Everyone in tech is talking about chatbots lately. So we decided we’ll finally start writing a bit about our own experiences with building mobile chatbots so far… and there’s still so much to do! Chatbot brain surgery never ends. 😜
While this post may be a bit outside the range of our normal, less tech oriented prose — considering we make selfie chat apps for iPhone and Android — as there is a fair amount of cutting-edge AI lurking beneath the surface of SelfieYo, we figured it might be fun to start talking a bit about it!
Our team has been experimenting with multiple types of AI in order to improve the experience of users. It’s been interesting, to say the least, as there is no perfect template for these things — despite what others may tell you…
We started reading about chatbots a year or so ago and wanted to figure out a way to use them for two primary purposes:
- To help our team scale in terms of creating a fun, energizing place for users that might not have people to talk to. You see, SelfieYo is primarily location-based photo chat, so if you are one of the first users in your area, it might get a little lonely… Sort of a classic ghost town problem that’s been around for ages. We think chatbots, when done smartly, could help lessen this challenge.
- We wanted to learn and explore the building and maintenance of chatbots so that we can eventually leverage the learnings to help brands and potential advertisers make wise decisions, implement effectively, and have a very clear idea of how these little autonomous critters work in the real world. To that end, obviously we needed to get ahead of the game and start practicing!
The technical challenges of building a chatbot on our system would be familiar to us as software product builders; assess the technology options, start exploring the details of how chatbots work and start coding our own implementations. 🤖
This is not trivial if you want to make a production quality chatbot that can scale across millions of sessions around the world, be easily modifiable, and have a variety of bells and whistles that would be required for customized mobile implementations.
The tech challenges were not surprising; what has proven more elusive is figuring out ways to program for situational and emotional intelligence that ultimately will separate “dumb” bots from the smart ones.
For example, Microsoft’s bot, Tay, got itself into hot water recently by making statements that clearly did not reflect the positions of Microsoft or the chatbot programmers, but were offensive reflections of some of the users who found ways to manipulate the responses of Tay.
It’s pretty easy to do — figure out how to encourage a bot to get salty. 😵
As many writers have pointed out, chatbot developers are not exactly generally situated in the same daily social circle as many of the users we are aiming to delight. The cultural milieu of chatbots and AI programmers is often quite different from that of end users. 🙀
We spend most of our time programming, tweaking, trying to improve… but so much of the next evolution will require a deeper connection with the emotional minds of human users, and we as builders will need to find ways to bridge the gap between “sophisticated bots” and truly emotionally intelligent bots that can get much better at discerning and reacting to the je ne sais quoi that makes the human mind, so far, infinitely more intelligent than any bot.
We’ll go into some more detail in future posts about what actually goes into creating chatbots, but for now we wanted to sort of set the stage for our own thought process of how the next wave of intelligence will need draw upon some of the “softer” qualities of the human brain.
Developers can continue to build hundreds of thousands or even millions of combinations in a bot’s brain — but we need to start trying to figure out the nuances of EQ to take the next leap.
In our next update on some developer thoughts around chatbots, we’ll dive a little deeper into how we’re trying to build the emotionally intelligent side of bot brains.
But for now, we’d love for you to snap a few selfies, chat with some friends on SelfieYo and see if you can help us understand EQ context a bit more. If you need a few tips on improving your selfie game, we’re happy to give a few suggestions. Perhaps this post on summertime selfie tips might be useful. 😂
You might end up influencing some synapses in a bot brain someday…