Awesome ❤

Thank you Mads for this greate update

Local Semantic Index... now THAT is awesome!

Great content as usual Ayça Baş!

I found the repo with the samples - it lives here: github.com/github/codespaces-models

Create Azure LLMops Video from scratch , first create a project and go further and finally LLMops step , also give us the code

Merci, c'est très complet comme explication !!

where I can find github of this

Is it possible to connect a data source like a blob container to document intelligence instead of only a single file like all the demos seem to show?

Great coverage and content guys!

Konkore police

I found this really informative thanks

Starting at 7:11

Lu Zhang is awesome!

Very Good Explaintion

can you make your own drone? If so, what hardware would it need in order for facial recognition.

Its free ?

how can i query the custom trained model with the api / sdk i.e. with curl? If I do this, I only get results from the default model, but not from my custom trained model

Start at 5:58 🙂

how to get unmanaged endpoint list by api? I am able to get only managed endpoint

Thank you for this video! :)

I am new to NLP. So, this presentation is very useful presentation and timely info. Thanks for sharing and your efforts at Global AI Community.

Who are these videos for? If you don't let me use your products, you let me test them and that's it. That doesn't sound normal to me, that's why you have many thousands of subscribers and no comments, because you're a one-gate operation.

Make the appearance of the chat beautiful, because as always with Microsoft inside everything is good, but outside file and sledgehammer, someone does. Visual part is 50% of success.

🎯 Key Takeaways for quick navigation: [00:17] 🤖 *Introduction to building a web-connected search assistant* - Overview of creating a "poor man's Bing chat" with GPT - Walk through reasoning process and considerations - Goal is retrieving recent real-world information [08:50] 🧠 *How LLMs work - predicting next words from context* - Models trained on huge datasets to predict next tokens - Fine-tuning and reinforcement learning improves quality - But doesn't provide real-time up-to-date information [16:13] 📚 *Large context window sizes enable more understanding * - GPT-3.5 Turbo increased from 4K tokens to 128K in one year - Allows summarizing longer texts like books and code - But still not enough to fit all of internet's data [18:58] 🤝 *Retrieve-Augment-Generate (RAG) architecture* - Retrieval finds relevant info, augmentation prepares it - Clean, relevant docs passed to LLM for generation - Performance depends heavily on retrieval quality [25:41] 🚚 *Build a chatbot with GPT + search engine results* - Optimize user query for search engine input - Get top search result, process to extract key content - Dialogue by appending messages to shared context [29:59] ☑️ *Capabilities demonstrated:* - Answering questions with latest info - Rendering results in custom formats (tables) - Basic manipulation of time and dates Made with HARPA AI

