interested.
