How to use the OpenAI Playground for CHEAPER generations
2023 ж. 18 Жел.
8 975 Рет қаралды
How to Build your own Custom AI app - wesgpt.mykajabi.com/
It might be time to cancel your ChatGPT Plus plan and use the OpenAI Playground instead.
Let's walkthrough the Playground and generate some text using our OpenAI API Key. We'll discover what Temperature means, how to adjust Top P, and setting Frequency Penalties.
Yes. This is very helpful. Keep expanding on this series 😁
I will for sure 😊
in chartgpt UI, you can regenerate a response with the little "rerun" icon on each response, and then you can toggle between versions. so too with edits of your input, you can track alternate versions and different streams
I totally forgot about this! You're right. Thank you 🙏
I have been waiting since yesterday for your genius advice broo❤❤❤
ha ha... same. I stayed up late waiting ... Gawd Wes is gooood.
Man, you guys are AWESOME! I appreciate you both 🙏
This was very informative. Thank you!
You're so welcome 😊
Thank you!! This is quite useful. I never thought about using the playground. Thanks!
Yeah, I'm slowly moving all of my workflows over to there! (and custom apps)
🎯 Key Takeaways for quick navigation: 00:00 📊 *Introduction: Cost Comparison and Tutorial Overview* - Introduction to the cost-effectiveness of using OpenAI Playground for generating less than 400,000 words per month. - Teaser on cleaner outputs and the upcoming tutorial on using the OpenAI Playground. 01:08 🗝️ *Setting Up OpenAI Playground and Billing Information* - Guidance on accessing the OpenAI developer platform and obtaining an API key. - Instructions on adding a payment method and clarifying that charges are based on generated tokens. 01:36 🎮 *Navigating the OpenAI Playground Interface* - Overview of the OpenAI Playground interface, including the system prompt, chat area, and model settings. - Explanation of the system prompt's role in creating custom GPTs within the playground. 03:51 🔍 *Text Generation Features and Flexibility* - Demonstrating the ability to delete and regenerate output in real-time. - Explanation of continuing conversations, changing context, and the history function. 06:08 ⚙️ *Model Settings and Customization* - Explanation of model options, including GPT 4 Turbo and GPT 3.5 Turbo, with details on their characteristics. - Overview of temperature, maximum length, stop sequence, and customization presets. 11:25 🔄 *Sharing, Code View, and Content Filter* - Instructions on sharing presets and the code view feature for API implementation. - Overview of content filter preferences for maintaining content moderation. 13:02 💻 *Usage Tracking and Cost Considerations* - Guidance on monitoring usage and tracking spending in the OpenAI Playground. - Encouragement to compare costs with ChatGPT Plus subscription for potential savings. 13:43 🚫 *Limitations: No Image Generation in Playground* - Highlighting the exclusive focus on text generation in the OpenAI Playground. - Mentioning the need for ChatGPT Plus or other tools for image generation. 13:57 🤖 *Future Tutorial Teaser: Assistance API and Custom GPTs* - Teasing a potential future tutorial on the Assistance API and custom GPTs in the OpenAI Playground. - Encouraging viewers to express interest in the comments for further content. Made with HARPA AI
Thank you for this 🙏
@@WesGPT Glad to help. Good for users and algorithm alike.
Great tutorial, made it much easier to understand all the technical terms. It seems you have two mouse cursors though, and the further from the center of the screen the more distant the cursors are from each other. It might be a recording issue.
I think this is an OBS issue - it only happens when I zoom in, correct?
@@WesGPT I think zooming in only intensifies the issue. When zoomed out the cursor still stays inside the other cursor sometimes.
Yes please continue on the subject and also interested in creating an agent to call for multiple assistants. Again, thanks @Wes
This is the current playlist (that will help you picture building Custom GPT apps alongside the course) - kzhead.info/channel/PL6_Yj1Axmkp8IiXbP4FeAZ9297PefHPPp.html
Yes, Assistants, please! Great job, by the way! Thanks!
It's in the pipeline!
