In Part I of the blog series “It’s No Big Deal, but ChatGPT Changes Everything”, we were introduced into the world of ChatGPT, chatbots, and generative Artificial Intelligence (AI). We ended Part I by giving ChatGPT a test run, by asking it “What would be a great vacation place for my family?” that gives us a basis for exploring further ChatGPT applications and potential areas of evolution.

Improving ChatGPT’s Vacation Planning Request Experience

The “Planning a Vacation” experiment in Part 1 was impressive and sets the stage for exploring further advancements for ChatGPT and other AI-infused, intelligent agents such as:

• What if ChatGPT could analyze your historical data (e.g., previous vacations taken and satisfaction levels, vacation articles read, vacation discussions on social media) to determine your personal vacation propensities and preferences?
• What if ChatGPT could take your personal vacation propensities and preferences and map that to what others with similar propensities and preferences preferred for vacation experiences?
• What if ChatGPT could ask diagnostic, exploratory, and validation questions to refine its understanding of your vacation intentions, objectives, ideal outcomes, and the metrics against which you would measure vacation success?

The types of questions that ChatGPT could employ to clarify your vacation intentions, objectives and ideal outcomes would include:

• Diagnostic Questions to clearly understand the objectives and intentions your vacation request, why you are taking a vacation, assess your current situation, clarify your desired vacation outcomes, and understand how you will measure the effectiveness of your vacation.
• Exploration Questions to explore similar or related alternatives vacation ideas, test different vacation rationale to spur ideation, and/or rank order your vacation rationale and success criteria in order to refine the vacation recommendations.
• Validation Questions to confirm and paraphrase what ChatGPT thought you requested and validate the intentions and objectives of your vacation request.

Going back to the original ChatGPT vacation planning request – “What would be a great vacation place for my family?” – might lead ChatGPT to ask the following diagnostic, exploration, and validation questions as it seeks to find the ideal options to my vacation request:

• How do you define and measure “great” from a vacation perspective?
• What vacations did your friends take that you found interesting?  And why did you find those vacations interesting?
• Please rank order these vacation destinations: skiing, hiking, beaches, cruises, casinos, and site seeing?
• How big is your family, what are their ages, and what are their areas of interest?
• What are your budgetary constraints? Is there a maximum that you’d be willing to spend for a “great” vacation?
• What are your travel limitations? What about travel do you find annoying and challenging?
• What was your favorite vacation and why was that vacation so special?
• What was your worst vacation and what made it your worst?

These are the types of questions that a good travel agent would ask, and evolving ChatGPT to enable this diagnostic, exploration, and validation questions would dramatically improve the relevancy of its answers…to you!

Creating “Your Own Digital Assistant” – YODA

ChatGPT demonstrates the potential of an AI-infused application or agent to 1) engage in a diagnostic and exploratory conversation with the user to 2) ascertain and validate user intent, objectives, and desired outcomes, and 3) identify how the user will measure the success of those desired outcomes.

Generative AI holds the potential to power Your Own Digital Assistant – or YODA – to help guide your analysis and decision-making. Yes, Generative AI holds the potential to create a YODA (app) that sits on everyone’s (virtual) shoulder helping them to assess their current situation, analyze their history of engagements and interactions, tap into a world-wide knowledge repository of experience and expertise, and recommend actions or options to make more informed, more relevant decisions (Figure 1).

Figure 1: Wise and Experienced YODA Sitting On Your Shoulder

For example, let’s say that you are a technician for a manufacturing facility, and you notice a problem with one of the plant’s motors. You need to engage YODA…

“Hey Yoda, my service motor increases its vibrations and electric current but there is no corresponding electric voltage increase.  What might be the problem?”

YODA: “There could be several potential causes for a service motor to experience increased vibrations and electric current without a corresponding increase in electric voltage, some of which include:

• A mechanical issue with the motor or its components, such as bearing failure or misalignment, which can cause increased friction and vibration.
• A problem with the motor’s load, such as an overload or an unbalanced load, which can cause increased current and vibration but not affect the voltage.
• A problem with the motor’s drive system, such as a malfunctioning variable frequency drive (VFD) or a problem with the power supply, which can cause increased current but not affect the voltage.
• A problem with the motor’s control system, such as a malfunctioning control circuit or a problem with the motor’s control settings, which can cause increased current but not affect the voltage.

