In this interview, Eric James McDermott, Senior Specialist AI Innovation, and Francesko Molla, Data Scientist, provide deeper insights into the alpha phase of Dicobot's development.
How did the idea for Dicobot originate?
Eric: Our internal research and development continuously explores emerging trends and technologies to assess their applicability in various business use cases. When we witnessed the substantial leap in capabilities from GPT2 to GPT3 last year, it became evident that something significant was on the horizon. After consultations with clients and internal stakeholders, it became clear that a chatbot tailored to company data would be incredibly versatile and powerful. This marked the inception of Dicobot, and from that point onwards, we embarked on its development journey.
What unique skills does diconium contribute to the development of an AI chatbot?
Eric: diconium is uniquely positioned to pioneer innovative, data-driven products such as AI chatbots. Our diverse skill sets allow for streamlined end-to-end project execution, encompassing everything from strategy and concept formulation to prototyping, business design, customer journey mapping, data and cloud engineering, machine learning, AI expertise, full-stack software development, and UX/UI design. Additionally, our proficient data scientists enhance product refinement based on behavioral data. These combined capabilities make diconium the ideal partner for forward-looking projects like AI chatbots.
The developer team
Our team transformed a PoC into a cloud-accessible solution, requiring collaboration from backend and cloud engineering teams. Simultaneously, frontend developers worked on improving Dicobot's accessibility.
Which experts played a role in the alpha phase?
Francesko: The alpha phase involved transforming a Proof of Concept (PoC) developed by our research and development team into a cloud-accessible solution. This required collaboration from our backend and cloud engineering teams, while our frontend developers concurrently worked on enhancing Dicobot's accessibility.
What preparations are necessary before commencing AI chatbot development?
Eric: Effective resource management is pivotal for success in cross-team projects like Dicobot. Fortunately, we have developed advanced internal dashboards that provide visibility into the availability of all diconium experts, including developers and engineers. This allowed us to align the right talent with the right tasks, facilitating targeted progress. In addition to resource allocation, we needed to ensure that our initial models were scalable and capable of withstanding the test of time, at least in the short to medium term.
What was the initial step in the Dicobot project?
Francesko: The initial stages involved setting up the chatbot to run locally, establishing API connections with services like OpenAI, and exploring open-source models. We aimed to make the chatbot focus on the data we provided, and this marked the inception of Dicobot's cognitive abilities.
What were your personal experiences and key takeaways from the alpha phase?
Eric & Francesko: In the tech industry, we often become consumed by the pursuit of the latest technologies and their applications, leaving little time for reflection. Working with chatbots that craft responses based on provided documentation, mimicking human-like interactions, was truly awe-inspiring. It's a testament to the collective achievements of humanity, and Dicobot stands as a remarkable project in this regard.
What lies ahead: What are the upcoming steps in Dicobot's development?
Eric & Francesko: Even in its alpha phase, Dicobot has proven highly effective for various user requests. The next phases will primarily focus on enhancing the user interface and chatbot experience. We will continue to extensively test Dicobot and fine-tune its performance to provide an even better user experience.
