Chapter 1
Introducing ChatGPT
IN THIS CHAPTER
Trying out ChatGPT
Comparing ChatGPT and search engines
Understanding ChatGPT
Choosing between ChatGPT and ChatGPT Plus
Grasping how ChatGPT is a harbinger of profound change
Conquering the fears that ChatGPT conjures
ChatGPT is a huge phenomenon and a major paradigm shift in the accelerating march of technological progression. It's a large language model (LLM) that belongs to a category of AI (artificial intelligence) called generative AI, which can generate new content rather than simply analyze existing data. Additionally, anyone can interact with ChatGPT in their own words. A natural, humanlike dialog ensues.
In this chapter, you learn where and how to access ChatGPT, why you should bother, the pros and cons of using it, and whether common fears are justified or wildly off base.
ChatGPT is often directly accessed online by users at https://chat.openai.com/, but it is also being integrated with several existing applications, such as Microsoft Office apps (Word, Excel, and PowerPoint) and the Bing search engine. The number of app integrations seems to grow every day as existing software providers hurry to capitalize on ChatGPT's popularity.
Setting Up an Account
One way to set up an account and enter your first prompt is to simply do the following:
-
Go to https://openai.com/blog/chatgpt.
Returning users can go straight to https://chat.openai.com/ and skip the rest of the steps.
- Click the Try ChatGPT button, as shown in Figure 1-1.
FIGURE 1-1: Go here to check out ChatGPT.
-
Follow the prompts to create your OpenAI account.
After you have registered for an OpenAI account, you can choose the free ChatGPT account or select the premium ChatGPT Plus subscription for a $20 monthly fee. Having an OpenAI account gives you access to other OpenAI models too, such as DALL-E and DALL-E 2.
-
When ChatGPT opens, enter your prompt (question or command) in the prompt bar.
ChatGPT generates a response.
- If you want to continue the dialog, enter another prompt.
-
When you're finished, rate the response by clicking the thumbs up or thumbs down icon.
Doing so helps fine-tune the AI model.
- Log out or simply close the window in your browser.
OpenAI's team can see any information you enter in the prompt and the entire conversation that ensues. This data may be used in training other AI models. See the disclosure in Figure 1-2. When using ChatGPT, don't disclose anything you want to keep private or confidential.
FIGURE 1-2: The ChatGPT data collection disclosure on the OpenAI website.
Comparing ChatGPT, Search Engines, and Analytics
ChatGPT is just one example, albeit the most publicly well-known, of a generative AI model. It represents a huge jump in AI capabilities.
Previously, ranking systems with more limited AI enablement would sort and rank information they found buried in massive datasets. You'll recognize examples of these ranking systems: search engines such as Google and Bing, recommendation engines used in coupon printing at retail checkout counters, GPS systems such as Google Maps that offer "near me" destination options, and personalized movie recommendations provided by Netflix and other streaming services.
Ranking systems shape how we think and function by the decisions they make in prioritizing vast amounts of information. For example, the Google search engine ranks and returns results from a user's input of keywords. Generally, users don't look further than the first three to five top ranking results. This in effect shapes our thinking by limiting the info we ingest and consider. Companies covet the top ranking spots for certain keyword results, which is why a huge search engine optimization (SEO) industry arose around keywords.
ChatGPT's capability to provide a unified answer is poised to affect our thinking and behavior to a far greater degree than ranking systems. For example, the prevailing public perception of this single-answer option is that ChatGPT is smarter, less biased, and more truthful than any other source. This perception is wrong.
ChatGPT's capability to generate new content deviates dramatically from that of previous software programs with which we are more familiar, such as other AI forms, search engines, chatbots, advanced analytics, and even business intelligence (BI) software. ChatGPT's accuracy arguably swings more widely than that of the more analytical types of software. Although I've seen poor outputs from a BI app, I've never seen one outright lie or hallucinate (generated responses that are convincing but completely wrong). But ChatGPT has shown it can do both on occasion.
ChatGPT differs from those other AI-enabled software categories because of its dialog format. Previous chatbots produce responses to natural-language queries by selecting from canned responses, meaning the content is prewritten and a response selection is triggered by keywords or the content of a user's question. ChatGPT generates its own response to the user's prompt. To the unsuspecting eye, the two types of chatbots may appear the same, but they are not.
The interaction with ChatGPT begins with someone typing a prompt in their natural language rather than in machine language. This means you can give the machine a command or ask it a question without using computer code. ChatGPT responds in the same language you're using. It continues building on the conversation as your interactions with it proceed. This threaded interaction appears as a real-time dialog and creates the semblance of a conversation or a highly intelligent response to your request.
However, the number of ChatGPT responses you can get in a single conversation may need to be limited to prevent this AI model from providing weird responses, making errors, or becoming offensive. To prevent this behavior, Microsoft limited ChatGPT in Bing to five responses per user conversation. You're free to start another conversation, but the current exchange can't continue past the capped limit.
ChatGPT generates rather than regurgitates content, which means it can make erroneous assumptions, lie, and hallucinate. ChatGPT or any other generative AI model is not an infallible source of truth, a trustworthy narrator, or an authority on any topic, even when you prompt it to behave like one. In some circumstances, accepting it as an oracle or a single source of truth is a grave error.
Understanding What ChatGPT Is and Isn't
The capability to produce a close semblance to human communication is primarily responsible for that skin-prickling feeling commonly referred to as the heebie-jeebies. ChatGPT sounds and acts almost too human.
The interaction between users and ChatGPT has a different feel than that previously experienced with other software. For one, software using earlier iterations of natural-language processing is generally limited to short exchanges and predetermined responses. ChatGPT can generate its own content and continue a dialog for much longer.
ChatGPT, like all machine-learning (ML) and deep-learning (DL) models, "learns" by exposure to patterns in massive training datasets that it then uses to recognize these and similar patterns in other datasets. ChatGPT does not think or learn like humans do. Rather, it understands and acts based on its pattern recognition capabilities.
ChatGPT supports 95 languages as of this writing. It also knows several programming languages, such as Python and JavaScript. Generative AI also differs from programmed software because it can consider context as well as content in natural-language-based prompts.
Chat in ChatGPT's name is a reference to its use of natural-language processing and natural-language generation. GPT stands for generative pretrained transformer, which is a deep learning neural network model developed by OpenAI, an American AI research and development company. You can think of GPT as the secret sauce that makes ChatGPT work like it does.
ChatGPT does not think like humans do. It predicts, based on patterns it has learned, and responds accordingly with its informed guesses and prediction of preferred or acceptable word order. This is why the content it generates can be amazingly brilliant or woefully wrong. The magic, when ChatGPT is correct, comes from the accuracy of its predictions. Sometimes ChatGPT's digital crystal ball is right and sometimes not. Sometimes it delivers truth, and sometimes it spews something more vile.
Unwrapping ChatGPT fears
Perhaps no other technology is as intriguing and disturbing as generative artificial intelligence. Emotions were raised to a fever pitch when 100 million monthly active users snatched up the free research preview version of ChatGPT within two months after its launch. You can thank science fiction writers and your own imagination for both the tantalizing and terrifying triggers that ChatGPT is now activating in your head.
But that's not to say that there are no legitimate...
