- 728
- 944 127
LLMs Explained - Aggregate Intellect - AI.SCIENCE
Canada
Приєднався 24 жов 2008
AISC is a community of intellectually curious individuals centred around technical review and discussion of advances in machine learning, including but not limited to Large Language Modesl.
Welcoming, supportive community of machine learning practitioner and researchers, coming from industry and academia, with an array of experience, from avid learners to experts in their fields. Our group is centred around consistent meetings focused on meticulous but friendly discussions of advances in machine learning, covering both intuitive understanding and foundational technical details.
Welcoming, supportive community of machine learning practitioner and researchers, coming from industry and academia, with an array of experience, from avid learners to experts in their fields. Our group is centred around consistent meetings focused on meticulous but friendly discussions of advances in machine learning, covering both intuitive understanding and foundational technical details.
Human Feedback Foundation - LLMs
Check out my essays: aisc.substack.com/
OR book me to talk: calendly.com/amirfzpr
OR subscribe to our event calendar: lu.ma/aisc-llm-school
OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems
OR book me to talk: calendly.com/amirfzpr
OR subscribe to our event calendar: lu.ma/aisc-llm-school
OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems
Переглядів: 86
Відео
LLM Products vs Traditional Digital Products
Переглядів 13114 днів тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems
LLM as Personal Financial Assistant
Переглядів 16028 днів тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems
LLM Products for Regulated Industries
Переглядів 116Місяць тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems
LLMs and Business Workflows
Переглядів 862 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems
5 Commandments of Building LLM Products
Переглядів 952 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems
Building a LLM Testing API
Переглядів 2842 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems ⃝ Challenges of testing Conversational AI systems: 🟢 There's no single agreed-upon approach for unit testing or regression testing in the world of chatbots. 🟢 Traditional metrics (accu...
LLMs - Chunking Strategies and Chunking Refinement
Переглядів 2642 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems 🟢 Chunking for precision, not just recall: The goal of chunking is to provide the LLM with the most relevant information possible, not just all of the information. This means that the ...
Large Language Models as a Building Blocks
Переглядів 3292 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems 🟢 Search systems can be improved by using language models to understand the meaning of queries and documents, rather than just matching keywords. 🟢 Semantic search can be broken down i...
Competitive Advantage for Startups in era of LLMs
Переглядів 1972 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems 🟢 Moats are difficult to build and maintain in the era of LLMs. Because LLMs are constantly improving and becoming more accessible, it's difficult for startups to build a long-term com...
Intersection Between LLMs and Products
Переглядів 1372 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems There are three important mindsets to consider when building LLMs (Large Language Models) products: rooting the problem, acting as a user, and having a growth mindset. When applying th...
What is the right team composition in era of LLMs?
Переглядів 662 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school OR sign up for our LLM course: maven.com/aggregate-intellect/llm-systems AF: It's a good two, three decades we've been talking about agile teams and in reality, they haven't quite happened in a lot of organizations. So given the fact that things are moving ...
Building an LLM Teacher-bot
Переглядів 1622 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school 🟢 The focus should be on user needs instead of technical feasibility. The user in this case is the teacher who is burnt out due to the workload. 🟢 Start by validating the desirability of the product. There are frameworks like Jobs to be Done that can help d...
What is the relationship between LLMs and multi-modality?
Переглядів 932 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school AF: One of the interesting strategies that Cohere is using versus other providers where you're specializing a bunch of different models for different types of things: you have a re-ranker, you have a few other things that you mentioned. In a lot of real wor...
What are the system level considerations for using LLMs?
Переглядів 683 місяці тому
Check out my essays: aisc.substack.com/ OR book me to talk: calendly.com/amirfzpr OR subscribe to our event calendar: lu.ma/aisc-llm-school AF: LLMs are not sitting in the vacuum. They're interfacing with your databases. They're interfacing your retrievals, probably some biological models, physical models. How do you go about designing a system around LLMs to make sure it is robust, make sure i...
What is the relationship between language and intelligence?
Переглядів 633 місяці тому
What is the relationship between language and intelligence?
How do you improve your RAG pipeline?
Переглядів 1273 місяці тому
How do you improve your RAG pipeline?
Are long context LLMs the death of RAG?
Переглядів 3463 місяці тому
Are long context LLMs the death of RAG?