It is recommended to conduct a thorough inspection of the motor and its components, as well as the load and drive system to identify the cause and find an appropriate solution. It is also recommended to check the wiring, controls, and settings of the motor to ensure that everything is connected and configured correctly.”

—————-

Note:  This is the exact answer I got from ChatGPT when I posed the question.  And while the answer may not be perfect, it does help the technician slim down and jump start that areas that they should explore to resolve the problem.  And the guidance from ChatGPT is invaluable in helping to guide the data science team in exploring the potential features of a Machine Learning (ML) model that can be used to help monitor, diagnose, and fix such operating problems this and other motors.

None of this great potential of an intelligent YODA sitting on our shoulder guiding our decision-making will happen if organizations haven’t started to think more holistically about how they are going to prepare and align their data and information across their organization, and across their ecosystem.  That requires a Contextual Knowledge Center.

Role of a Contextual Knowledge Center

Contextual Knowledge Center is an AI-infused, continuously learning and adapting directory that leverages an extensive knowledge graph to identify, locate, categorize, cluster, access, reuse, and refine the organization’s information assets within the content of the problem you are trying to address.

A Contextual Knowledge Center is a next generation extension of an organization’s data catalog by integrating AI/ML, contextual search, Natural Language Processing (NLP), asset scoring, graph analytics, and a decisioning (recommendation) engine to recommendation information assets based upon the context of the user’s request.

More than just another contextless (dumb) data repository, a contextual knowledge center frames the organization’s information assets in context of the organization’s business and operational objectives. The heart of the Contextual Knowledge Center is the construction of a knowledge graph (Figure 2).

Knowledge Graph is a data model that represents information as a set of entities and their relationships to one another.

Figure 2: Turn a Harry Potter Book into a Knowledge Graph

Knowledge graphs help applications understand and make sense of large amounts of complex information. For example, knowledge graphs are used in search engines to better understand the intent of a user’s query and return the most relevant results.

ChatGPT Data Management Ramifications

So far this has been an interesting conversation (at least for me), but now I want to apply this to the world of Data Management.  What role can a YODA play in accelerating my data management journey from business need to business outcomes?

For example, let’s say that I am presented with the following request:

“How do I identify and score which of my most valuable customers are at risk of attrition over the next 6 months?”

Wouldn’t it be great of my Data Management YODA came up with the following recommendation (Figure 3).

Figure 3: Data Management YODA and my Customer Retention Challenge

Again, I don’t need YODA or ChatGPT to give me the exact answer, but what if YODA could get me started with an intelligent, interactive framework around which to diagnose, explore, and validate the business request and even start to assemble some of the key data sources, ML models, analytic scores, and other relevant data and analytic assets, then I’m well down the path of accelerating the journey from business need to business outcome.

Heck, I’m one big step closer to the vision of Automated Data Management or AutoDM.

To achieve AutoDM requires an extensive data management contextual knowledge center. A data management contextual knowledge center would include internal information assets such as workflows, data pipelines, data transformation and enrichment algorithms, critical data elements, composite metrics, propensity scores, entity or asset models, ML models, standard operating procedures, governance policies, reports, dashboard widgets, and design templates.  But a contextual knowledge center could also external sources (in red in Figure 6) such as academic and industry research papers, analyst reports, press releases, patent filings, trade journal articles, published success stories, and more (Figure 4).

Figure 4: Data Management Contextual Knowledge Center

Heck, if you integrate my “Thinking Like a Data Scientist” methodology and design templates into the AutoDM solution, then I’ll probably be out of a job (or have more time for my next great adventure)!

ChatGPT Changes Everything Summary – Part II

In the same way that Netflix, Google, Spotify, Waze, and Amazon changed users’ expectations about AI-powered recommendations, ChatGPT and Generative AI will change our expectations with respect to how we want to engage with AI-infused apps and their associated operating environment.  And this could lead to the achievement of the ultimate data management vision – Automated Data Management or AutoDM.

AutoDM would be driven by an AI-fueled agent (YODA) built around an intelligent, interactive data management / value realization framework around which to diagnose, explore, and validate the user’s business request and coupled with a holistic and intelligent clustering of the organization’s information assets to unleash the value of the organization’s data and analytic assets (Figure 5).

Figure 5: “The Art of Thinking Like a Data Scientist” methodology

Yea, a Data Management YODA might make me more productive and enable me to focus my attention and teaching efforts on those next level challenges that we have in empowering a culture of AI & Data Literacy (hint, hint, hint).