How Philly’s biggest startup expo helped these founders train their own ‘artificial intelligence’

As a veteran entrepreneur who has invested in a couple successful Philly startups, I was pleased when I learned that we were able to score a last minute high top for my company, uTu.ai, at last week’s Philly Startup Leaders Entrepreneur Expo during Philly Tech Week 2017 presented by Comcast. Tech Week can be super helpful for early-stage companies new to the scene — and that was definitely the case for us. It was a great event that allowed us to give folks an inside look into what we’re building. As I reflected on our experience the next morning, it occurred to me that Wednesday night unfolded in exactly the same way we use artificial intelligence to teach a chatbot to interact with a human.

Chatbots, Facebook Messenger extensions and Alexa Skills are all names for bots on different platforms and they all use artificial intelligence to learn how to converse with their users. The ultimate goal is for the bot to pass the Turing Test. Doing so indicates that a human can’t determine whether they’re talking to the AI or another human. In practice, this takes time and inputs to “train” the AI to respond properly in natural language. And as I was sipping my coffee the morning after the Expo, it occurred to me how analogous the experience at the Entrepreneur Expo was to training AI. Here’s how.

We started uTu six months ago, raised a seed round and we have just gone to market in March. Our early sales calls have been structured. We knew who we were targeting and specifically how uTu could help their business. However, this wasn’t the case at the Entrepreneur Expo as we never knew who was coming up to our table or what they knew about bots/AI. Among our visitors were hiring candidates, SVP’s of product, even Mayor Kenney. Much like a bot with AI, we knew nothing about our audience at the start of our conversation and so my cofounder, Marcus Tewksbury, and I simply started with our pitch.

In the beginning of the evening, we took our time with a detailed description of the business.  As the night progressed, more variables were introduced such as, more people and louder music. My partner and I found that we could not be as descriptive due to the time and the crowd. We also quickly established what features and concepts of the business people responded to most. So, we then used this information to gain a connection and understanding with our immediate audience quicker and earlier in the conversation.

Midway through the evening when the venue was very crowded, I had learned some useful questions to ask a visitor first and then start our pitch, which by that point had been condensed to about half the time. Trained by both my brief interactions and external variables (in my case, crowd size and noise level, but for a bot it could be things like time of day), I learned to respond in a more efficient and effective manner.

This process works the same as how we train a bot’s AI. The combination of inputs, processing, and incorporating learned data results in a better experience for the user. The next morning my cofounder and I participated in Technical.ly Introduced, a rapid meet-and-greet event where we had ten minutes to pitch regional companies. Our “training” from the previous night paid off. We had condensed our story to a couple minutes, which left ample time for questions. Our company then received several requests for follow-up meetings from several of the companies. It was a very valuable 48 hours and we were very grateful to the Technical.ly team for their support and hard work.