Great video! Really loved the quick breakdown of the playground...you made it super easy to understand. Hey, I've got an idea for your next vid: how about showing us how to build a no-code app integration with ChatGPT? I think it'd be awesome to see an end to end take on that, including content warnings. Can't wait to see what you come up with! (this comment was generated using AI, but I swear I am not a robot.)
Can you tell me more about this idea? What type of app? I'll definitely do it 😊
@@WesGPT I was thinking of a simple workflow... User inputs a description, maybe a few checkboxes or radio buttons for turning on or off toggles and then once submitted chatgpt runs a prompt+parms to generate a dalle or Leonardo image and returns it to the webpage.
Check out this recent tool I built around DALL-E 3. Would you add anything to it? - kzhead.info/sun/grmjj6h6jn-AmGw/bejne.html
Thanks for doing this , could you please do a video on the assistants and explore that a bit more
Definitely. Assistants API video is in the pipeline 💯
@WesGPT Legend !
Great tutorial, thanks!
Glad it was helpful!
Great Tutorial 👍 Waiting for Assistants API
It's in the video pipeline!
Great video. I use PlayGround, but it would be great if you made a video on how to use the assistants, so I can get the most out of it.
That video is in the pipeline!
does it changes something if you write your context (your are best selling author Mark Manson ...) in the system window or before your prompt (i met a girl today ....)
System prompts seem to have a higher "order of importance," so it should be weighted slightly more. Though, both should do the trick.
Can I stop the model from opening every email draft with "I hope this message finds you well." by increasing the frequency penalty... ? I even tried to stop it using that by putting something in the custom instructions but it continues to use it... so bizarre. Anyway, yes yes yes... let's create GPT's using assistant. The other feature missing on the playground (besides Dall.e) is the code interpreter/data analysis. I want to be able to upload files and links and have the information simplified and summarized. Thank you SO much Wes. I appreciate you! (and the other people who comment on your videos)
Does it still start with "I hope this message finds you well." when you use the latest GPT-4-Turbo model? My initial hunch is that Frequency Penalty would NOT help this particular problem, because it's generating this phrase at the beginning and nowhere else in the text. So the Assistants API does have code interpreter as a feature! I'll be making a video about it soon 😊
It won't let me add a billing method even though I used the API. It shows personal for organization even though I tried to give it a different name
Can you buy credits instead?
Hi Wes! Is there a way we can use Playground to access the internet?
No, unfortunately not yet. This is coming to their Assistants API in the coming months, though. Check out this video: kzhead.info/sun/a6Z-orKsgJlpoHk/bejne.html
So how do I automate this to generate articles and post to my wordpress now that i have the code?
Are you talking about this free Bubble template (AutoBlogger?) If so, you'll have to download the data as a CSV and then upload to your Wordpress CMS.
Since 3rd party custom instructions are currently made for GPT+, will you please consider making a workaround to import Custom GPT instructions into Playground.
Yes, of course! This can be done with the system message OR by the Assistants API. Will make a video about both soon 😊
Another banger!
My man 🙏
do we need gpt 4 for that
Nope, you can use GPT-3.5-Turbo!
Is creating api key suposed to take long? Its been buffering for a few minutes.
It should be instantaneous. Try refreshing and starting again!
You have really great voice
Aww, thank you so much for saying that!
How to get the API of an “assistant”?
I'll be working on that video in the coming days 😊
top top top 👏👏👏👏👏👏👏👏👏
I appreciate you 🙏
great for the assistant so we don't need anny temaplte lmao i can't wait for the next vid
I'm slowly transitioning all of my Custom GPTs to the Assistants API...
Assistants. Yes, please.
Added to the pipeline!
the main difference is that ChatGPT can access the internet, and Assistant API can't. Can you do a video on setting up a function to let the assistant access the internet?
An Assistants API video is in the pipeline!
Where yu explain about cheaper?
When I talk about the pricing. ChatGPT = $20 per month. GPT-4-Turbo generations are $0.03 per 750 words. Which means, ~400,000 words per month is the same as a ChatGPT subscription.
👋
I appreciate you 🙏
Open ai playground for the YT algo. Yes interested in assistant tutorial.
Good idea! I'll think about how to make a video about the YT algo one 😊