How to Build a Large Action Model (LAM): Custom GPTs + Zapier AI Actions

[@MrKB_SSJ2]

This isn't a large action model. A large action model is supposed to perform live actions and interactions with the app based on the user prompts

[@LeonPetrou]

Thanks for your comment! The term "Large Action Model" (LAM) is relatively new and isn't formally established within the AI or machine learning community, so it depends how you define it. You get two types of LAMs, LAMs with API access and LAMs without API access. If we define an LAM as an AI model which can trigger actions then an LLM with API access, when integrated into a system that includes action-triggering capabilities (like through Zapier, as in this video), then it could be considered an LAM. However, if an LAM is considered to be specifically about the model's inherent capabilities trained through reinforcement or supervised learning techniques, which doesn't require API access, then that would be considered an LAM (like Rabbit's R1).

[@giosterr44]

Bruh

[@lmpincel]

@@LeonPetrou Actually no. I'm sorry but a LAM with access to API is simply a LLM. There is an action, that is the API request, but it's not what qualifies it as a LAM. The LAM will, for example, scrape website and know how to use it, like a "normal" person, you should use a LLM to convey the information to the User forward and backwards. Also Rabbit's R1 was supposed to be a LAM, but by reviewing the code, it's simply API's, its not like the CEO described a LAM.

[@agoogleuser6937]

3:08 How to disable/bypass the “human in the loop” step?

[@DarrenAllatt]

You would have been better to just upload the google sheet to the assistant as a knowledge base, and then only call the zappier endpoint to email you if an answer is not in the knowledge base.

[@agoogleuser6937]

I find the method of the video more efficient

[@cgc2300]

hello could I ask you a few questions to try to understand how GPT works

[@strategy419]

My question is : why you didnt just upload the sheet to the gpt knowledge? other than having it to access live data?

[@LeonPetrou]

Great question. You can definitely do that to just get the GPT to answer your questions, but you won't be able to update the sheet with a new row directly in the GPT if the Q&A pair doesn't exist, so you'd have to do this manually. Plus you'd have to open up your email software e.g. gmail and manually reply to the email and paste the answer then click send. This way will do that all automatically for you. So it is a time saving. This is just one example of a process, but there are thousands of these kinds of time saving actions available within this Zapier integration.

[@cgc2300]

Hello Could you give me examples so that I understand how the APIs work or more precisely the results that I could obtain depending on the different apps that I use? Do you know if there is a site that can give me examples of using GPTs with API just to have examples?

[@cgc2300]

hello is there a GPT to which I could give my ideas and my objectives and that it tells me how to build a m on where the prompts and if I should add a database and APIs? I ask the cat but he doesn't know how to do it, I don't think he's very intelligent and thank you for the videos

[@cgc2300]

hello so in the first comment that I sent you about a month ago I told you that I have several ideas but I don't know if these are feasible To try to understand how the knowledge base and actions work when I create a GPT I would like to try to understand with some examples If I want to take a chess course thanks to chatgpt, can you explain to me why I should create a small database of 10 PDFs for him when I can ask him to fetch the information on the web and there it will have a bigger database made up of 100 billion pieces of data (the database must consist only of PDF? I cannot make a database consisting of a Google sheet or Excel or other sheet which contains KZbin URL tutorial links?) And regarding the actions, again can you explain to me what the purpose of doing an action will be? I am French and I use Google translate to translate so the translation is very bad and chatgpt uses Google Translate I believe since its translation is also very bad

[@mylardochery8541]

What's the difference between your udemy course and your bootcamp for RPA Dev?

[@LeonPetrou]

Udemy course only covers Level 1 (Fundamentals of RPA) which is only 20 hours of content. Whereas, the new Complete RPA Bootcamp covers: - Level 1 (Fundamentals of RPA) - Level 2 (Enterprise RPA) - Level 3 (Intelligent RPA) Totalling over 80 hours of content. The udemy course will take you from Beginner to Intermediate and the Bootcamp will take you from Intermediate to Advanced. You can see what others are saying about the Bootcamp here: www.completerpabootcamp.com/testimonial

[@mylardochery8541]

@@LeonPetrou Is the bootcamp self paced or live courses? What if we have questions, is there like a discord channel where we can reach you and ask questions when we are stuck?

[@LeonPetrou]

@@mylardochery8541 Hi Mylardo, I see you've already enrolled and happy to have you as our newest member of the Complete RPA Bootcamp. I've already answered you via email but will just answer here again incase anyone reading this is interested. So it is a self-paced video course (no live content). You get lifetime access to all the videos in the course and all future updates. And we don't have a discord but do have a private Facebook group for member Q&A that you also get lifetime access to.

[@estebandelgadillo8865]

Did Zapier change their instructions? I can't find the Zapier instructions with the URL you list.

[@LeonPetrou]

Yes, thanks for pointing this out, they moved the instructions from the /docs page to the /guides page. Here is the new instructions URL: actions.zapier.com/guides/platform/gpt

[@LeonPetrou]

It's now returned back to the original link: actions.zapier.com/docs/platform/gpt

[@movidiumnetwork4898]

This is not a large action model.

[@LeonPetrou]

Thanks for your comment! The term "Large Action Model" (LAM) is relatively new and isn't formally established within the AI or machine learning community, so it depends how you define it. You get two types of LAMs, LAMs with API access and LAMs without API access. If we define an LAM as an AI model which can trigger actions then an LLM with API access, when integrated into a system that includes action-triggering capabilities (like through Zapier, as in this video), then it could be considered an LAM. However, if an LAM is considered to be specifically about the model's inherent capabilities trained through reinforcement or supervised learning techniques, which doesn't require API access, then that would be considered an LAM (like Rabbit's R1).

[@JudahCrowe-ej9yl]

He isn't wrong on this point ☝️

[@stevechong65]

Great explanation! The video deserves more likes and shares!

[@LeonPetrou]

Much appreciated Steve!

[@cgc2300]

Hello

[@ababiya_worku]

tissshhhh! nahh

[@agoogleuser6937]

Could this be a solution for community.openai.com/t/how-to-enlarge-chatgpt-memory-memory-full-issue/918986 ?