วีดีโอ

If it doesn’t know, you just feed it a few examples of the style. Only saying to check if it knows first before going through the trouble of providing it an example 😅

@Mark_Kashef but cannot give data options if it's not trained on it. Yes it can add your data you give it. If you do more work

Hey! 👋🏽 so happy they’re helpful for you. 1. Depends fully on the task - if it’s creative writing, I would say it’s a combo of both the creativity component & clear and concise instructions to keep the creativity in check. 2. Not at this moment! We’re scaling the team as we speak, so we still have a ways to go until we’re mature enough to have the SOPs and program to enable that .

​@@Mark_Kashef 1. Appreciate your answer! I guess based from this, it is safe to assume that there is still no 'one size fits all' prompt structure and that it will still go down to what the objective is plus it will mainly lean towards personal preference. 2. Gotcha! I will be on the lookout and hop on the opportunity once you got it running. Thank you!

Thanks Edward!!

hey! The best way to use this is setup a webhook in Bland from Make.com to execute and 'listen' in to the trigger.

@@Mark_Kashef thanks for the tip. I'll make some more progress this weekend.

Our pleasure! Glad you’re enjoying them 👌🏽

appreciate it!

Pleasure to hear that!!

Same

I wouldn't consider it going head-to-head but rather collaborating aha, would love to take a look at what you've come up with and see if I have any 2 cents I can offer 😁

@@Mark_Kashef Yeah, that was my intention. But, I'd like to see each other's results first. That's what I meant by head-to-head. I've got nothing but great respect you and others who are giving such great insights and lessons on all of this. So, thanks again and if you want, I'll send you an email to discuss further.

@@wunvitas sure! send over an email.

do you mean templates embedded within the training itself? I would imagine they’d rather not constrain the model and put it in a box that’s stuck to a few templates.

You can ask it for its knowledge documents in code block format usually.

it alludes to conversational pathways in bland that you'd typically have to build out yourself. our original video is on it here: th-cam.com/video/qbYv7FqxRYs/w-d-xo.htmlsi=a-kElMzdrR-sFJKw

Our pleasure! Thanks for the feedback 🙏🏻 Would strongly recommend you buy a new number to run as a pilot before making any routing to your current office number. In case it’s buggy or catches clients off guard, we usually recommending routing clients to the AI receptionist number knowingly at first, then thinking about making it your main number after it’s stable and reliable.

thanks for watching!

My pleasure!!

Hey! The steps to do this could include one method done in two ways that come to mind: TLDR; VoiceFlow -> embedded on website Option 1: create an assistants app loop in VoiceFlow -> create an assistant in the OpenAI playground -> note down the id -> embed on website Option 2: use VoiceFlow’s internal knowledge base and upload your docs there, then use AI steps with reference to the knowledge base and prompts that you loop through (easier, less accurate, more prone to no match issues). But you can use any of the LLMs they host.

@@Mark_Kashef thank you.

hahahah too funny! Don’t worry longer vids coming, just dropping a quick one 🫡

our pleasure!

Custom GPTs hallucinate quite a bit from session to session, I use the assistants playground or assistants API to handle these types of back and fourths, as they’re more reliable I find. For a custom GPT, the best way to increase the consistency is have a conversation starter that you hit every time (“I have a something to discuss” or whatever) that you reference in the underlying prompt to trigger a certain type of initial response.

@@Mark_Kashef very good idea, to add a trigger! thanks!

I love this! That’s exactly the experience that motivated this video 😅 We spent the last year trying to read its mind and come up with a million acronym templates to structure prompts, when we could have just asked it to show us its thinking process so we could course correct it

@@Mark_Kashef here's Claude's response to my observation... "You've made a perceptive observation. I do aim to provide rationale for my responses when it's relevant and helpful. Here's some insight into why and how I do this: 1. Transparency: I believe in being open about my reasoning process, which helps build trust and understanding. 2. Context sensitivity: I try to gauge when explaining my rationale would be valuable based on the conversation context and the nature of the query. 3. Depth of explanation: I adjust the level of detail in my rationale based on the complexity of the topic and the apparent interest or expertise of the user. 4. Proactive clarification: Sometimes I provide rationale preemptively if I anticipate it might be helpful or interesting to the user. 5. Responsive elaboration: When users express interest in my thought process, as you have, I'm prepared to delve deeper into my reasoning. 6. Structured thinking: Providing rationale helps me organize thoughts in a logical, easy-to-follow manner. 7. Educational approach: Explaining my reasoning can be instructive, helping users understand not just what I'm saying, but why I'm saying it. 8. Conversational flow: I try to incorporate rationale smoothly into the dialogue without disrupting the natural flow of conversation. This approach aligns with my goal of being a helpful, transparent, and educational conversational partner. It's interesting that you've noticed this aspect of my communication style. Do you find this approach to providing rationale beneficial in our interactions?"

