Let me help you build your project - no fees

Hi @nelson, I am interested in attempting to utilize GPT to help streamline and advance various processes within the manufacturing sector. Everything from teaching GPT how to create g-code programs to run on various CNC machine controllers (the examples it outputs currently need a lot of work). To conversationally controlling CNC machines, and even using GPT to create 3D models in various CAD software based off uploading blueprints. Let me know if you can help or offer guidance.

Possibly, joining my project would be of interest to you: https://www.codeassist.tech . If it is, then let me know. Iā€™m continuing to improve it, itā€™s not where I want it to be yet. Thereā€™s lots of work to do.

Just a note, the email on the website was wrong - it should be contact@codeassist.tech , instead of contact@codeassist.com . I have corrected it now, but if someone sent an email to that email address, then please know that I havenā€™t received it.

Iā€™d like to fine-tune GPT3 on the works of various public domain authors, e.g. Bram Stoker, F. Scott Fitzgerald, Edith Wharton, etc.

Iā€™ve got as far as creating CSV files with the Prompt column blank and the Result column being the text of these novels, with each row a different paragraph. But I run into errors when I convert these into .jsonl files, or trying to upload them, and Iā€™m not sure what the issue is.

Iā€™d like someone who knows what theyā€™re doing to show me what Iā€™m doing wrong. Iā€™m new to all this and could use some advice. I learn best by watching people do something and then imitating what theyā€™ve done, so if you could show me how youā€™d create a fine-tuned model this way, then it would make it much easier for me I think.

Hi @schneidjar
Thank you for sharing your project with me. Iā€™m not familiar CNC machines, but perhaps I can provide some insights on how to utilize GPT with your current system. Feel free to PM or add me on linkedin, Iā€™ll be glad to provide any assistance if I can. Thanks.

Hi @damc4,

Thank for sharing your project. Would it be similar to codex?
Can you share some more information how codeassist will work in a different way?
Much appreciated.

Hi @elizabeth.m.main
Usually some common problems with converting jsonl files include handling special characters.
If this is not the case, feel free to reach out to me and show me the errors you are getting. Thanks.
Here are some examplesā€¦

1. Newline characters (\n) should be escaped as "\n" to prevent them from breaking the JSON format
2. Carriage return characters (\r) should be escaped as "\r"
3. Backslashes () should be escaped as "\\"
4. Double quotes (") should be escaped as "\""

CodeAssist is built on top of Codex.

CodeAssist has the following features that Codex doesnā€™t have:

  1. It has a natural language interface, so you can talk to it like you talk to a human.
  2. It can see your code and make changes in the file that is open in the editor.
  3. It can execute actions like: opening files or moving through a file (although it doesnā€™t execute them often for now).
  4. It can generate code that is consistent with the context of the codebase.

Additionally, Codex is an API and CodeAssist is a chatbot and a plugin/extension to IDEs like PyCharm or Visual Studio Code.

In other words, CodeAssist is like ChatGPT that works in the IDE and can directly edit the files in your codebase.

Thereā€™s more to come.

Oh I see the differences now, I can see these features being very helpful for a developer. For now Iā€™m going to focus on helping folks who are not experts in NLP and OpenAI. Best of luck to your project, and thanks for sharing your project with me.

Ok, best of luck to you as well.

Hi @nelson, itā€™s very generous of you to offer your help. I have no developer experience but I have a big idea for a functional app that allows openai api to interact with another very large database api using something like bubble.io to build it out.

My hope is that the app would allow for the NLP interface to help you search this large database returning information and ai assisted analysis and even ideas based on the search. I know that is not much info but I would be happy to talk more frankly about it if you thought it was doable and interesting enough to assist with. Thanks for taking the time to read this.

2 Likes

Hi Luke,
Thanks for sharing your project with me.
This sounds like one of the project I am currently working on, feel free to reach out to me on PM or linkedin and see if we can collaborate more on this, thanks!

@nelson Hi nelson. i just wanna need some advise for my new programmer life. I have 3 month exp in Python. At this point i just wanna start my own job in internet but not likeold version i wanna use ai for creating my web and when i give data for for my web site ai will do it for me. That is it :smile: , iā€™m not lazy but just wanna laern how to do it

1 Like

Hi @alita_lol
Thank you for sharing your thoughts with me. I think this is a difficult questions to answer because there are many different ways to build website, but I can try to break it down to two different areas.

Static Website:

If you are looking to build static website using AI, yes, you can generate html script with tools like ChatGPT and make minor adjustments to the output. However, there are many no code tools out there that you can use to build static website pretty fast. So I donā€™t think there is a lot of value on that unless you can find a way to build beautiful designs that stands out from all others.

Web Apps

If you are looking to build applications that run on the web, then you will need to build front-end, back-end, database and potential many other stacks in between. There are other tools out there that help you to build the entire stacks, but in my experience, they donā€™t work well for complex applications. So if you are able to use AI to create code for you for each layer in your stacks, I can see values in there because you will be able to build custom app faster and at a lower cost. However, I still think you will need to understands how each components work and be able to put it together yourself as well, otherwise it will be hard to support the apps you build, especially when it comes time to troubleshooting it in production.

