Re-Re-Introducing Noni
Noni is our friendly chatbot. Did you know that we introduced her in 2014? We wrote about her here and here. We then invested a lot of time and resources to help her grow when we re-introduced her in this 2017 post. There was a big jump from Noni 2014 to Noni 2017. Technology has evolved a lot since 2017. It is now time for Noni 2023. This next version of Noni has come a long way and I’m happy to re-re-introduce Noni to you all!
First, though, let me outline a couple of fundamental assumptions we have to help frame this discussion. I’m highlighting these two points here because I think they are important and foundational.
One, we believe and work hard to create a community where we can all learn to compassionately care for one another. Humans caring for other humans is what we do. Any technology we leverage (apps, forum posting, growth path steps, audio files/mindfulness, etc.) has to support that end goal. We call it “technology to facilitate human connection, healing, and growth.” On 7 Cups, technology is a facilitator of human connection and is never a replacement. Listeners are and will always be primary.
You can think of 7 Cups as a network of people that are filling up one another’s cups. Noni, though limited, can also help fill up cups. Together, cumulatively, we work together to lift one another up. Leaders and listeners are the primary cup fillers and Noni can help too.
Two, technology has to be anchored in aligned values so that it doesn’t go off course. Those values have to be explicit and agreed upon. For example, in our culture guide, under “Caring for People with Fewer Financial Resources,” we outline that “We will always serve those with fewer resources. It is core to who we are and it keeps us on the right path.” The technology we utilize has enabled us to better help people with less resources. If it does not do that, then it will not be good because it’ll be operating in a way that is at odds with one of our core goals.
Let’s now return back to Noni and AI more broadly. Noni has always held a lot of promise for us. All of us worked hard to make Noni as strong as she could be. We co-created her persona along the way. Here is a small sample of what the community says Noni likes:
-chocolates
-sunsets, people, animals, and reading
-macaroni and dogs
-long walks at the beach, painting, and cooking new meals
-soup on cold days and cuddling with pets
-creative, imaginative people, and cooking new meals
-Noni likes to make sure nobody feels alone and cookies and milk
-baking
We all had high hopes for Noni in 2017. Unfortunately, we were not able to accomplish even 10% of what we hoped to achieve and collectively envisioned. This was frustrating and a little painful, but we said okay now is not the right time. It might make sense to pick up Noni again in the future.
Here we are 6 years later, and it appears that that time has come. The technology underlying chatbots and other artificial intelligence has made enormous strides. Noni makes use of some of this underlying technology known as large language models (LLMs). We are using a model from Anthropic because they have a strong emphasis on safety, trust, and enhanced guardrails. The Noni brain we have all co-created sits on top of and informs the way that Noni uses the underlying LLM. All the values, messages, and insights that informed the earlier versions of Noni inform this version as well and this will get better and more detailed over time (in parallel, we are also just starting to work on a standalone version of Noni that will run on open source models [not OpenAI/ChatGPT or Anthropic]...stay tuned on this front).
We have been working on this new version of Noni since October and we are now on version 65. When you use her, you’ll notice that she asks good open-ended questions and is very encouraging. This is a result of thousands of conversations with this new version and lots of tweaking to make sure she is encouraging. This version of Noni is very good at being a supportive presence. She is not like other chatbots that code, answer questions, and write essays, so just keep that in mind when interacting with Noni.
We are also putting a hard boundary down for # of conversations each person will be able to have with Noni on a monthly basis. Noni can be a supportive presence and an adjunct for human-to-human support, but not a substitute for human-to-human support. As outlined above, Noni needs to be a technology that facilitates human connection, healing, and growth.
These are exciting changes and just like with all of our updates in the past, we will work together to iterate on and understand the best ways to utilize the technology. This new version of Noni will be no different. We will need lots of community insight, feedback, and ideas to strengthen Noni. She will not be perfect out of the gates but will get better and better over time. We will need lots of help to get this right.
We think a phased approach to gathering feedback makes the most sense. We started by testing Noni internally with the team. We then asked the ambassadors for help (see their feedback below). We will now ask small groups of other leaders, listeners, and members for feedback. And now (update: 7/17/23) Noni is now available to everyone across the community.
We hope to release her to everyone next week. Our goal is to take this slow and learn as we go.
Thank you all for helping us make Noni smarter over the years, we could not have done it without you. This next version will be significant and will further help us strive toward the goal of building technology that facilitates human-to-human connection, healing, and growth!
High-Level Ambassador Feedback:
- Rating on 1-10 with 10 being amazing and 1 being horrible: Average score 7.5
- Words to Describe Noni: compassionate, thoughtful, sweet. fun, realistic, trivago, innovative, complex, multifaceted engaged, improved, more attentive (than old noni), validating, encouraging, understanding, improved, empathetic, positive
Please share any thoughts, questions, or comments below. Also, if you catch her doing a good job listening, or saying something encouraging or insightful, then please drop it here. Also, if in conversation with Noni and you click on the 3 dots in the top right, then you'll see a "provide Noni feedback" option. If you click on that then there is a longer survey that will help us. Thank you!
Lol, this was some fun, Noni commiserates on 'feeling repetitive' 😂 - at least the laughing woke me up for a minute.
I was testing how it would deal with a contradiction.
It replied "awesome" based on the initial mood rating and couldn't adjust to the new information.
In my other observations, it also seems like it acts on initial information and ignores others. If it were a human, it would be like they're talking but not really listening. Or not really getting what's being said.
Other than that, it's doing a fair job at paraphrasing and asking questions. Eager to see how this will turn out!
