Understanding Generative AI

Generative AI is a form of AI that creates new content based on a prompt. Chat GPT is a form of generative AI called a large language model (LLM) that produces text-based output. While it appears to be magic, the algorithm has been trained to predict the text response that makes the most sense based on the massive amounts of content that it has been trained on.

But there are other forms of generative AI that can produce different kinds of outputs, from artwork, to music, to analytic insight. AI is perfect for combing through huge amounts of data, extracting meaning, and generating a new way to consume that meaning. Virtualitics uses generative AI to explore and explain data so that data analysts and data scientists can quickly grasp what’s going on inside data sets.

Using Generative AI to Power Intelligent Exploration

Intelligent Exploration is the use of AI coupled with multidimensional visualizations to do rich data exploration of vast, wide datasets. Using forms of generative AI, amongst other advanced analytic tools, analysts can enter prompts to produce meaningful insight. Virtualitics AI Platform incorporates powerful generative AI capabilities, including:

Natural Language Query
Natural language query (NLQ) broadens the audience who can take advantage of Virtualitics’ already easy-to-use AI tools. Less technical users and new users alike can perform powerful analysis by using simple commands like “What drives ‘Sales’?” or by taking advantage of our Suggested Queries. Additional syntax and industry-specific language can be customized through use of Keywords.

Smart Mapping
Smart Mapping is a supervised machine learning routine that allows analysts to discover which attributes are driving a selected target, and to what degree. The mapping results lists all of the attributes ranked by importance. There may be a number of different ways to review the results so Smart Mapping suggests several visualizations that are best suited for the data types. For example, if Longitude and Latitude are drivers, the routine suggests a geo-spatial view. The suggested mappings consider the relative importance of the attributes and any correlations existing between them.

Network Extractor
Virtualitics’ patented Network Extractor builds networks and creates communities based on a selected target. For example, to create customer personas, the Network Extractor will sort through hundreds of attributes to find the ones that best defines customers, and then generate and visualize a network graph of customers based on how similar or dissimilar customers are from one another. Additionally, it will detect groupings of customers that can be defined as communities. Network graphs are an incredibly powerful way to visualize and get a rich understanding of the data.

Insights
Insights are plain text observations about statistically significant regions of data within the analysis that the user is currently working with. Insights can call attention to regions of obscured data points while working with big datasets, ensuring that every user, regardless of data skills, are getting at the meaningful information within their analysis. Insights can highlight an area of the plot where specific values for key metrics lead to higher shipping delays or identify important combinations of characteristics that lead to higher spending, or any other finding of note. Insights can be saved for later review or to share with others.

Explainable AI
Explainable AI (XAI) provides succinct descriptions of relationships and insights within a network. It analyzes relationships that define each community and provides succinct descriptions of these insights to users. XAI can be run on any type of network visualization within the platform, regardless of whether the Network Extractor was used to construct the network or not.

Advancing Virtualitics’ Intelligent Exploration

Applications like Chat GPT deliver answers to conversational questions in text. Intelligent exploration, on the other hand, can make sense of text, numerical information, proprietary data, and lots of other multidimensional data visually. Pairing these two AI technologies together will result in major advances in analytics. Data analysts will be able to seek and deliver answers from huge amounts of data using everyday language. The outputs—visualizations, reports, recommendations—will become more understandable to decision-makers who don’t have a background in data science or data analytics. And this will make taking action on findings easier and faster.

We will be incorporating an LLM on our product roadmap for the near future, but like all of our features, we’re focused on adding functionality that will enhance the usefulness of our solution, helping users discover and understand more from their data.

We’re also focused on addressing three significant concerns when it comes to LLMs:

• Prompt Design
  For an LLM to return a useful output, it needs to have interpreted the user’s prompt the way it was intended. There is a lot of nuance in language that can lead to misunderstandings and we’re focused on a solution that has guardrails to ensure consistent results that meet expectations.
• Hallucinations
  LLMs, including ChatGPT have been known to simply make up data to fill in the gaps in their knowledge just so that they can answer the prompt. They are designed to produce answers that feel right, even if they aren’t. Hallucinations in analytics will undermine the trustworthiness of the results so our solution must control for this.
• Security
  Keeping our customer data secure is our top priority. We currently adhere to a number of security standards and we will ensure that the LLM solution that we move forward with will keep us in compliance. That means that ChatGPT and other external models like it are not viable options. There are a growing number of LLMs on the market that are suitable for commercial use and we’re working on the options that will best suit the needs of our users.

Telling the Story of Your Business

One reason AI is such a game changer for data analytics is the ability to look at the full universe of information that’s available. It’s like being able to look at the full vault of a starry sky, getting guided to the patch that’s most important, then zooming in to find new discoveries.

So many organizations are sitting on a wealth of data they can’t make sense of. They may have a shortage of people qualified to sort through it. Or their leadership finds it hard to trust AI findings because conclusions aren’t backed up by explainable AI. They might lack an AI engine powerful enough to ingest all the relevant types of data, from sales reports to PowerPoints to webinar transcripts to spreadsheets.

Using generative AI with intelligent exploration will help companies move beyond these challenges. These AI technologies use natural language and visuals to tell the full story hiding in data. They are poised to automate much of the work of data analysis, so analysts can do more.