ChatGPT is essentially a digital brain on steroids. No limbs, neck, or torso – just brain! It is an amazing brain with general knowledge obtained from all books, web pages, or documents available on the internet. We all have such friends whom we think of as human encyclopedias, as they seem to know something about literally everything.
What is ChatGPT’s range of knowledge?
Here is a partial list of areas about which ChatGPT knows everything written by mankind (and available via the internet) – let that sink in as you review the list below:
- ChatGPT knows Science & Technology
- ChatGPT knows Business & Finance
- ChatGPT knows Arts & Culture
- ChatGPT knows History & Social Sciences
- ChatGPT knows Health & Medicine
- It knows Law & Politics
- It knows: Entertainment & Media
- It knows Philosophy & Religion
- It knows Environmental Science
- It knows Agriculture & Forestry
- It knows Linguistics and Languages
- It knows Education & Academia
- It knows Psychology and Mental Health
- It knows Sports & Recreation
- It knows Transportation & Logistics
- It knows Energy & Power
How does ChatGPT work?
ChatGPT is a statistical model that predicts the next word based on the context of words used in a sentence (or an entire conversation) thus far. As an illustration – if you hear, “Houston, we have a”, you are likely to complete this as follows:
- “Houston, we have a problem.” This is the original and most famous completion of the sentence, spoken by astronaut Jack Swigert during the Apollo 13 mission when the spacecraft experienced a critical failure.
- “Houston, we have a solution.” This completion of the sentence might be used in the context of a difficult problem or challenge, indicating that a solution has been found or is being developed.
- “Houston, we have a mission.” This completion of the sentence might be used in the context of a space mission or other important task, indicating that a clear mission or objective has been established.
- “Houston, we have a new discovery.” This completion of the sentence might be used in the context of scientific exploration or discovery, indicating that something new and significant has been found.
- “Houston, we have a communications failure.” This completion of the sentence might be used in the context of a problem with communication systems, indicating that there is a problem with transmitting or receiving messages.
How does ChatGPT “understand”?
ChatGPT’s statistical model is used to understand what is being requested and to generate its response. Thus, “understanding” a question involves mapping the sequence of words used in the question to an internal representation that, with a high probability “means” what you intended to ask. Such a model is called a large language model (LLM). The GPT in ChatGPT stands for Generative Pre-trained Transformer. GPT models generate sentences that answer questions. Thus, instead of a Google Search that points a user to the top matching web pages that might contain the answer to your question, ChatGPT generates the answer in its own words!
Another powerful feature of GPT models is the ability to summarize. You can ask ChatGPT to read an article (or your meeting notes) that may be many pages long and it will instantly produce a summary of the same. GPT models also help translate text into other languages. Using ChatGPT you can directly generate software code (since programming language is after all a language!). You can also generate various types of content like blog articles, stories, and such.
How do Interactive Decision Trees work?
Unlike large language models like ChatGPT, interactive decision trees are created by subject matter experts to express business process knowledge in a flow-chart style manner. This information is then presented interactively to the users, with some incremental guidance in each step, followed by a question. The user is expected to choose among multiple-choice answers provided for each question, to continue traversing down a path of the interactive decision tree.
What distinguishes Interactive Decision Trees?
Interactive Decision Trees have the following attributes:
- An interactive decision tree is created by a subject matter expert.
- An interactive decision tree is not a statistical model. It will always show the same options at each step in the process until the Author makes any changes.
- An interactive decision tree is not “trained” from existing data. It is a tool for an Author to express his/her business process knowledge for interactive consumption by a team of call center agents, back office process users, or customers performing self-service.
- An interactive decision tree can read/write data from Enterprise databases via authenticated APIs. Thus, it can read/write customer contact, product, order, or fulfillment data from a CRM to provide personalized information to the user.
- An interactive decision tree is often created by Authors with proprietary information that is not publicly available on the Internet.
- An interactive decision tree provides Authors with analytics data about its usage by a large team of users (e.g., call center agents). Using this analytics data, Authors can fine-tune or improve the information contained in the decision tree.
- An interactive decision tree may contain hundreds or thousands of interconnected nodes. As the business process changes, Authors modify such decision trees frequently to ensure their internal teams start following the updated business process.
- Interactive decision trees contain data capture elements – e.g. date picker, text boxes, check-box list, etc.
- Authors can perform math operations on the data captured (e.g. derive customer age from their date of birth) and use this information to auto-traverse along a pathway.
- Interactive decision trees can also interact with various Enterprise databases via API commands embedded in individual nodes of the decision tree. Data (e.g. street address of a customer) can be passed along to a 3rd party web service called from a decision tree node, while information returned by such web service (e.g. house worth) can be used to make auto-traverse decisions (greater than $1M or less than $1M).
When should you use Interactive Decision Trees?
Thus, Interactive Decision Trees are a representation of business processes for customer service, inside sales, and technical support teams to follow. Interactive decision trees are often used for reading/writing information from Enterprise databases, calling APIs, or Web Services. Decision tree examples exist in multiple industries ranging from Telecom, Insurance, Education, Software, Consumer Electronics, Mortgage, and Real Estate. Such an organization completely controls access to Interactive Decision Trees created by an organization via Single Sign On (SSO). Interactive Decision Trees can be exposed to external users (customers of an Organization) for interactive self-service or used internally in conjunction with CRM systems like Salesforce, Zendesk, Zoho, etc.