🎯 Key Takeaways for quick navigation: 03:22 🤖 *Generative AI models can complete text by predicting the next most likely word word by word.* 23:08 🛡️ *Generative AI models are versatile and can perform multiple tasks like content generation, summarization, code generation, and more.* 24:13 🧠 *Foundation models serve as building blocks for various AI tasks, and their usage goes beyond their initial training tasks.* 30:03 🏭 *Generative AI has the potential to create new jobs, particularly those requiring AI knowledge across different industries.* 36:44 🌐 *The future of generative AI holds numerous creative possibilities, with applications expanding into various domains.* 55:12 🧩 *Language models take in a sequence of tokens as input and generate one token as output. Tokens can be whole words or parts of words, depending on their length.* 57:57 🔄 *Language models work in a sliding window manner, taking in n tokens as input and producing one token as output. This process repeats until a stopping condition is reached.* 59:19 🔄 *Language models are not deterministic; they generate a probability distribution over possible tokens, and sampling from this distribution can yield different results for the same input.* 01:02:41 🧩 *Transformers, introduced in 2017, form the basis of many language models, including GPT models. They utilize the attention mechanism to pay more attention to specific input tokens.* 01:04:06 📚 *GPT-3.5, ChatGPT, and GPT-4.5 are different versions of GPT models, with variations in base models, fine-tuning, and reinforcement learning with human feedback. Choosing the right model depends on specific use cases.* 01:06:06 🧠 *Generative AI, like GPT models, has exciting potential, and not everyone fully understands its capabilities yet.* 01:07:03 💬 *The speaker has been working on a chatbot prototype using Microsoft semantic kernel in C# and is excited to introduce C# developers to large language models.* 01:11:02 🛠️ *The speaker discussed the architecture of using large language models in applications, emphasizing the importance of a kernel component for controlling AI models.* 01:15:28 🖥️ *To use Microsoft semantic kernel, you need to build a C# application, configure the kernel, and import skills, allowing you to harness the power of language models in your application.* 01:25:11 🧠 *Microsoft semantic kernel can be enhanced with memory, such as Azure Cognitive Search, to connect business documents to your AI conversation and provide grounded responses.* 01:30:16 🛠️ *Generative models can be used to augment AI product development by optimizing recipes or formulations, making the process more efficient and faster.* 01:38:11 🧪 *Generative models have applications in simulating events for scientific research, including physics phenomena and drug development.* 01:49:02 🥃 *Generative AI can be used to create unique products, like AI-generated whiskey, which can be well-received in the market.* 01:51:23 💼 *Generative AI can learn from small data or even no data, making it versatile for various industries, but compliance and testing are essential.* 01:56:13 📊 *Generative models can extend beyond optimizing recipes to develop formulations for drugs, chemicals, or materials. Human expertise remains crucial in this augmentation process.* 02:05:37 🖼️ *AI models like DALL-E are being used to generate synthetic visual content for various applications, reducing the need for physical samples and photo sessions in industries like fashion and e-commerce.* 02:07:09 🎮 *DALL-E and similar AI models are accelerating game development by generating NPC appearances and reducing the carbon footprint associated with game development.* 02:09:38 🚫 *Stopping AI research doesn't make sense; instead, focus should be on responsible AI use and addressing ethical concerns in areas like data scraping and copyright.* 02:19:45 🎓 *Educating people about AI, its capabilities, limitations, and ethical considerations is crucial, especially for various user levels, from students to professionals.* 02:26:40 🌍 *Climate change is a growing concern, and open-source AI models may play a significant role in addressing it by facilitating the development of more efficient models and solutions.* Made with HARPA AI

Thanks a lot Usman.

Hi peton. Can you re-do a short tutorial video with steps of 1,2,3 on how to work with your excel file downloaded from github? thanks

Why are you using Third-party OpenAi Api to run LLM after discussing in length that Rust is faster than python.

With no script this is just another waste time video. If I have a script: I don't watch this video. If I don't have the script: Why the f**k waste my time watching this video on how it works... but don't have the f*****g script?

I totally agree! I am very grateful for AI. It's fascinating how AI comprehends and encapsulates the entirety of my unique experiences like being in an intercultural relationship for a decade and living the digital nomad lifestyle for two years.

I love his point of view

It is the future. so those that learn to adapt first will be ahead

it's all about knowing how to use a tool well.

It's scary but very exciting ....................

This was very informative. Thanks Sammy and Farzad.

Amen to that! Let's be positive and never be afraid )) yeah!

Could you provide the script you used in this video?

This is great to see the history as to where we are now. This video helps in the next generation, considering I was around learning computers in the late 70's. And now working in this field and skipping all those university years. Well done.

thanks for the presentation. I am thinking of scenario where I want to implement Q&A dialog in a chat bot. How could I integrate SK completion dialog with memory with chat where single prompts generates single response ?

Nice presentation on SK.. it s a very valuable AI tech

Good video

Great video.

A combination of AI augmented methods and DoE ? Would try it.

Thanks so much, guys

Title and content are not really related at all!

I love the clip, thanks!!

Maybe the script would have been more helpful!!!

This is lovely. Keep growing & continue to be the AI (Always Inspiring) person that you are ❤️