How We Taught Artificial Intelligence to Sellby@lensai
1,194 reads
1,194 reads

How We Taught Artificial Intelligence to Sell

by LensAIApril 23rd, 2020
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

Pavel Shkliaev, a co-founder of LensAI, created an AI-powered contextual computer vision ad platform that monetizes any visual content. The idea's global nature and potential caused him sleepless nights. The prototype was able to identify key objects and calculate the accuracy of the object detection algorithm. The one thing we didn't foresee is a crazy amount of manual work that we disliked the most but teaching machine associative thinking to machine thinking to the most. It took us three weeks to create a prototype, and that we got what we got all unbridged unbridging ways to develop the project.

Companies Mentioned

Mention Thumbnail
Mention Thumbnail
featured image - How We Taught Artificial Intelligence to Sell
LensAI HackerNoon profile picture


I am Pavel Shkliaev, a co-founder of LensAI, and this is my story.

After six months of work on the previous project, I finally got a free week or two. I was finally able to unload my brain, but my irresistible craving to create was too intense to ignore.

It's been a while since thoughts about the world's overproduction of goods and how it created the problem to sell rather than produce occupied my mind. I kept thinking of how Instagram turned into a vast marketplace, where every homemaker sells another luxury wrinkle cream. It doesn't take a brain surgeon to figure out that all these led to such an oversupply of information that an average consumer can't focus on one specific thing for a long time, forget about focusing on the text.

Even though I always enjoy the process of solving the problems, it felt like I had enough of it. Anyway, a solution to specific issues is a niche market, and I was ready to solve universal truths. It was time to shift my focus to creating new opportunities.

The Birth of An Idea

It all started with a sofa. Precisely, with me sitting on that sofa and clicking remote control. After changing over two hundred channels, surprisingly, I reached the Shopping Channel. First of all, I was shocked it existed. Secondly, I was more shocked to realize it existed because someone still buys such crap as super-duper knives for $50 , questionable diamonds, and magic steamers. Channel switching continued until the "Casino Royale" movie with Daniel Craig caught my attention. While watching, I could not stop wondering why would they not sell items that they show in the films? I liked the suit James Bond was wearing in that scene, and if given a chance right now, I would probably place an order for it. Although I did not need the suit, I liked how it suited the actor and just an idea of having one gave me a good feeling. As I proved to myself the known truth that the purchase of consumer goods is often a spontaneous event, I began to grow this idea in my mind. The first question arose: what if the concept of the shopping channel would be implemented right into the movie giving me, a potential consumer, an opportunity to buy things that were shown in that same movie? And then, the second question followed: what if I could instantly buy all the things that I see in the digital space without spending time searching for them or remembering the exact source where I saw them?

The idea tempted me. I did not think about AI and computer vision at that time, but some visuals began to emerge in my head. I kept imagining some object selection tool that would enable me to pick a product from the movie and buy it straight from that movie in real-time. The idea's global nature and potential caused me sleepless nights. I started to gather information and research other similar projects. It looked promising.

Once I finished the online research part, it became evident that similar ideas and projects exist in one way or another. However, all of them are niche projects that use computer vision to solve only one specific problem, such as video search for weapons, smartphone applications for searching for particular goods, etc. While researching, I crossed only one American project that was based on a similar idea and had a similar implementation, but it specifically concentrated on media advertising and stopped there. Undoubtedly, Internet giants such as Google, MS, and Amazon were ahead of all. I had to consider new inputs to make sure my idea could work, but everything else was pretty stable. During the development process, the main points of my idea had to shift more than once since I knew how essential it was to enter the market as quickly as possible.

"Hi, Konstantin! I have an idea!" I called the partner from the previous projects. Before making a call, knowing Konstantin and his tiredness from the cut-and-dried work, I knew for a fact that he would get involved as it was not another application for Apple but a damn artificial intelligence and machine learning! You know, some good food for an analytical mind. 

"Yes, you've got some interesting idea, but I have to check the technical feasibility before giving you an answer. I will need a week," Konstantin replied. That's when the technical feasibility study began.  We quickly concluded that the project was technically feasible, and we expected no unusual restrictions. The one thing we didn't foresee is a crazy amount of manual work that we disliked the most but teaching associative thinking to the machine was too tempting.

It took us three weeks to create a prototype, and that's what we got:

The prototype was able to identify both key objects and positions of them and calculate the percent accuracy of the object detection algorithm.

We moved to the brainstorming the ways to develop the project. Among all unbridled creativity and non-stop idea-generation, we highlighted four main components:

Development of an advertising platform for Publishers and Webmasters to join.

A computer vision technology that processes both images and videos found on the website and then, depending on the detected objects and other data obtained through the analysis, automatically embeds advertising straight into visual content.

A pay that comes from Advertisers for the traffic that such advertisements bring.

A commission that we take through our platform.

The output should be:

A user clicks on an advertisement and finds himself/herself on the website of the store where that particular product is available for sale. Yes, everything is that simple. The technical implementation was a fascinating process that generated lots of obstacles that we overcame during the development stage. We wrote another post regarding some of them but be aware of a lot of incomprehensible words we used to write it.

"Hello, Max! We are ending the work on the "Bread Card," I called my guy to bring some news that it was time to put our previous project on the shelf. "Oh, why? Damn, so much work has been done on it," Max sounded upset. "Yeah, no one pays for it... Anyway, I've got a better idea, and it is far away from being as banal as this "Bread Card," I cheered Max up.

I knew Max well, so his immediate desire to engage was not a surprise…

My team of engineers has been building software for the past ten months that hopefully, will bring a change to the digital advertising industry in the nearest future. I am looking for some feedback on using AI-powered ads that would turn the visual content of any website into a shoppable one.

Last summer, I got that crazy idea to create a tool that would allow users to shop items that they see in the video/image straight from that video/image. LensAI, that's how I named it, is B2B software that would analyze any visual content found on your website and then automatically deliver hyper-relevant ads of the products that were detected and available for sale on leading marketplaces like Amazon, Walmart, eBay, etc. Also, I choose the design of the ads to be as native as possible, and the ads would look more like helpful hints rather than irrelevant ads.

What do you think about the idea in general? Can you see yourself using it if it doesn't cost you anything but unlocks new income for you? LensAI has already joined the first round of various Affiliate Programs, and we would share generated revenue from those programs with our Publishing Partners/Websites, so it's a win-win.

We are making our baby steps to bring some innovative changes to the digital advertising industry. Here's what we got for you! We worked hard 🤓(from home, of course 😉) and finally finished Slide #2 on our website at Yay! 💥Make sure to check it out to see how LensAI Advertising Technology works on the famous "Diana" movie trailer. 🎥 Grab popcorn! 🍿


Subscribe on PH:

Stay tuned for more cool AI-Powered stuff to come!