Entertaining AI: Chai Research is Making Conversational AI Fun

3341
woman on smart phone making conversational AI fun

The hype around AI language models has exploded since OpenAI released ChatGPT in November 2022. But two years ago, when I first started getting into the space it wasn’t like that. Everyone thought chatbots were boring, and that they weren’t very cool. That sparked an idea within me: this was a new technology, and it shouldn’t be boring. Why not build a chatbot that was entertaining? Making conversational AI Fun!

How it started

After graduating from the University of Cambridge, I developed an artificial intelligence (AI) to predict the horse racing markets. We used AI to gather all the data, then fed it back in to crunch the numbers and run the algorithms. The system then spit out a prediction on which horse was the most likely to win the race. We then used algorithmic execution to place bets and it proved successful in making money, which led to the conception of my first company Seamless Capital.

Based in Cambridge, Seamless Capital is an AI-driven hedge fund and algorithmic trading firm. I had a team of 10 really smart astrophysicists and quant researchers working alongside me to build the business. At the time of Seamless Capital’s apex, I reached the ceiling of strategy. You can only make so much money betting on horses. I could choose to take the low-risk choice and keep doing the same thing or I could aim for something bigger and better.

At the time GPT-2 had just been released, and I saw the excitement surrounding the technology. It sparked my imagination and thought this was where the future was going, so, I decided to found Chai Research with the goal of building our own language. This was making conversational AI fun in the making.

Removing the barriers and making chatbots cool

If you speak to ChatGPT, it’s very useful but it’s not very fun. In fact, when interacting with ChatGPT, if you ask it certain questions, it will respond with “I’m not allowed to discuss this,” and doesn’t want to engage. It’s not limitless. We want to make chatbots that are limitless — that can have any type of conversation, and in essence, remove the barriers and restrictions in the way you’re allowed to use this technology.

Users want to be empowered to build the bot that they want. They want to have their own interactions and their own connections. There’s so much exciting technology that’s on the horizon that will keep making these bots more and more powerful. Making conversational AI fun. There are so many people working in this space, but we’re the ones that want to push it forward to create something different, new and appealing: entertainment.

How its going

OpenAI has a huge lead in the language processing space. They published a lot of papers and shared a lot of code. In all of this they, in a way, gave away a lot of their secrets. We read through it, picked it all apart and then put it back together. Except, rather than optimizing for accuracy or for factual correctness, we optimized for something that we thought was entertaining.

We wanted our chatbots to be fun, not factual. As we built this AI, we released it to users and the results were like lightening. Users loved it. The chatbots had personality, they had energy. It’s as if they had emotions. It was totally unexpected.

With all of this positive feedback, we released our API model to users through our app, Chai. Within the app users are able to build their own bots using our AI and interact with them in a way that they wanted to, not in the way that open AI says is allowable under the terms and conditions of the API.

The Chatbot of the Future

For Chai Research, half of the vision and the excitement around what we are doing is taking this amazing technology and making it entertaining. The second half is seeing the users’ reactions. We want to keep making this AI more and more entertaining and we really think there’s blue sky in how far we can go.

This AI is on an exponential trend, but you can’t take it for granted that it’s just going to happen by itself. Right now, the AI has the intelligence of about an eight-year-old and it’s only just beginning, so how long until it has the reasoning abilities of an 18-year-old? Once it gets to that point, you’re not going to be able to tell the difference if you’re talking to a human or an AI. And that’s the future that we’re trying to build – making conversational AI fun!

Subscribe

* indicates required