Video

Would you be willing to set up a bot for me? Ill give your channel a few shout outs.

Great overview. I’m sticking with Zapier for now but also exploring Make just to learn.

I've a question. What if on the webhook node you want to reference specific data from earlier on in the zap workflow. e.g. model: "llama-3.1-sonar-large-128k-online", messages: [ { role: "system", content: "You're a smart data analyst providing summaries." }, { role: "user", content: "Provide a detailed report on the follow [Google Sheet Zap that has lists of Data fields] " } ] }; ive been trying to get the information that is sent to perplexity to be dependent on what's in my dataset but zapier keep throwing an error. It works perfectly fine without referencing prior zaps and fails whenever I reference.

Is it possible to link that fone tuned model to live data. Like train it, where to find informations in a google spreadsheet.

Great video man. You deserve more subs! 👍

Didn't know perplexity had an API! VERY useful information brother 🙌

Thank you for this video very helpful. How much did that cost you with Zappier?

Great Video! Could you also make a in debth Video from the Automation you used from the video [How To AUTOMATE Your Business in 2024 (Step by Step)] which you posted 4 weeks ago?

Thanks for your video. I would like to ask, why didn’t you use perplexity instead of using chat gpt for the last step? Many thanks

Introduction and Purpose of the Live Stream - 00:00:00 Accessing OpenAI Fine-Tuning and Tier System Explanation - 00:01:24 Threads and Interface in OpenAI for Tracking Usage - 00:03:56 Overview of AI Fundamentals and Past Content - 00:05:13 Introduction to Fine-Tuning Concepts and Neural Networks - 00:08:47 Importance of Fine-Tuning and Adjusting Model Weights - 00:12:09 Fine-Tuning Use Cases and Preparing Training Data - 00:21:07 Creating Training Data for Fine-Tuning in JSON Format - 00:24:12 Uploading Training and Validation Data for Fine-Tuning - 00:38:20 Running the Fine-Tuning Job and Understanding Parameters - 00:43:01 Live Fine-Tuning Demonstration and Analysis - 00:45:29 Exploring Fine-Tuned Model in OpenAI Playground - 00:50:46 Comparison of Standard and Fine-Tuned Model Responses - 00:52:12 Using Fine-Tuned Models in Assistants and Applications - 00:56:29 Conclusion and Closing Remarks - 00:57:17

how to make this project?

Thanks, keep going! really good video.

Great but too bad Zapier is overpriced!

This is Gold! Thank you for sharing your experience. Subbed!!

Nice video @sameer. It would be great if you create video on automating Microsoft outlook, for sending RFQ or emails to vendors with standard template which will save a lot of time & improve productivity at our workspace. Or integration of ChatGPT + outlook automation will be really helpful.

This was an absolutely amazing video! - Thank you so much for your insights and teachings 😃🙏💛 Also love that you elbowed Wized a bit about their content 😁

Wow 👀

You mentioned a more detailed discussion about the testing results for fine tuning (training loss and convergence). But I need to find out which video you mean. Can you share! This was a good video!

Thank you, this is amazing!