How Do You choose between training, fine-tuning, and using small models?
Переглядів 1383 місяці тому
How Do You choose between training, fine-tuning, and using small models?
LLM Evaluation, Validation, and Verification
Переглядів 2333 місяці тому
LLM Evaluation, Validation, and Verification
How Do You Validate LLM Systems Beyond Benchmarks?
Переглядів 1033 місяці тому
How Do You Validate LLM Systems Beyond Benchmarks?
Can Sherpa (multi-agent llm) Handle Multi-modality?
Переглядів 943 місяці тому
Can Sherpa (multi-agent llm) Handle Multi-modality?
What Kind of Risks Are Specific to LLMs?
Переглядів 673 місяці тому
What Kind of Risks Are Specific to LLMs?
LLMs, What Skills to Learn? and What a Time to be Alive!
Переглядів 4823 місяці тому
LLMs, What Skills to Learn? and What a Time to be Alive!
How do you Force an LLM to Keep Track of the Assumptions a Document Makes?
Переглядів 1013 місяці тому
How do you Force an LLM to Keep Track of the Assumptions a Document Makes?
How to Annotate Data for LLM Applications
Переглядів 4033 місяці тому
How to Annotate Data for LLM Applications
What is the Role of Data Quality and Diversity in LLM Systems?
Переглядів 1353 місяці тому
What is the Role of Data Quality and Diversity in LLM Systems?
Testing Strategies for LLMs - SHERPA - Open Source Project Update, 2023-12-08
Переглядів 3906 місяців тому
Testing Strategies for LLMs - SHERPA - Open Source Project Update, 2023-12-08
Evaluating Job Exposure to Large Language Models
Переглядів 1816 місяців тому
Evaluating Job Exposure to Large Language Models
Exited!!
I totally Agree. There is no Free lunch. OpenAI, Google and others are benefiting from giving free access to their models They get feedback from millions of people they use to improve their proprietary models more and more.
Why am i watching 4 scientists complaining about how they can’t find similar research groups and yet you all have been trained to take the human out of you. You guys talk about connections but none of you will take the time to establish an equally respectful connection with let’s say someone me and discuss in layman’s terms just what you said in this podcast in about an hour. What’s your point. You’re also bottom tier scientist and the likes of you can actually save science. I know you guys want these big jobs but remember this science was invented to answer humankind’s questions not do all that hard work to sit here with a bunch of a holes that mainstream science doesn’t even accept. Just do your job and educate, don’t worry about your lucrative jobs. Also you have all the right to have a lucrative job end goal, but please don’t lie to people. Hope you read that a few times before you get angry and respond because I feel your authentic existence is the only way you can be relevant and not “et all.” Grow some relatable personalities while you’re at it. Your lack of relatability to daily emotional life and your fundamental troubling belief that these are separate individual phases in actual true science. Just like context is important in a headline claiming to report the truth but at the same time prompts the audience when to clap for the recorded episode. When the fuck did you guys lose any practical logical and scientific thought process? Was it when views on UA-cam became more important? What a waste of your parents dollars to go these prestigious universities, wish I had even one semester with privileged even AP classes you guys were getting in public high schools even if I had better grades. I was targeted for military recruitment. Hope you guys see what I’m saying. Willing to discuss this openly.
There’s no room for courtesy in science. You guys self imploded because neutrality doesn’t advance science
Link to the slides docs.google.com/presentation/d/1VRDLt78bkYArC7y6WBnoxis79Y6gOrquFnRUYPkchYk/edit?usp=sharing
based liam
Can you upload source code to GitHub ?
ua-cam.com/users/redirect?event=video_description&redir_token=QUFFLUhqbUlQb05zTWszT1VHWF94aEt4aTBEX0FCYld1d3xBQ3Jtc0ttN29JcEp1Wmo4dlprTnZHdlNxdGNLb1M3dFVhU1FxUDhuSGNNM0Y0el91aHpJRV94cHotdENRV0otWmNPeGFuYUhmdDNkV0ZpU3dBR1ZSWlQ4MTdHS0NKUDdvOWllZnhxWWt2VVRiOXdQQ0FKS2NCTQ&q=https%3A%2F%2Faisc.ai.science%2Fevents%2F2019-08-12&v=jio04YvgraU ---> link for slides doesn't exist or has been moved.
