Communication of an algorithm put together with chat gpt to program or operate something else can does be done?
You can query external sources so yes, even with basic ChatGPT you can interact with and operate something else. GPTs can access APIs and the API you can do lots of interesting things with.
Query external sources means?
Sounds like dropping a lot of information then building with that information.
By query I mean you can access and search information on an external website… If you can call a website URL with a simple query you can execute a specified function externally to a chat.
ChatGPT will also read the contents on the queried webpage and then include this in it’s knowledge for the current chat but it wont copy 100% accurately (At least I dont think so) it will understand a summary of the contents.
Here I search for this forum thread and summarize it
I plan to design an engine for this to help
The thing to note is that while it does access webpages this method also hallucinates so it may return false information. It may sometimes return some made up information without even accessing a webpage. It is ‘non-deterministic’. Not 100% reliable to follow the same process every time.
Designing an engine than can it recognize the speed of what it’s moving would help
I built a gpt if you’d like to test it
accessing the link in a web browser from chat gpt?
Hey,
Sorry busy week.
You can find an example of a GPT accessing an external webpage here.
There have been some updates to GPTs and this sometimes breaks their functionality but this example appears to work well.
An installer,agent, macro to do?
I am not entirely clear what you are trying to do.
In your original question you asked if you could access an external source in Chat GPT.
ChatGPT
The answer to this is yes, ChatGPT can access/read external URLs in limited contexts.
ChatGPT will access them kind of like a person reading a webpage, and then doing a reading comprehension afterwards from your queries.
ie it wont be able to tell you exacting details from the page 100% accurately but will be able to summarise it
GPT Builder
You can build ‘Custom GPTs’ which can access external APIs. I haven’t tested the accuracy of these but it is better than the method I showed above for accessing external sources.
API
If you want full interactive access to exact data on webpages and advanced interactions with external sources you would need to use software and OpenAI API.
You have a further question
I take it that you mean the results are not always 100% accurate. This is where you would use logical programming languages and API.
My goal in this example was to show how ChatGPT/GPTs are able to operate and interact with external sources in a basic way as per your original question.
In this case it reads all the issues from a post from the OpenAI forum
The post is describing:
Dall-e image creation best practices
- Handle blocked prompts by rephrasing or simplifying them due to strict content policies.
- If nonsensical text appears, try rephrasing the prompt to avoid DALL-E 3’s text misinterpretation.
- Manually save metadata for prompts, as DALL-E 3 does not auto-save them.
- Address orientation and negation issues with precise wording; avoid overly complex prompts.
- Use caution with common templates like "alien" or "moon" for unique or novel results.
- Simplify prompts to improve output consistency and quality.
- For photo-realistic images, specify “photo style” rather than “realistic.”
- Avoid excessive embellishments to help DALL-E produce clearer visuals.
- Take advantage of DALL-E's landscape generation capabilities and straightforward stylistic requests.
It rolls it into a macro so you can create images simply and avoid common issues
Image("prompt", "size", "style")
Image(“prompt”, “size”, “style”)
{
- Handle blocked prompts by rephrasing or simplifying them due to strict content policies.
- If nonsensical text appears, try rephrasing the prompt to avoid DALL-E 3’s text misinterpretation.
- Manually save metadata for prompts, as DALL-E 3 does not auto-save them.
- Address orientation and negation issues with precise wording; avoid overly complex prompts.
- Use caution with common templates like “alien” or “moon” for unique or novel results.
- Simplify prompts to improve output consistency and quality.
- For photo-realistic images, specify “photo style” rather than “realistic.”
- Avoid excessive embellishments to help DALL-E produce clearer visuals.
- Take advantage of DALL-E’s landscape generation capabilities and straightforward stylistic requests.
}
You could then call this macro when creating images without having to type lots of caveats on generation
Image(“Teddy Bear”, Landscape, “Cave Painting”)
There is a progression in expertise required to develop on the platform
ChatGPT
Custom GPTs
API - Requires some programming knowledge to start
ChatGPT and Custom GPTs are mainly for simple chats but with a little additional functionality.
Tried designing a logical language to generate images
Designing an image generator capable of generating images of custom Gerber files, custom Arduino boards, custom boards
If you provide specific requirements (e.g., microcontroller type, features, or board size), I can guide you in designing an example Arduino schematic, PCB layout, and Gerber file, and generate corresponding visuals
Any suggestions, advice or examples of this to test it?
I’ll have to leave your specific GPT designs and testing to you but happy to help with any method improvements in the context of OpenAI products/features where I can.
Challenges and Solutions
Challenge | Solution |
---|---|
High Complexity of PCB Design | Focus on beginner-friendly features with tutorials. |
Compatibility with Manufacturing | Partner with popular manufacturers like JLCPCB. |
AI Accuracy in Suggestions | Train models using real-world design examples. |
Market Competition | Highlight AI-driven simplicity and cost-efficiency. |
Example Workflow for Users
- User Input: “I want a Wi-Fi-enabled Arduino board with 10 GPIOs.”
- AI Suggestion: Propose ESP32 microcontroller with power and I/O components.
- Schematic Builder: Drag-and-drop components into the editor.
- PCB Layout: Auto-route traces with AI suggestions for placement.
- Validation: AI detects issues (e.g., unconnected pins) and provides solutions.
- Gerber Export: Generate files and share with a manufacturer.