I think advices are hard to give because it can change at different stage in your career. But after working in this field for 20+ years, the advice I will give to a younger version of myself is to find an area in technology you are passionate about, and get extremely good at it. Because if you can do that, you will make good money doing things you love.

Best of luck to you, and welcome to the programmer life.

2 Likes

There are other tools out there that help you to build the entire stacks, but in my experience, they donā€™t work well for complex applications

Had To Mention AppSheet

*While AppSheet isnā€™t an exact match for this sort of thing, it is capable of any level of complex logic you can throw at it.

  • The limitations center around the ux experience and how itā€™s rigid and not very customizable, in the sense of how most people are approaching no code app builders (drag and drop, that sort of thing).
  • Itā€™s all centered around using a Google sheet, or some other cloud hosted spreadsheet, as the database
  • using an unstructured database like this, where you can just go to your Google sheet and do whatever really, really opens up a lot of flexibility
  • it contains all sorts of ways to slice and dice your data, literally creating any level of complex data chain that you can imagine
    • (seriouslyā€¦ Iā€™m kind of looking for a challenge to be honest)
  • You can create scheduled bots that run on a timer or that are triggered from data changes
    • these have the ability to natively call an app script, and return of value which you can then continue on in your data processing chain

If youā€™re looking for a system that can take you from MVP all the way to final platform with thousands of users - AppSheet might be worth a look.


Iā€™ve personally found the combination of Google drive, sheets, AppSheet, apps script, open AI to be an incredibly easy thing to get working.

  • the hard part is now prompt engineering

But I just thought that I would throw this out there for those who are looking for a possible solution thatā€™s fairly straightforward.
AppSheet does have a steep learning curve yes, and it helps if you have some traditional developer concepts under your belt, but the community is an extremely helpful place - with a mentality and heart aligned with what Nelson has going on here.

  • And Iā€™ve personally witnessed hundreds of people significantly improve their life situations by figuring out how to use that platform.
    :muscle:
3 Likes

Thanks for sharing @multitechvisions
Itā€™s always good see others sharing their insights and knowledge while maintaining a positive and professional attitude.

2 Likes

Hello @nelson, and thanks for your work here. Seems like youā€™ve helped a great many people just in this thread alone.

My use needs are very simple. Basic Q&A for 3 categories: company service info, technical support and training data. The training I am looking to do is in the areas of dance and parkour (this data would include video links).

The problem I am running into, like many, many others here, is that fine-tuning is not doing the job. The AI still has a tendency to return itā€™s own answers rather than those in the datasets ā€“ even if the questions match word for word.

The solution (or one of them) to this is embeddings, that much Iā€™ve learned. And, your Google spreadsheet to GPT3 appears to be a very good solution. My problem is that I struggle with understanding what embeddings are (no matter how many times itā€™s been explained to me). In your Spreadsheet, I get Step 1. But, after that, I donā€™t know what to put in for these:

  1. Go to the (Completions Sheet) and fill in the context and prompt for GPT3.
  2. Fill in the values in the Context Column, in the Prompt Column we use the function getCompletions to build a prompt using either a cell or input value
    For example: getCompletions(A2, ā€œWhat is Open AI all about?ā€) will get the value of cell A2 and append "What is Open AI all about as the prompt
  3. If you like to search using your own data and feed the most relevant information into GPT3, you can store your text in the Data Sheet and use the Search sheet to create a dynamic prompt

My mind understands how to format my data for questions and answers, prompts and completions, but Iā€™m a bit lost in this embeddings thing. I donā€™t even understand the flow here: After the data is embedded, where does it actually go?

I asked in another thread if there was an ā€œEmbeddings for Dummiesā€ book I could pick up.

Can you, or anyone, point me to a step by step tutorial which would help me to understand this process and how to utilize it to get the responses I seek?

Thank you!

1 Like

Hi @SomebodySysop,

Thank you for your kind word. The concept of embedding is kind of confusing, but think of it as a language that only computer understands. It will probably be much easier if you see how the spreadsheet work with your examples. If you like, you can send me 1-10 examples of your QA and I can fill them into a spreadsheet for you.

That would be awesome! Thank you! Messaging 10 prompt-completion lines I used to fine-tune to you now. Iā€™m sure seeing examples with my data will help my mind begin to re-envision how to present it.

Oh I didnā€™t think of connecting NLP and GPT. Thatā€™s actually genius. I got some minor experience in DialogFlow, Iā€™ll give it a go.

I also asked ChatGPT for suggestions to NLP platform to use, and it really seems to like BotPress :exploding_head:

I used to be a huge fan of ChatFlow, made by Kitt.ai and owned by Chinese Google competitor: Baidu.

However, it seems the service has been shut down. Iā€™ll attach a video how it worked. Maybe someone would make something similar :innocent::crossed_fingers:t3: