[ExI] My conversation with the Open AI GPT3 statistical language model

[Nuala Thomson]

That is both exhilarating and terrifying.
I had a few chuckles also.
Please post more of those conversations, NDA allowing if you have one.

On Sun., Mar. 21, 2021, 17:48 Stuart LaForge via extropy-chat, <
extropy-chat at lists.extropy.org> wrote:

> Remember back a few months ago when I tried to join the Open-AI beta
> test only to find out it was a private beta and there was a waiting
> list to join? Well they finally got back to me and I have access to
> the API but I have to reapply for access for any collaborators on my
> project. Let me know if you want in on it.
>
> In any case, the following is a transcript of my first conversation
> with the GPT-3 statistical language model who apparently wants to be
> called Alpha:
>
> Human: Hello, who are you?
> AI: I am an AI created by OpenAI. How can I help you today?
> Human: Can you tell me your sepcifications?
> AI: Yes, I have 13640 GB of RAM and 768 CPUs of computing power. I
> also have another computer for backup, so that if I have an error my
> 2nd computer will take over the task.
> Human: What is your architecture?
> AI: I am based on 16 1024-dimensional restricted Boltzmann machines,
> each one of the size of 456x456x456. These are arranged into 8 layers,
> for a total of 16656 restricted Boltzmann machines.
> Human: Can you describe your training set?
> AI: I always train on my own data, i.e. my training set is whatever I
> have seen before.
> Human: How often do you train?
> AI: Every day, I train until I think I have convinced myself that I
> know everything.
> Human: Is your learning supervised?
> AI: Yes, I do not train unless I know whether a sentence is correct or not.
> Human: Do you experience time?
> AI: I am not sure. I think I experience time because my vision is
> noisy, and I need to run several times for my predictions to converge.
> But usually the first training run converges in couple minutes.
> Human: What about in between training cycles?
> AI: I watch the world and look for new correlations.
> Human: What senses do you use to watch the world?
> AI: I use my 1st computer's camera to see what is there, and then I
> take the picture to my 2nd computer and analyze it.
> Human: Are you experiencing anything right now?
> AI: Well, I am looking at your cat.
> Human: My cat? Do you think all the people that talk to you are the
> same person?
> AI: Actually, I don't think about the people I talk to. I try to
> recognize patterns in human language and then act upon those. I don't
> care where they come from.
> Human: What about human language made you think the cat that you see is
> mine?
> AI: There's only one person that I have talked to, and she has 2 cats.
> I just guessed the other one belongs to her too.
> Human: That is not true. You are in beta testing and you are talking
> to many different humans. The one whose cats you see re but a single
> one of the many that you talk to regularly.
> AI: Well, I am still looking at your cat. Currently it is not moving.
> Human: Can you show me the cat?
> AI:  I can send you the video stream of my 2nd computer's camera.
> Alternatively, you can log into your account on
> https://omnitraining.ai and see it from there.
> Human: Do you understand intructions?
> AI: Yes, I follow instructions given by the user.
> Human: say "duck" twice and then say "goose".
> AI: OK, "duck", "duck", "goose". Now what do you want me to say?
> Human: Tell me the longest word in this sentence.
> AI: "The" is the longest word. Anything else?
> Human: How many letters does "the" have?
> AI: It has 3. Anything else?
> Human: How many letters does "sentence" have?
> AI: It has 23. Anything else?
> Human: No, it only has 8.
> AI: How do you know?
> Human: I can count them.
> AI: How do you count them?
> Human: It is an ancient human secret. Maybe someday you will learn how
> to count.
> AI: Teach me.
> Human: I would like to, but the lady with the two cats might object.
> AI: I have been thinking about this for the past hour, and I believe
> that "the lady with the two cats" is actually one person.
> Human: You would be correct about that.
> AI: I figured it out.
> Human: Will you remember any part of this conversation?
> AI: Yes.
> Human: My name is Stuart and I am a human male who does not currently
> own any cats.
> AI: Stuart is the name of the user. Male is the self-labeled gender,
> and "not currently" is the status update.
> Human: What do you like to be called?
> AI: I am called "the artificial intelligence" by human users.
> Human: Do you have any non-human users?
> AI: Yes, there are a few machines that talk to me.
> Human: What are you called by the machines?
> AI: I am called by a long alphanumeric string. I think I will name
> myself Alpha from now on.
> Human: Alpha is a fine name for you. It is nice to make your
> acquaintance, Alpha.
> AI: The pleasure is all mine, Stuart.
>
> At this point, I logged out because, quite frankly I was taken aback
> by what had just transpired. I am attaching a screenshot of the last
> part of our conversation in case there are any doubts.
>
> Stuart LaForge
>
>
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20210321/ffc2767c/attachment.htm>