Time to catch the ChatGPT craze! Yes, everyone is flocking to the ChatGPT AI-driven chatbot and asking all sorts of life altering questions such as instructions for removing a peanut butter sandwich from a VCR in Biblical verse in Figure 1 (my answer…you’ve still got a VCR?).
Figure 1: ChatGPT in Action!
Heck, Ryan Reynolds even asked ChatGPT to write a TV ad for his company Mint Mobile.
But why is this ChatGPT fascination so important? Because in the same way that recommendations engines from the likes of Amazon, Netflix, Spotify, Google, and Waze changed users’ expectations on how they engaged with apps and websites, ChatGPT will change end users’ expectations on how they engage with their operational systems. And that’ll impact all of us.
What is a Chatbot?
Chatbots are an AI-driven computer program designed to simulate conversations with human users.
Go to almost any ecommerce or support website, and a soon a chatbot will pop up. And to be honest, most of them totally suck and are as annoying as hell.
Early chatbots were often programmed using simple decision tree structures that consisted of a series of “if-then” rules. If the user asked about this, then show that. These chatbots would evaluate a user’s input and determine the appropriate response based on the keywords or phrases that were present in the input and the underlying decision tree.
For example, a chatbot might be programmed to recognize the phrase “My computer is not working?” and respond with “Have you tried turning it on.” If the user’s input did not contain any of the keywords or phrases that the chatbot was programmed to recognize, it might simply respond with a generic message such as “I’m sorry, can you please rephrase your question?”
While this approach was relatively simple and easy to implement, it had many limitations.
- Chatbots based on simple if-then statements were not able to understand natural language or handle complex or open-ended questions.
- They were not able to engage in a diagnostic or exploratory conversation to better understand the user’s issues, objectives, and intentions.
- They were not able to learn and adapt from their interactions with users, so their responses were limited to the pre-programmed set of rules.
But ChatGPT is different, and different in ways that really matter.
What is ChatGPT?
ChatGPT (which stands for Generative Pre-trained Transformer) is a chatbot launched by OpenAI in November 2022. It is built on top of OpenAI’s GPT-3.5 family of large language models and is fine-tuned with both supervised and reinforcement learning techniques. The OpenAI–Microsoft partnership began in 2019 when Microsoft announced that it would invest $1 billion into OpenAI’s project of building Artificial General Intelligence (AGI) that had wide economic benefits.
From the article “ChatGPT-what is it and how does it work exactly?”, we get the following insights about the workings of ChatGPT.
“ChatGPT was trained using Reinforcement Learning from Human Feedback (RLHF). Reinforcement Learning from Human Feedback (RLHF) uses reinforcement learning to directly optimize a language model with human feedback. RLHF has enabled language models to begin to align a model trained on a general corpus of text data to that of complex human values.
To create a reward model for reinforcement learning, we needed to collect comparison data, which consisted of two or more model responses ranked by quality. To collect this data, we took conversations that AI trainers had with the chatbot. We randomly selected a model-written message, sampled several alternative completions, and had AI trainers rank them. Using these reward models, we can fine-tune the model using Proximal Policy Optimization. We performed several iterations of this process[1]. (Figure 2).
Figure 2: ChatGPT using Reinforcement Learning from Human Feedback to…well, Learn!
ChatGPT is actually a type of Generative AI.
What is Generative AI?
Generative AI is a type of artificial intelligence system that relies on unsupervised or semi-supervised learning algorithms to create new original digital content (e.g., articles, program code, poetry, photographs, artwork, music) by learning from existing data or content.
ChatGPT how applies Generative AI to drive a human-like dialogue is impressive. ChatGPT is a variant of the GPT (Generative Pre-trained Transformer) language model, which is a machine learning model that has been trained to generate human-like text.
However, while ChatGPT can generate text that is similar to human language, it is not able to understand or learn in the same way that a human can. ChatGPT is not Artificial General Intelligence, the holy grail of the AI community.
Artificial General Intelligence (AGI) refers to a type of artificial intelligence that is capable of understanding or learning any intellectual task that a human being can, rather than just being able to perform a specific set of tasks. AGI is sometimes referred to as “strong AI” or “human-level AI,” as it would be able to perform any intellectual task that a human being can.
By the way, AGI does not exist outside of “AI gone rogue” movies like Eagle Eye, Terminator, I Robot, and Colossus the Forbin Project. But remember that with enough smart people with enough time collaborating, exploring, and trying, sometimes “science fiction” becomes “science fact.”
Let’s go through a simple example of using ChatGPT to help plan my family vacation. That’ll give us a basis for exploring further ChatGPT applications and potential areas of evolution in Part II of this blog series.
Using ChatGPT to Plan a Vacation
(The following is an actual conversation that I had with ChatGPT)
Bill: “Hey ChatGPT, what would be a great vacation place for my family?”
ChatGPT: There are many great vacation destinations that would be suitable for a family. It really depends on your interests and what you’re looking for in a vacation. Some popular family-friendly destinations include beaches, theme parks, national parks, and cities with a variety of kid-friendly activities. Have you considered any specific locations or are you open to suggestions?
Bill: “I’m open to suggestions”
ChatGPT: Sure! Here are a few ideas for family-friendly vacation destinations:
- Disney World in Florida is a classic choice for a family vacation. There are four theme parks to explore, as well as water parks and other family-friendly attractions.
- The beach is always a great option for a family vacation. Some popular beach destinations for families include Myrtle Beach, Florida’s Panama City Beach, and California’s Huntington Beach.
- If you’re looking for a more natural setting, consider a national park. Many national parks, such as Yellowstone and Yosemite, offer a variety of kid-friendly activities, such as ranger-led hikes and educational programs.
- For a mix of city and beach, consider San Diego. The city has many family-friendly attractions, including the San Diego Zoo and Legoland California, as well as beautiful beaches.
I hope these suggestions are helpful! Do you have any specific interests or preferences that you would like me to consider when making further recommendations?
Bill: “Cost is a big factor. Which of these options are more cost effective for a family of 5?”
ChatGPT: All of these destinations can be relatively budget-friendly if you plan carefully and are willing to compromise on some amenities. That being said, some options may be more cost-effective than others. Here’s a breakdown of some potential costs for each destination:
- Disney World can be expensive, especially if you stay in a hotel on Disney property and eat most of your meals at the theme parks. However, there are ways to save money on a Disney vacation, such as staying in a hotel off property or bringing your own food into the parks.
- A beach vacation can be relatively inexpensive, especially if you’re able to find a good deal on a rental property or hotel. Beaches are also a great option if you plan to bring your own food and spend most of your time enjoying the beach and nearby attractions, rather than eating out and participating in organized activities.
- A national park vacation can be budget-friendly, especially if you’re able to camp or find a reasonably priced hotel or rental property. Many national parks have free or low-cost ranger-led activities and educational programs, which can be a fun and inexpensive way to spend your time.
- San Diego can be a bit pricier, especially if you stay in a hotel in the city and eat most of your meals out. However, there are ways to save money, such as staying in a hotel or rental property outside of the city or bringing your own food.
I hope this information is helpful! Let me know if you have any other questions.
ChatGPT Changes Everything Summary – Part I
In the same way that Netflix, Google, Spotify, Waze, Amazon, and others changed our 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.
In part 2 of this series, I will explore the ramifications of Generative AI, future ChatGPT improvements, and how something like ChatGPT to change the world of data management.
In the meantime, carve out some time and play with ChatGPT. Don’t cost nothing…
[1] ChatGPT-what is it and how does it work exactly? https://link.medium.com/EWzIlREV6vb