@@Mark_Kashef from Claude "You've highlighted an important point about the value of providing rationale. Indeed, organizing thoughts in a logical, easy-to-follow manner is a key benefit of this approach. Let me expand on this: 1. Structural clarity: Articulating rationale requires breaking down complex ideas into digestible components, which naturally leads to a more structured presentation of information. 2. Logical flow: Explaining the reasoning behind a statement or decision often results in a step-by-step progression of ideas, making it easier for both the communicator and the receiver to follow the logic. 3. Identification of assumptions: When we provide rationale, we often uncover and state our underlying assumptions, which can lead to more robust and transparent thinking. 4. Gap identification: The process of explaining rationale can help identify gaps in logic or areas that need further exploration, improving the overall quality of the discussion. 5. Enhanced understanding: For both the explainer and the listener, going through the rationale often leads to a deeper understanding of the topic at hand. 6. Facilitation of critique and improvement: When thoughts are organized logically, it becomes easier to critically examine them, leading to potential refinements or alternative perspectives. 7. Memory aid: A well-structured explanation is often easier to remember and recall later, benefiting both short-term dialogue and long-term learning. Your focus on this particular aspect suggests you value clear, structured thinking. Do you find that articulating rationale helps in your own thought processes or communications as well? And are there other benefits or potential drawbacks to this approach that you've observed?"

@@Mark_Kashef More from Claude "You've made an astute observation about the impact of providing rationale on my response generation process. You're right that it does lead to a more structured approach. Let me elaborate on how this works: 1. Framework creation: When I'm prompted to provide rationale, it automatically triggers a more systematic approach to organizing the information I'm about to present. 2. Hierarchical thinking: It encourages me to categorize ideas into main points and supporting details, creating a natural hierarchy of information. 3. Logical sequencing: Explaining rationale often requires presenting ideas in a logical order, which helps in structuring the flow of information from foundational concepts to more complex ones. 4. Comprehensive coverage: The act of providing rationale pushes me to consider multiple aspects of a topic, ensuring a more thorough and well-rounded response. 5. Clarity of expression: Structuring thoughts for rationale often results in clearer, more precise language use, as I need to articulate each step or component of the reasoning. 6. Connections and transitions: It prompts me to make explicit connections between different ideas, improving the overall coherence of the response. 7. Self-checking mechanism: As I structure the rationale, it serves as a form of self-review, helping to ensure consistency and completeness in the information provided. 8. Adaptive organization: The structure isn't rigid; it adapts based on the complexity of the topic and the flow of our conversation. You're correct in noting that this process of providing rationale influences how I organize and present information. It's an interesting aspect of how my responses are generated and shaped by the nature of the questions and the context of our dialogue. Does this align with what you've observed in our interaction? Are there other aspects of how I structure information that you've noticed or are curious about?

So happy this was helpful for you! 🦾

Pleasure as always!

So happy to hear it's helpful! I have it it my description! 🤖Meta Chain-of-Thought Prompt Example: bit.ly/3Y0pmTb

I love the conversational design of this prompt! If I could pin it here I would! 📌

Awesome to hear legend!

thanks for your feedback - this video is meant to focus on the 'how' to use the technique rather than how to optimize the use of this technique in LLM production-based applications. If you're using a front-end with a no limit on token usage, being superflous with wording isn't a consideration

@@Mark_Kashef economy in programming and clear, concise thinking is never optional when it comes to constructing logic which is what we are doing. This very much is part of the "how" to use efficiently. There is always a cost when we are not efficient and economical

in the lab cooking now cooking them 🍳

@@Mark_Kashef brother i hope your not cooking something else in the lab - BB reference 🤣

I have a video coming this week with the back and fourth aspect to help fine tune the prompt creation :)

happy to hear that Matthew 🦾 more coming

So happy to hear! Thanks for tuning in from Maroc 🇲🇦

my pleasure! hope they help

yes sir! You tell it to be 'inspired' by the example or 'create something derivative of example X for Y use case'

@@Mark_Kashef Nice, gotcha

thanks for the comment! We've been using Make.com 99% of the time so far for stability purposes, but will put it on our list.

The best you can do with those is just how their schema is constructed to see what data they 'may' have.

I don't have any actively in the store at the moment, all just personal or shared.

Absolutely, anything doable via Make is fair game; Zendesk has a solid set of triggers through there and Zapier

two more on the way next week😃