iphone photo. a python carrying a torch, riding a llama. --ar 5:3 --v 6.0 --style raw

Tech and Development News: Week Ending 04/12/2024

April 12, 2024

iphone photo. a python carrying a torch, riding a llama. –ar 5:3 –v 6.0 –style raw (that didn’t work)

“absolutely incredible solution to the @VictorTaelin challenge (by anime pfp anon ofc) look at this peak prompt engineering llms let you construct entire programming languages from thin air limited only by your imagination https://twitter.com/swyx/status/1778285624006778924

“My favorite part about @honicky’s Paper Club session this week on the 1-bit LLMs paper – relating it to @jefrankle’s Beyond Chinchilla laws and adjusting the equations for the memory/latency characteristics of 1-bit LLMs to derive an optimal param count/data size to aim for. no https://twitter.com/swyx/status/1778941946444279809

“”Through multiple controlled datasets, we establish that language models can and only can store 2 bits of knowledge per parameter, even when quantized to int8, and such knowledge can be flexibly extracted for downstream applications. Consequently, a 7B model can store 14B bits of https://twitter.com/rohanpaul_ai/status/1777638750740210175

“DSPy is a methodology, programming model, and set of optimizers for building Language Programs: arbitrary control flows that call LMs multiple times in a system. It can optimize the prompts & weights (incl. w/ PEFT) of the LM calls to maximize program quality on a given metric🧵” / X – https://twitter.com/lateinteraction/status/1777731981884915790

“Pretraining LLMs is expensive, but scaling laws allow us to accurately predict the performance of larger training runs from much cheaper ones. Recent research has drastically improved our practical understanding of scaling laws for LLMs… “We fit scaling laws that extrapolate https://twitter.com/cwolferesearch/status/1777424149415145882

“AI will improve, but will it ever replace a professional programmer? Current evidence says no. 1. AI can turn non-coders into average programmers 2. AI can help average programmers become better 3. AI can’t help expert programmers much Here is the evidence I have in mind:” / X – https://twitter.com/svpino/status/1777430219785130067

“Have you ever wanted to train LLMs in pure C without 245MB of PyTorch and 107MB of cPython? No? Well now you can! With llm.c: https://twitter.com/karpathy/status/1777427944971083809

[2404.05405] Physics of Language Models: Part 3.3, Knowledge Capacity Scaling Laws – https://arxiv.org/abs/2404.05405

“Instead of doing naive text splitting, extract a document knowledge graph to power your advanced RAG pipeline 💡 This tutorial by Fanghua Yu shows you a novel use case of LlamaParse – extract it into structured markdown that you can then convert into a document graph that you https://twitter.com/llama_index/status/1777348428755820849

“An agentic extension for RAG is to treat documents as tools and agents instead of just text chunks – this allows you to dynamically interact with these documents beyond getting back a fixed list of chunks. This is a great blog post by @andysingal and diagram by @clusteredbytes https://twitter.com/jerryjliu0/status/1776971813874028694

“Building Multi-Document Agents with @llama_index RAG with simple questions over a small set of data is easy, but a key goal for @llama_index is to solve complex QA over many docs. @andysingal presents an excellent overview of our multi-document agents. Instead of treating each https://twitter.com/llama_index/status/1776627066126901311

DataStax Acquires Langflow to Accelerate Making AI Awesome | DataStax – https://www.datastax.com/blog/datastax-acquires-langflow-to-accelerate-generative-ai-app-development

“AWS just released new GPU Instances! 🤯 @awscloud EC2 G6 with NVIDIA L4 GPUs (24GB) of memory with up to 8 GPUs (192 GB) on a single instance. Starting at $0.805 and are cheaper 25% cheaper than G5 with NVIDIA A10G. 🤑 Available in AWS US East (N. Virginia and Ohio) and US https://twitter.com/_philschmid/status/1776172921230123178

“New Text-to-SQL dataset from @gretel_ai! 🤯 retelai/synthetic_text_to_sql is a high-quality synthetic Text-to-SQL dataset released under Apache 2.0. 😍 👉 https://twitter.com/_philschmid/status/1776154264944931014