The Code by Zapier Step to parse JSON (use after each search): in the input data fields add two: some_text_field_with_json (use this exact name as it is in code) > put the output of any AI with Json Code below: import json def extract_json(input_string): stack = [] start_index = -1 # Traverse the string to find the complete JSON by counting braces for i, char in enumerate(input_string): if char == '{': stack.append(char) if len(stack) == 1: start_index = i # Mark the start of JSON elif char == '}': if stack: stack.pop() if not stack and start_index != -1: # When the stack is empty, the JSON is complete end_index = i + 1 return input_string[start_index:end_index] return None # Return None if no complete JSON object is found # Assuming input_data['some_text_field_with_json'] is a field from a previous step json_string = extract_json(input_data['some_text_field_with_json']) if json_string: json_data = json.loads(json_string) output = {'json_data': json_data} else: output = {'json_data': {}} return output Perplexity first search body request: { "model": "llama-3-sonar-large-32k-online", "messages": [ { "role": "system", "content": "Search for 15 random (not only top search results to minimize repetition) LinkedIn profiles based on the specified terms and return the results in JSON format. The search should focus on profiles from the United Arab Emirates that are active and seem influential and include the following details if available: complete name on LinkedIn, profile URL(i need verified and active urls avoiding not found 404 errors), current company, current location, email address, phone number, area of expertise, and profile headline." }, { "role": "user", "content": "Search LinkedIn for profiles related to the following terms: head hunter, recruiter, HR, human resources, talent acquisition, recruitment specialist, and also, as secondary criteria directly or indirectly are related procurement, supply chain, and c suite hire. Only include profiles based in the United Arab Emirates. The JSON response should have strictly this structure: {\"Profiles\": [{\"CompleteName\": \"Full name here\", \"ProfileURL\": \"Profile URL here\", \"CurrentCompany\": \"Current company here\", \"CurrentLocation\": \"Current location here\", \"EmailAddress\": \"Email address here\", \"PhoneNumber\": \"Phone number here\", \"AreaOfExpertise\": \"Area of expertise here\", \"ProfileHeadline\": \"Profile headline here\"}, ...]}" } ], "temperature": 0.5 } Perplexity second detailed search request body: { "model": "llama-3-sonar-large-32k-online", "messages": [ { "role": "system", "content": "Search online for additional contact information based on the provided LinkedIn profile details. Return the results in JSON format. The search should focus on finding the profile owner's information across various sources and include the following details if available: name, current location, current role, email address, phone number, mobile number, current company, profile headline, LinkedIn profile URL, education, experience, skills, recommendations, and connections." }, { "role": "user", "content": "Search online for the profile details based on the following information looking at all possible links for the same individual matching ANY of the following : 1-Name: {{244664614__name}}, 2-ProfileURL: {{244664614__profile url}}, 3-profile headline: {{244664614__headline}}. The JSON response should have strictly this structure: {\"Profile\": {\"Name\": \"Full name of the person as displayed on LinkedIn, identified value to replace value here\", \"CurrentLocation\": \"Current city and country where the person is located, identified value to replace value here\", \"CurrentRole\": \"The job title or position the person currently holds, identified value to replace value here\", \"Email\": \"Email address if available or inferred from other sources, identified value to replace value here\", \"PhoneNumber\": \"Landline phone number if available, identified value to replace value here\", \"MobileNumber\": \"Mobile phone number if available, identified value to replace value here\", \"CurrentCompany\": \"The company where the person is currently employed, identified value to replace value here\", \"ProfileHeadline\": \"The brief description or tagline under the person's name, usually summarizing their professional identity, identified value to replace value here\", \"LinkedInProfileURL\": \"The URL of the LinkedIn profile, identified value to replace value here\", \"Skills\": \"List of comma seperated string of skills and endorsements the person has received, identified value to replace value here\", \"Connections\": \"The number of connections the person has on LinkedIn, identified value to replace value here\"}}" } ], "temperature": 0.5 } Code by zapier step to make sense of HTML: in the inputs: html_response > response from get request Code: const htmlResponse = inputData.html_response; function checkPageStatus(html) { // Check for indicators of an active LinkedIn profile page if (html.includes('window.location.href') && html.includes('/authwall?trk=')) { return 'Active (Requires Authentication)'; } else if (html.includes('Page Not Found') || html.includes('Profile Not Found')) { return 'Not Found'; } else { return 'Unknown Status'; } } output = [{ "status": checkPageStatus(htmlResponse) }]; >>>>Second way through code by zapierto confirm linkedin URLs, better tested after vid (recommended - no get request only this-<<<< input is URL > insert the LinkedIn URL code: import re import requests def validate_linkedin_url(url): def is_valid_url(url): try: headers = { 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36' } response = requests.get(url, headers=headers, allow_redirects=True, timeout=10) return response.status_code == 200 except requests.RequestException as e: print(f"Error checking URL {url}: {str(e)}") return False # Normalize the URL url = url.strip().lower() if not url.startswith(('', '')): url = '' + url # Check original URL if is_valid_url(url): return {"valid": True, "url": url} # Extract username from the URL username_match = re.search(r'(?:\/in\/|\/pub\/)?([\w\-]+)(?:\/|$)', url) if not username_match: print(f"Could not extract username from URL: {url}") return {"valid": False, "url": url} username = username_match.group(1) # List of possible LinkedIn URL formats url_formats = [ f"www.linkedin.com/in/{username}/", f"www.linkedin.com/pub/{username}/", f"linkedin.com/in/{username}/", f"linkedin.com/pub/{username}/" ] # Try each format for fixed_url in url_formats: print(f"Trying URL: {fixed_url}") if is_valid_url(fixed_url): return {"valid": True, "url": fixed_url} # If no valid URL found, return the original invalid URL print(f"No valid URL found for username: {username}") return {"valid": False, "url": url} # Zapier input input_data = { 'url': input['url'] # Replace with actual Zapier input } # Process URL and get result result = validate_linkedin_url(input_data['url']) # Zapier output output = { 'valid': result['valid'], 'url': result['url'] }

If you could also give the code or steps on how one can implement (github etc) it so that it’s actually helpful for the community, it would be great.

