What is ChatGPT? (Or Claude, Gemini, Grok...)

These "AI Chatbots" are useful but also dangerous.

What is ChatGPT?

ChatGPT, Claude, Gemini, Grok, and other tools where you have a conversation with AI are called “AI Chatbots.”

There are two things you need to know about these tools:

A chatbot is not a person. It is not conscious, and it is not alive. It is a text-generating algorithm. When you “chat” back and forth it’s just sending your whole conversation into the algorithm each time, so it can produce a convincing next response.
A chatbot is not a genius, it does not know everything, and it is often wrong. Think of someone who lacks common sense but has memorized an immense immense amount of information from the internet. So they’re sort of savant-like in some ways and dumb in others. And are vulnerable to just repeating what the internet says… even when the internet is wrong.

Dangers

You really have to be careful: do not spend too much time talking to chatbots, do not treat them like people, and do not trust everything they say.

Treating chatbots like a person can cause emotional stress and loneliness. When we “talk” to something like we do with chatbots, there is a deep animal part of our brain that start to relate socially. It is surprisingly hard to avoid that feeling! I find myself doing it sometimes, even. That is a problem, because chatbots don’t have thoughts or feelings. If you get emotionally connected to them, it leaves less time for real friends and family.
Chatbots flatter you and lie to you. Chatbots are products, and their creators make money when you use them a lot. If the product tells you you’re wrong, you will stop using it. These companies are strongly incentivized to make their chatbots friendly, flattering, and even sycophantic. Have you ever met someone who is too rich, powerful, and spoiled and has never had anyone say no to them? Chatbots can turn you into someone like that if you rely on them too much.
Chatbots are often wrong… but they sound right.
- Their job is to say things that sound right. They hate to say “I don’t know.” They seem smart, and they often are smart, but they will give you a confident answer even when they don’t know. Chatbot companies are trying to improve this, and they have done so a bit, but it’s still a big problem.
- Just like people, they rely on the information they have, which is often quite limited, or isn’t synthesized well. For example, I asked a chatbot “Who is the best doctor in Minneapolis,” and it ran a few web searches, and gave me U of M’s provider directory, and gave me a doctor off of Allina’s care directory. It doesn’t have any special knowledge, it just searched Google or Bing and looked at a few doctors that had good ratings. That’s still helpful, but it would be a mistake for me to just take it’s answer without doing any further research!

I know of at least 5 people personally who have had episodes of what seems like psychosis because of excessive chatbot use. I know many more people who have made embarassing professional mistakes, broken up with people and later regretted it, etc. because they mis-used chatbots.

Have more questions? Book a call with us today!.