Asked

Chatgpt Integration?

I have been using ChatGPT a lot lately just to see what the AI can do.  I am impressed in many ways over previous Chatbot AI engines.   Integrating ChatGPT into ARC/Synthiam environment could lead to some very sentient like robot applications.   Is anyone working on a plug in for ChatGPT?


Related Hardware EZ-B v4
Related Control OpenAI Chatbot

ARC Pro

Upgrade to ARC Pro

Don't limit your robot's potential – subscribe to ARC Pro and transform it into a dynamic, intelligent machine.

#2   — Edited

Jeremie, I should have been clearer.  Yes I have used this one that is based on GPT-3.  ChatGPT is a significant step up over the ai chatbot that is in ARC as it stands now.

PRO
Synthiam
#3  

There's some confusion about what Chat GPT is - and how you can use it. Here's an example.

User-inserted image

It is a chat with gpt. So you can chat with it. It's an interface that talks to the Open AI GPT product. So, chat gpt is chatting with the gpt. Does that make sense? Because that already exists with the Open AI robot skill. So by using the robot skill, Jeremie presented, chatting with gpt. That's the same thing, except presented as a robot skill vs a website

If you want to use Open AI's chat gpt website, you'd need to visit the chat gpt website, and you can type (chat) with the gpt. But you can't incorporate that into a robot because it's a website. If you want to use the gpt in a robot, you would add the Open AI robot skill. They both "chat" with the GPT product.

PRO
Synthiam
#4   — Edited

Oh, I should also add some more info that I think might be relevant. The GPT product is trained with data from the internet. That means the response or information you receive during the chat comes from previous conversations on the internet through forums. There is no reliability or accuracy regarding the response. The answers are not facts; if anything, they're majority opinions based on the popular redundancy of previous internet conversations.

You'll get some pretty unfactual responses if you ask the chat gpt for information about building a robot with synthiam arc. I was asking about using a camera with synthiam arc, and it was telling me to connect the camera's ribbon cable to the raspberry pi. shrug

So I guess talking to GPT is about as accurate as talking to your neighbor - or anyone you come across during the day. You'll get an opinion based on what is heard most. That's how GPT works - it'll tell you what it heard most on the internet (so to speak)

#5   — Edited

DJ, I think I got it.  I have installed and operated the AI Chatbot in synthiam using the very process described. It just feels so different than ChatGPT in its responses.  Also, mine talks as slow as Elmer Fudd. Any idea how to speed up the readback response?

PRO
Synthiam
#6  

What do you mean by "talks slow"? As in, the speech is spoken slowly?

I can say that the Chat GPT web UI uses a significantly large response size, so that's why it is so lengthy. But that would be costly to implement on a robot because the web UI is accessible, and you'd have to pay for your own application. Also, I'm not sure if you want the robot to have lengthy responses to everything.

I had read where some articles try to say the Chat GPT uses a model that has been massaged slightly for conversation. I'm not sure if that's true, but if it is, I guess there could be an advantage to creating our gpt models. We have discussed having a gpt model built around the forum and support section content of synthiam.com. It would be interesting because you'd be able to "talk to the website," I'm not sure what it entails, but it has piqued my interest.

#7   — Edited

DJ, Yes it talks very slowly.  I am wondering if that is a setting on my laptop or maybe Bing Speech? I’ll try to record it and let you hear the cadence of the speech.

W/respect to GPT, I hope you’ll look at it further. I’m quite taken back by it!