Thank you It would be lovely if you could create a process video that could cover this scenario A daily technology update from google news saved in Airtable The link is researched on Perplexity Outcome sent to ChatGPT for content repurposing - blog / shorts / instagram reels The outcome save in Google docs or in airtable for review When approved, automatically posted on LinkedIn, Facebook or Insta

I just put a complicated math word problem into bot Claude and ChatGPT. Claude failed hard. It even showed simple math equations that were clearly wrong and then came back with one of those "thank you for pointing that out" replies. ChatGPT solved the problem in a completely different way and correctly.

So, what I missed to get the Artifacts is turning it on. To do that click on your name icon of your account top right, then go to feature preview and once there toggle on Artifacts

Great video this would be a gamechanger!

How to update my existing Fine-tune model without create new model

Thanks for this video man you are doing a amazing job. A costumer asking me for a assistant doing exact this what you are showing in this video. I was starting by myself but you helped me a lot. Thanks

Thanks community for having my back! i cut out the showcase part as I realized maybe too late that when you talk to GPT-4o on the mobile app it is actually still 4 turbo! Well that's something I leaned about going live on whime to get a scoop, you tend to get thing wrong :) I appreciate everyone who made a comment and alerted me, and apologies if this disappointed some

It is a more natural coouter-human interaction when computer give the same boring useless responses like humans ? "It's great to be here", "How are you today", "Oh its so great you asked this" blah blah for nuttin. This bots are also trained to get anxiety and depression most of us-folks need to have ?

I cant wait till the app sees live video

That's regular GPT 4 bro

GPT4o voice has still not been released to the public, the version you are testing of the voice is the same as the GPT-4. Later this month they probably release the full version.

thumb up for your action you show up the whole steps. it is a little diff fron official demo due to latency

I really enjoyed this video. I would highly recommend it for other learners. Thank you for such a nice job

Groq is running the model through an LPU. It's always faster, regardless of the model. It spits out 900+ T/s on the 8B model.

ChatGPT Summary to avoir the video : It seems like you've conducted a thorough comparison between the three language models: ChatGPT4, Microsoft Copilot, and Lama370B. Let's summarize your findings: ### Summary of Comparison: 1. **Speed:** - Lama370B was the fastest in providing solutions, followed by Microsoft Copilot, and then ChatGPT4. 2. **Complexity and Detail:** - While Copilot tended to simplify solutions, ChatGPT4 and Lama370B provided more detailed and complex responses. 3. **Handling PDFs:** - Copilot disappointed as it didn't handle PDFs directly, requiring integration with the Microsoft ecosystem. - ChatGPT4 excelled in handling PDFs, demonstrating its strength in document-based tasks. 4. **Cost Consideration:** - Copilot comes at a higher cost, especially for the business version, compared to ChatGPT4. - Lama370B's cost-effectiveness depends on factors like hosting and usage, especially if integrated with Azure. 5. **Tool Integration and Familiarity:** - ChatGPT4's integration with tools like Zapier and familiarity were cited as advantages, contributing to ease of use. 6. **Overall Decision:** - Despite Copilot's strengths, you've decided to stick with ChatGPT4 due to factors like complexity, cost-effectiveness, and familiarity. ### Conclusion: Your decision to stick with ChatGPT4 seems well-considered, taking into account factors like functionality, cost, and integration with existing workflows. It's essential to choose the language model that best aligns with your specific needs and preferences, and your thorough evaluation process should serve as a helpful guide for others navigating similar decisions. If you have any further questions or need more assistance, feel free to ask!

Hi Samer, Your AI video was amazing! I liked how you broke down such vast topics into something everyone can grasp. The references you included are super valuable for anyone who wants to learn more. Subscribed!

The repeating of the text is the fault of however that Azure playground is processing responses not the fault of the Model. You can see the application missed a stop token in the text, that is just a bug in Azure unfortunately.

What I absolutely disliked about local LLAMA3 was the data theft going on in the background. I was at no point made aware of the fact, that this thing sends EVERYTHING back home, for data mining.

A power packed content. Thanks for the amazing video

Nice video. However when i try the same, I get a method not allowed error on running the endpoint generated. Any idea what that's about?

Great video! Can't wait to see the next part 😄

Great content. Zapier is so powerful. Keep it going 🙏

test test chat gpt hahah i dunno if it will reply to this!

Hello, you are really doing a great job. Lots of inventiveness and creativity which makes me always surprised and delighted to watch your videos. Thank you !

good job bozo

Habibi ya Samer now, listen. .... What is the return of investment of creating my own GPT in the marketplace? ....Ok? Now Second. ....What are the metrics contributing to Obesity? Let's hold there and revisit.

One of the best videos, thank you for taking the time to explain it thoroughly

How do I deploy the bots?

Please register below to gain access to free cool stuff I share! givemethemicofficial.com/collections/exclusive-access-boutique