ChatGPT and the Wonderful World of Information Personalization

Imagine having a conversation with your smartest friend, the one who knows everything about everything and everyone, except that they have no social skills or real-world experience apart from what they’ve read online. Granted, they’re voracious readers. Welcome to ChatGPT, or generative pre-trained Transformer chat box, launched by OpenAI in November 2022.

How does it work? You type a question, give it a task, and faster than you can say ‘my robot Genie,’ it delivers the answer to you. To understand it, it’s helpful to compare it to a Google search. You type in keywords and sometimes a sentence in a Google search. For example, ‘What is the average cost of a one-bedroom apartment in Portland, Oregon in 2023?’ Without a Google search, I can tell you it’s way too expensive. Google returns a list of links to websites that are likely to have the information you are looking for. I can search for listings on the second and perhaps the third page, though almost no one looks past page one or two.

w let's compare that to ChatGPT. Instead of giving you a list of web pages, ChatGPT will do its best to answer your question. Here was the response I got from ChatGPT: 'If you're looking for the most current factual information, you will still need to do a Google search.' Unlike a Google search, you can give ChatGPT feedback. It also processes the information in ways that you can specify. For example, 'Can you create a table comparing average rental costs of a one-bedroom apartment in Portland, Oregon; Seattle, Washington; Bloomington, Illinois; and Charlottesville, North Carolina in the year 2019?' Not only can I request information be delivered in a certain format, such as a table, a script, a blog post, or a cover letter, I can also request the level and kind of information. For example, 'Give me three different definitions with examples that a fifth-grader would understand.' This is just scratching the surface of what I can ask AI to do.

Instead of sending me to a web page to find the answer, ChatGPT delivers specific results tailored to my questions and requests. Given what ChatGPT can do, some compare it to having a personal assistant who is willing to help 24/7. Furthermore, ChatGPT detects emotional tone and can provide encouragement and respond as a human might.

But there are some problems too. When you get information from ChatGPT, you have no way to evaluate the credibility of the information or to know where it’s coming from. In a Google search, you are given a link to a website and can evaluate the credibility of that source of information. Because we don’t have the ability to evaluate the credibility of the information ChatGPT gives us, we run the risk of even greater confirmation bias, that is, getting information which already agrees with what we believe to be true.

With ChatGPT, the information superhighway has been replaced by an endless echoing Supermall of information that reinforces our pre-existing biases. Collectively, we have seen how the lack of a good flow of credible information reinforces not just confirmation bias, but contributes to polarization and division, the spread of misinformation and disinformation, and the undermining of democratic processes.

How is ChatGPT using the detailed personal information we so willingly and freely give it? That’s the question of who the man is behind the curtain. But before we get to that question, what about your job and mine and the future of work? That’s the subject of our next video. If you like this video, subscribe and share it with a friend.