GPT & LLM. How many people know what these acronyms mean but don’t actually know what the letters stand for? These phrases essentially became a household name (primarily in tech hubs) overnight all thanks to OpenAI with the release of ChatGPT.
Let’s take a quick look at the surge in popularity of these terms. Google Trends confirms the recent surge in the two acronyms.
Starting around December 2022, there was a massive jump in searches for these types of terms. However, if you poll a room, most people don’t know what LLM (large language models) and GPT (generative pre-trained transformer) stand for, much less how they work.
To understand how these terms became omnipresent, let’s breakdown of timeline of recent events that underscored their development.
You might notice something: despite the recent popularity of these terms, OpenAI has been releasing various forms of GPT for a while. GPT-1 came out in June 2018 and GPT-2 came out in February 2019, both long before COVID had impacted the world. GPT-3 was released in May 2020 and, even more recently, GPT-3.5 was released in March 2022. And, while GPT-3.5 was released in 2022, you most likely didn’t hear about it until ChatGPT was released in late November 2022.
It was ChatGPT that led to an explosion in popularity. It took the world by storm, getting to 1 million users in 5 days of the launch. By January 2023, the product hit 100 million users. ChatGPT started this new craze and is the reason why searches for phrases like GPT and LLM started climbing in December.
OpenAI isn’t done—they release an even more powerful model, GPT-4, in March 2023.
Impact on the Startup World
With OpenAI’s revolution of Generative AI technologies, the startup world’s eyes were opened to amazing possibilities. Despite a bear market, venture capitalists started drooling over Generative AI companies. Most deals that got done in the last few months more or less have some relation to Generative AI. There are a couple of interesting data points about how Generative AI has permeated startup (and general tech) land:
- Funding in Generative AI has increased by 425% since 2020 to $2.1 billion (source).
- A significant percentage of Y Combinator’s W23 batch were AI focused (source).
- Notion introduced Notion AI in November 2022, beating most third-party GPT integrations (source).
These statistics underscore a bigger phenomenon: industries are getting rocked and are changing overnight because of how this new technology is changing previous assumptions industries are built on. One of the big questions is how Generative AI will impact analytics, specifically embedded analytics.
Generative AI and Embedded Analytics
There are two user groups in Explo’s ecosystem:
- Explo’s customers pay for the Explo software, building dashboards in the Explo web app, and embedding the dashboard into their own applications or websites.
- Explo’s end users are customers’ customers, the users that consume the embedded dashboards.
There are two main ways to segment how Generative AI can impact embedded analytics. The first is the impact on customers, and the second is the impact on end users.
Generative AI has a huge potential to help customers with the dashboard creation process. There are 3 powerful process changes that can be disrupted:
- SQL writing
- Dashboard creation
- Custom styling
Today, customers need to write their own SQL to create datasets that feed visualizations in the platform. ChatGPT has proven to be very good at writing SQL. Explo’s query editing interface can utilize GPT to write queries, edit queries, and even interpret previously written queries.
Behind the scenes, dashboard configs are all just a bunch of JSON blobs. Models can be taught to understand how the JSON blobs are structured and correspondingly create these blobs on their own, effectively autonomously creating dashboards. In short, GPT can be leveraged to build powerful dashboards and give customers the end solution or a starting point to work with.
Dashboards need to be styled to look like the application they are being embedded into. Customers have a UI interface in Explo they can utilize to match the style of dashboards to their application, but this is something that GPT can also tackle just as well as humans can, if not better. The styles config is just a JSON blob and these models can be taught to analyze an existing design system, and tweak the JSON config till it precisely fits the given designs.
End User Impact
This is where the real magic happens. When Generative AI can properly be used to improve and enhance the end-user experience, embedded analytics will start to instrumentally change. There are many hypotheses around how this can be done in the industry, but no one has quite cracked the code just yet. This is what we know about end-user behavior:
- End users want to have the flexibility to customize the analytics they see.
- Sometimes they know how they want to change the dashboard and sometimes they don’t know how they want to change it, but they know they don’t like its current state.
- They know how they want to change the dashboard, but don’t understand the underlying data enough to make the change.
- They don’t know what they want changed, but they know the question they want answered.
Embedded analytics can leverage Generative AI to improve and change the end user experience. It’s not just about answering the data question (although extremely important), but also how to best answer it (bar chart, line chart, or table?), and how to most effectively position it on the dashboard itself.
At Explo, we are working on a few really exciting ways that will completely change the end user experience. The feedback we get directly from end users, our abundant amount of usage analytics, and our clear understanding of the space has helped direct how we’ve gone about integrating Generative AI into the end user experience.
🚀 Book a meeting with Explo to learn more - let's talk