“Visualization-of-Thought Elicits Spatial Reasoning in LLMs Inspired by a human cognitive capacity to imagine unseen worlds, this new work proposes Visualization-of-Thought (VoT) prompting to elicit spatial reasoning in LLMs. VoT enables LLMs to “visualize” their reasoning https://twitter.com/omarsar0/status/1776082343813403063

“LLMs as Compilers This work proposes a think-and-execute framework to decompose the reasoning process in language models. This helps to improve algorithmic reasoning in LLMs. – It first THINKS to discover a task-level logic to solve a task and express logic in pseudocode – It https://twitter.com/omarsar0/status/1776248188707430719

“RAG pipelines have a lot of hyperparameters, and every new use case requires tuning these parameters for optimal performance. We’re excited to present AutoRAG (by Marker-Inc-Korea)🔥- given an evaluation dataset, AutoRAG will automatically find and optimize RAG pipelines for https://twitter.com/llama_index/status/1776289203459858849

“LangChain 🪢 Weaviate @weaviate_io’s open source vectorstore has features ranging from native multi-tenancy to advanced filtering, and it’s now accessible through a standalone integration package: `langchain-weaviate`! Its wide range of features even power the retrieval behind https://twitter.com/LangChainAI/status/1776301091375948244

“ReFT: Representation Finetuning for Language Models 10x-50x more parameter-efficient than prior state-of-the-art parameter-efficient fine-tuning methods repo: https://twitter.com/arankomatsuzaki/status/1776057023697731913

“Introducing Rerank 3! Our latest model focused on powering much more complex and accurate search. It’s the fastest, cheapest, and highest performance reranker that exists. We’re really excited to see how this model influences RAG applications and search stacks. https://twitter.com/aidangomez/status/1778416325628424339

“Introducing Rerank 3: our newest foundation model purpose built to enhance enterprise search and Retrieval Augmented Generation (RAG) systems, enabling accurate retrieval of multi-aspect and semi-structured data in 100+ languages. https://twitter.com/cohere/status/1778417650432971225

“# explaining llm.c in layman terms Training Large Language Models (LLMs), like ChatGPT, involves a large amount of code and complexity. For example, a typical LLM training project might use the PyTorch deep learning library. PyTorch is quite complex because it implements a very” / X – https://twitter.com/karpathy/status/1778153659106533806

“Adapting LLaMA Decoder to Vision Transformer https://twitter.com/arankomatsuzaki/status/1778237179740688845

“Google announces Leave No Context Behind Efficient Infinite Context Transformers with Infini-attention This work introduces an efficient method to scale Transformer-based Large Language Models (LLMs) to infinitely long inputs with bounded memory and computation. A key https://twitter.com/_akhaliq/status/1778234586599727285

Groq CEO: ‘We No Longer Sell Hardware’ – EE Times – https://www.eetimes.com/groq-ceo-we-no-longer-sell-hardware/

Heads up! You’ve scrolled to the end of this category. There may have been just one or two links (above), so go back up and double check to be sure you didn’t quickly scroll down past it.

Be Sure To Read This Week’s Main Post:

This week’s executive overview and top links are here:

AI News #28: Week Ending 04/12/2024 with Executive Summary and Top 48 Links

The post you just read is an deep dive extension of my weekly newsletter, This Week In AI, an executive summary of the top things to know in AI. Each week, I create an accessible overview for laypeople to feel confident they are conversant with the week’s AI developments. I include a curated list of must-click links of the week, to offer everyone a hands-on opportunity to explore the most intriguing updates in artificial intelligence across various categories, including robotics, imagery, video, AR/VR, science, ethics, and more. Beyond the overview, I post these topic-based deeper dives (below). If you haven’t read this week’s overview, I recommend starting there.

Credits/Sources

Most of these weekly links come from just a few prolific oversharing sources. Please follow them, as they work hard to find the news each week and they make it a lot easier for me to compile.