Prompting Your AI Agents Just Got 5X Easier...
HTML-код
- Опубликовано: 28 май 2024
- If you want to master Prompt Engineering, join my community: www.skool.com/new-society
Follow me on Twitter - x.com/DavidOndrej1
Please Subscribe.
Anthropic Console: console.anthropic.com/
Anthropic Cookbook: github.com/anthropics/anthrop...
Anthropic just released a new Claude 3 prompt generator that makes prompt engineering easier than ever before, such as with AI Agents.
If you want to master Prompt Engineering, join my community: www.skool.com/new-society
Dont need to now lol
David did you read the paper with the 26 Principles of prompt engineering? On Arxiv its called 'Principled Instructions Are All You Need for Questioning LLaMA-1/2, GPT-3.5/'
Now the interesting thing here is they measured the before and after effect....
I didn't like that anthropic workshop reqd phone number verification, and has no 'delete my account' function. So no eu users then? Violates GDPR if I am not mistaken?
@@_0xle0_ no they don't have to provide a button for 'delete all my data. Its more that if there is a request, to the data controller, even by email for example, then they have to delete it, or explain why they can't or that they don't have it (isn't recorded). The GDPR does force a lot of complications on tech companies, but autodelete of the data they have on you isn't one of them I'm afraid!
@@_0xle0_ Also here is my solution for what its worth to the whole 'give us your email and phone number and personal details'
1. I lie, blatantly, unless I have to use a credit card or bank card (In which case I'll have scrutinized the hell out of them.
2. Since I build a fair number of mobile apps I have five phones and three tablets that I consider essentially burner phones' for dev. really cheap phone is €80 in Ireland.
Great video David! The new Anthropic prompt generator looks like an incredibly useful tool for prompt engineering. It could be a big time saver, especially for those just getting started with building AI agents. Thanks for the detailed walkthrough and sharing your insights. Looking forward to trying it out!
Great content. Thank you
They have had this in a free colab in their repo for a while - good that they added it to the site - i use it constantly
Is the free tool stil live?
Great guide
You and David Shapiro are the best AI RUclipsrs out there.
matthew and david is a duo i never thought i wanted but would be sick
Same ikr! ✌🏻
Thanks!
I’m not seeing how to get to the workbench and when I ask Claude Claude denies that one exists. What do I need to do? Where can I go to find it? What you’re working on?
That's so funny, I immediately checked your channel after seeing matt's vid to see your thoughts. But I think he might be mistaken, OpenAI may have been talking about their own gpus not everyone else's. Would love to hear your interpretation!
Awesome!
would love to see that one. Mathew Burman if you see this please come on his show. You guys are awesome
Hey there. Thank you for these videos. I'm wondering since I know you're in EU as well, how you charge your Anthropic account? My credit cards get refused all the time.
Anthropic made a prompt generator so you can give it your short prompt and it will turn it into a masterpiece. My guy went ahead and wrote a full blog-post worthy prompt for the prompt generator 😀
I just wish this wasn't dependent on the most expensive LLM out there. I'm interested in tools that work within private AI, totally walled off from dependencies on token costs, and APIs to closed source AI projects controlled by BigTech companies chock full of MBAs thinking day and night on ways to increase dependency on users, to eventually leverage them financially. No thanks.
i don't see this option. Did they remove it ?.
I think they did cuz I don't see the option either smh
Yeah, the HUGE problem with Claude, though, is that no one can even sign up to it Lol ,even with the phone number, email and everything. I read about it on reddit and it seems to be the same problem using a non-gmail email address (i.e. "verification code failed"). They only seem to want to accept gmail addresses for sign up now.
I had trouble in vivaldi, I used my naked chrome with no plugins, I was able to sign up just a few mins ago. Hope tihs helps.
How is this any different from Custom GPT?
It's nice to see Anthropic making their models more useful for specific workloads.
This is insane
Do a podcast with David Andre.
😂😂😂😂😂😂😂😂
Obviously, He meant do a podcast with Matthew Berman.
Good listening ears though.
Sure, do the podcast
Matthew Berman would be great to see on David Ondrej's podcast!
мне очень нравится твои видео я подписан давно! до этого момента было круто но на этом видео немного смешно ))) так как это уже решено давным давно )
Is this not DSPy?
Prompting Your AI Agents Just Got 5X Easier...
00:00 Новая функция от Anthropic
• Компания Anthropic выпустила новую функцию, которая может навсегда изменить разработку Prompt Engineering.
• Функция позволяет создавать расширенные подсказки с использованием новейших принципов разработки Prompt.
01:48 Примеры использования функции
• Примеры использования функции включают написание электронного письма, подготовку ответа на электронное письмо клиента, классификацию стенограммы чата и перевод кода.
03:39 Создание собственной подсказки
• Создание собственной подсказки включает описание задачи, предоставление контекста и использование переменных для упрощения процесса.
• Пример использования функции для создания краткого изложения стенограммы еженедельного общественного собрания.
10:43 Использование подсказки в Anthropic Console
• В видео демонстрируется использование подсказки в Anthropic Console для создания и редактирования системных подсказок.
• Обсуждаются различные варианты подсказок и их использование для создания более информативных и полезных сообщений.
14:37 Примеры использования подсказки
• В видео приводятся примеры использования подсказки для создания системных подсказок, которые могут быть использованы в различных ситуациях.
• Обсуждаются преимущества использования подсказки для создания более информативных и полезных сообщений.
17:09 Использование подсказки для разработки программ
• В видео подчеркивается, что использование подсказки может помочь разработчикам начать работу над проектами и сэкономить время на решении проблем.
• Обсуждаются различные аспекты использования подсказки для разработки программ, включая настройку, управление пакетами и понимание нюансов различных платформ и фреймворков.
Did you take the subtitles and asked an AI to generate a summary ?
nose strip ? hey you forgot to write GPT on it ! 😂
Interesting
So it's basically a compiler for prompts ?
Seriously though, wouldn't it make more sense to update the AI to understand the shorter prompts directly ?
Hell yeah, you and Matt Berman are the two people I watch. lol
Not 100% sure it's there yet, at least for more advanced usages. Not seeing use of expert role, emotionality or additional context which are all proven prompt techniques. Not bad though if you're a beginner and want some decent prompts to get you going.
Do a podcast with david ondra
even the prompt engineer just lost his job 🤣
Heres what I don't get. IF Anthropic are sure this prompting philosophy works then why have the prompt generator at all, just have a switch that allows the model to itself change the prompt using the cookbook before it starts the task. You could simply control whether this is turned on or off. Do they have that? A sort of 'magic prompt' button? Ideogram and many image models have it and all of the proprietary models do it anyway in their guardrails!
this will reduce the flexibility, these prompts should be subject to editing.
If they get really positive responses to this feature, they will probably add the feature you suggest. But they wouldn’t want to add that feature until they’ve confirmed that users actually find the generated prompts useful.
I’ve been writing Prompt generating prompts since day one. Seemed obvious to me. Of course, you still need to be good at writing prompts, but it makes it a lot easier
They call them foundational models for a reason. The are as generalized as possible
I did find a sort of API thing they have going and a colab demo but its not optimising the prompt just making one up.
DO A BROADCAST WITH HIM, EVEN ME WHILE TRAINING THE NEW LEADERSHIP AGENTIC MODEL WANT TO SEE THAT BROADCAST, IT'S OBVIOUS THAT'S OPENAI LIES ALL THE TIME, IT'S A SIN
David and matthew please make it happen❤
How many days your nose is broken?
Whats that in your nose?
Do a podcast with David Andre.