Cut the Bull shit short next time
Excellent Lecture/key note
looking forward to LLMMs (large language middle managers)
Rich compilation of wisdom nuggets. Great channel.
Really nice I have been searching around no such updated content on this topic please looking forward
Thanks for the Review :) 1:36 One Question! If we were to talk about "Generalization Gap", shouln't Train Score be higher for LBs?
Share the repository dadash!
44:14
44:35
49:07
No cybersecurity...?
source code and dataset please
Could you please share the ppt? The provided link is no longer valid.
It's sad that he kinda dodged to tell the full answer (or at least his speculative answer) to the question at 20:45, I'd like to hear more on this.
Looking forward to it Amir!
cheers Arash
maven.com/forms/30a683
This was very insightful. Thank you !!!
Can anyone please provide the notebook for this talk?
Garbage in garbage out
Thank you
Thanks very nice 🌹, How to use open source LLM from hugging face locally using llama.cpp or ctransformer library with PandasAi without internet connection offline mode? for example codellama or pandalytic gguf type
This is an incredible series! Thanks for sharing, I learned a lot :)
Great lecture! Do you know what the 3D Scharr filter would be for 3x3x3?
0:00: 📚 The presentation discusses the concept of creating an oracle that collaborates with humans, emphasizing the need for close collaboration between humans and machines in knowledge-intensive workflows. 4:17: 🤖 The video discusses the evolution of AI, from descriptive analytics to the emergence of narrow AI and the potential of large language models in crossing the narrow AI chasm. 8:03: 🤖 The video discusses the development of language models and their potential to enable machines to perform complex tasks. 11:41: 🤖 A framework called Sher AI is being developed to augment knowledge-intensive workflows using large language models and data connectors. 15:22: 🤖 The video discusses a multi-agent system project aimed at creating a thinking companion to assist humans in problem-solving and decision-making. 18:44: 📝 The video discusses opportunities for contributing to a project, current tasks, and future plans. 22:15: 🔧 Setting up and using Shera for virtual environment and database integration. Recap by Tammy AI
Will the notebook link be provided?
Thanks for sharing!
Grear video, Thank you so much. But if you can please explain or provide the streamlit code pleaaase
Can I get the slides of this lecture? Thanks a lot
'promo sm'
very good presentation. very interactive session. thanks
@35:01 A question: How do we know that p(x|z) is normally distributed? I can see that p(z|x) must be normal, but not p(x|z). I am asking this question because I see it in a lot of VAE related talks, stated without any proof or justification.
where can i find this slide
Great intro. thanks!!
I am really excited by your project. Thanks for sharing. I wish the best from Morocco
I know it’s “old” now (by today’s standards!) but I really enjoyed this conversation. Thank you for posting it 🙏
could you please share the slides
There are two mistakes on slide at 40:11: 1. u is content bias, not position bias. Vector u is firstly appears in term (c) from the paper. And term (c) governs a global content bias according to aurhors' words. 1. v is position bias, not content bias. By the analogical reason. It appears in term (d) from the paper. According to paper authors, term (d) encodes a global positional bias.
I don't understand why "v" is content bias and why "u" is positioal bias. Vector "v" is multiplier of R[i-j] so it's more similar to positional bias, because it's like linear function on relative positon R. And "u" is content bias because it's like linear function on k.
"41:56" that guy can understand it perfectly and use his mobile too.😅
hey thanks for the great intro but i cant find any slides in provided link or on your website..can u provide available links for slides?
thanks
how to apply this architecture to predict deformation of solid structures under static load ? this problem has long spatial dependencies which are difficult to address with local edges in the graph.
Asc
Good presentations. Thanks Gordon Can I have your twt profile, I would love to get connected with smart minds
windows install failed: ERROR: The tar file (C:\Users\xx\AppData\Local\Temp\pip-unpack-plha6gd0\tf-encrypted-0.9.1.tar.gz) has a file (C:\Users\xx\AppData\Local\Temp\pip-install-bj8xrdtv\tf-encrypted_08c1f7c5ab8c46a19975b7aaec907caf\tf_encrypted/operations/aux) trying to install outside target directory (C:\Users\xx\AppData\Local\Temp\pip-install-bj8xrdtv\tf-encrypted_08c1f7c5ab8c46a19975b7aaec907caf)