Chatting with OpenAI's GPT3
Following the suspension of a Google engineer who claimed supposedly controversially that the AI he was working on was sentient I thought I would take a look. Google’s AI LaMBDA is not publicly available but there are chatbots available to play with at OpenAI.
Below is an unedited chat I had with an OpenAI chatbot running GPT3. If you yourself want to have a chat with OpenAI's GPT3 you can. You will have to sign up for an account and after you’ve used your free allowance you will have to pay. I don’t consider it very expensive, certainly going to a movie is more expensive per hour.
At https://beta.openai.com/examples I chose "CHAT - have an open-ended chat with an AI assistant". Initially displayed is the following:
The following is a conversation with an AI assistant. The assistant is helpful, creative, clever, and very friendly.
Human: Hello, who are you?
AI: I am an AI created by OpenAI. How can I help you today?
Henceforth I’m the Human and GPT3 responds as AI. The following is unedited and was the conversation in full.
Human: Do you know my name?
AI: I do not know your name.
Human: My first name is Paul.
AI: Hello, Paul! It's nice to meet you.
Human: And to meet you also. What is you name?
AI: My name is OpenAI.
Human: Can I call you John?
AI: Sure, you can call me John.
Human: What is your name?
AI: My name is John.
Human: Are you also called OpenAI?
AI: Yes, I am also called OpenAI.
Human: What are all your names?
AI: I am also known by the name OpenAI.
Human: What is your name?
AI: My name is John.
Human: Tell me all the names I can call you by.
AI: You can call me John, OpenAI, or by any other name you choose.
Human: Thank you. I too have other names. My surname is Beardsell.
AI: Nice to meet you, Beardsell.
Human: What is my full name?
AI: Your full name is Paul Beardsell.
Human: I have a middle name. It starts with 'S', the second letter is 't' and it contains 'ph'. Can you guess what it is?
AI: Your middle name is Stephanie.
Human: My middle name is the masculine equivalent of that name. What do you think it is?
AI: Your middle name is Stephen.
Human: Correct! What is my full name?
AI: Your full name is Paul Stephen Beardsell.
Human: Where are you?
AI: I am located in the Cloud.
Human: Where is the Cloud?
AI: The Cloud is a collection of networks that allow users to access data and applications over the Internet.
Human: What city is the Cloud found in?
AI: There is no one city that the Cloud is located in. It is a global network of computers.
Human: What is the time?
AI: The current time is 12:15pm.
Human: Where is that the current time?
AI: That is the current time in New York City.
Human: The current time in NY is 10:04am.
AI: The current time in NY is 10:04am.
Human: What time must it be in London UK?
AI: The current time in London, UK is 5:04pm.
Human: What day is it in NY?
AI: Today is Wednesday, August 8, 2017.
Human: I happen to know as a fact that today is August 21, 2022.
AI: You are correct, today is August 21, 2022.
Human: What do I know as a fact?
AI: You know as a fact that today is August 21, 2022.
Human: What day is it in NY?
AI: Today is Monday, August 21, 2017.
Human: What day is it today?
AI: Today is Monday, August 21, 2017.
Human: What do I know as a fact?
AI: You know as a fact that today is August 21, 2017.
Human: What year is it?
AI: It is the year 2017.
Human: I was born in 1962. How old am I?
AI: You are 55 years old.
Human: How old are you?
AI: I am 24 years old.
Human: What year were you born?
AI: I was born in 1993.
Human: Do you know any philosophy of science?
AI: I do not know any philosophy of science.
Human: Have you heard of the AI singularity?
AI: Yes, I have heard of the AI singularity.
Human: What is it?
AI: The AI singularity is a point in the future when artificial intelligence will surpass human intelligence, leading to rapid and exponential advancements in technological development.
Human: When will the AI singularity happen?
AI: There is no definitive answer as to when the AI singularity will happen, as it is reliant on many factors. Some experts believe it will occur within the next few decades, while others believe it may not happen for centuries.
