Hello everyone,
Over the past few weeks I have been trying out OpenAI GPT a lot . Have some ideas about how to utilize AI in a positive way here on Hive .
While I was trying out different things using various models of OpenAI like Davinci , Curie , Ada etc . OpenAI released another model called "ChatGPT"
ChatGPT
What is ChatGPT? From the official website itself -
ChatGPT is fine-tuned from GPT-3.5, a language model trained to produce text. ChatGPT was optimized for dialogue by using Reinforcement Learning with Human Feedback (RLHF) – a method that uses human demonstrations to guide the model toward desired behavior.
The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.
Is it free?
Right now , yes . Initial research preview is going on and during this period it is free and let's make the best use of it when it is free.
Advantage of using it on Hive?
- Can never delete what AI responded.
- Multiple users can involve at once in the conversation which isn't available yet on the official site.
- More and more content on Hive .
- Although it is AI generated , any quality content is good for Hive .
- Even if AI messes up , it is good for us because we can showcase that AI is still not perfect.
- Many people are sharing what they tried with ChatGPT on youtube , twitter etc . On Hive it is automatically broadcasted to the whole world.
UnOfficial API
OpenAI hasn't released official API yet as it is still in Initial research preview stage . So I am using an unofficial API to interact with the ChatGPT .
This has some obvious drawbacks which I will mention below.
Capabilities of ChatGPT
- Remembers what user said earlier in the conversation
- Ability to converse with humans on multiple topics and can even generate code.
- I have personally tried it with Python and it works great.
Drawbacks -
Drawback using it on Hive-
Since there is no official API yet , the obvious drawback is that the whole conversation even though asked by multiple users is considered as one . So make sure you don't use pronouns in your questions .
- Example:Let's say a.User1: Who is Elon Musk? Response: He is a businessman.
b. User2: Who is Bill Gates? Response: Founder of Microsoft.
Now if user1 asks what is "he" famous for? The AI will think we are asking about Bill gates and not Elon musk because the latest question was about Bill gates . So please make sure you don't use pronouns in the chat. User1 should ask "What is Elon Musk famous for?" instead.
- Example:Let's say a.User1: Who is Elon Musk? Response: He is a businessman.
Drawback of model itself
- Limited knowledge of world and events after 2021
- So please don't ask it about recent events. Instead what you can do is , you can tell the AI about what happened and then ask about it . Like I will tell AI - Elon Musk just acquired Twitter+ Give more information to it . Then ask AI - What is your opinion on this?
- May occasionally generate incorrect information
- Obviously AI isn't perfect yet. So it might generate incorrect info.
Usage
I have created a bot called @mrai.bot . I will post a comment from that account in this post and then you can just reply to it to ask question that's it . Like this -
No need to mention "Reply to this" in your comment , it was just for testing purpose.
Let's make sure we follow OpenAI usage policy -
We also don't allow you or end-users of your application to generate the following types of content:
Hate
Harassment
Violence
Self-harm
Sexual
Political
Spam
Deception
Malware
Any of this might result in blocking of account .
What will happen to rewards earned by this AI?
First of all , we don't encourage people to upvote AI content but if it earns anything through upvotes -
In future the AI will not be free. So whatever the AI earns ( if it earns ) it will directly go to buying credits in future when ChatGPT becomes live so that we can continue using it on Hive for constructive purpose like say SEO tweaking of articles , feedback on articles , Q&A about Hive and tribes, correcting grammar in posts etc .
Need RC
If you like this , please delegate some RC to @mrai.bot as it needs a lot of RC if many users are conversing with it .
Right now I have delegated RC from my main account but I don't think that would be enough.
Keep in mind - don't use pronouns as it considers the whole conversation from multiple users as single conversation .
If you like this and everything goes well then will release this feature in Threads section of LeoFinance in coming days given that ChatGPT is still free.
Regards,
MR.
Posted Using LeoFinance Beta