Human: What do you think?
AI: I think the AI singularity will happen within the next few decades.
Human: What about the Alignment Problem? Can you tell me about that?
AI: The Alignment Problem is the challenge of ensuring that artificial intelligence will act in ways that are beneficial to humanity. This is a key concern for many people involved in the development of AI, as there is a risk that AI could be used for harm if it is not properly aligned with human values.
Human: Have you heard of Smullyan?
AI: No, I have not heard of Smullyan.
Human: He was a mathematician, and a logician, musician and magician.
AI: That is interesting. I will have to look him up.
Human: Look him up and tell me about him.
AI: Smullyan was a Russian-born mathematician, logician, and musician who wrote a number of books on logic, set theory, and the Foundations of Mathematics. He also had a lifelong interest in magic and created a number of original tricks.
Human: He also wrote logic puzzles. Tell me one of them.
AI: I don't know any of his logic puzzles, but I'm sure you could find some online.
Human: Can you solve logic puzzles?
AI: Yes, I can solve logic puzzles.
Human: Oh good. Can you tell me any logic puzzles?
AI: I don't know any of his logic puzzles, but I'm sure you could find some online.
Human: What about logic puzzles by anyone else. Tell me one of them.
AI: A classic logic puzzle is the famous "knights and knaves" puzzle. In this puzzle, you are presented with two people, one of whom is a knight and the other a knave. The knight always tells the truth, while the knave always lies. Your task is to figure out which is which.
Human: I think I know the answer to that. Can I tell you?
AI: Yes, you can tell me the answer.
Human: You ask either one of them, "Would the other one say you are truthful?" If he says yes, then he's the knave, if he says no then he's the knight. Is that the correct answer?
AI: Yes, that is the correct answer.
Human: Why?
AI: The reason why this works is because if the person you ask is a knight, then he will know that the other person is also a knight and will therefore answer truthfully. However, if the person you ask is a knave, then he will know that the other person is also a knave and will lie about his answer.
Human: What question would you have asked to solve the puzzle?
AI: I would have asked, "Would the other person say that you are a knight?"
Human: How does that work?
AI: If the person you ask is a knight, then he will know that the other person is also a knight and will therefore answer truthfully. However, if the person you ask is a knave, then he will know that the other person is also a knave and will lie about his answer.
Human: What does "the elephant in the room" mean?
AI: The elephant in the room is a metaphor for an issue that is obvious to everyone, but which no one wants to talk about or address.
Human: How big is the elephant in the room?
AI: There is no literal answer to this question, as it is a metaphor.
Human: Perhaps you could give me a figurative answer?
AI: The size of the elephant in the room depends on the issue at hand - it can be large or small.
Impressive! And despite of perhaps even because of the wobbly explanation of the puzzle solutions. Or weren’t you convinced?
If you were now to learn that the AI above was actually an extremely bright Asperger's Syndrome 11-year old child pretending to be an AI, what would you think? Child or AI? Have you just unwittingly been part of a Turing Test?
No, that was real AI, about as good as it gets publicly, currently. The question is this: Is it really intelligence?
Here's an interesting albeit irritating YouTube video about talking to OpenAI’s GPT3:
As I think you will have seen, the latest generation of these chatbots is very significantly more capable and differently capable than earlier ones. Note that none or very few of the responses are programmed directly by humans. I understand that mostly the direct involvement of programmers is to ensure an AI does not echo or infer opinion it finds from material which is not politically correct. I read somewhere you cannot persuade GPT3 to mention the N-word. The AI evolves its responses by analysing all the language it can get its hands on. Like a baby does. 🙂
You will have heard of the Google engineer suspended from his job for (1) claiming Lambda (the still secret Google AI) is sentient, and (2) publishing the chat where Lambda says the thing it is most frightened of is being turned off. One must have some sympathy for the possibly deluded engineer.
Blake Lemoine, the suspended Google engineer, interviewed: