Chat over your data with a personality 🕺 LlamaIndex 0.7.2 brings the ability to add a system prompt to our chat engines/OpenAI agents. Other updates: - Generate callback events for SubQuestionQueryEngine (@thesourabhd) - @Azure OpenAI support (@disiok) https://t.co/tntCOrbeFx

📣 Exciting news! Prodigy v1.12 is now available! 🚀 ✨ It includes OpenAI integration, prompt engineering workflows, custom task routing and more! 🧵 Check out this video from @fishnets88 for a quick tour of the new features 👀 https://t.co/JowsF14IYL https://t.co/2EfLq2Hm7Y

SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis paper page: https://t.co/Ztk9rYU1Gs present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet… https://t.co/HB1prIm6Lh https://t.co/0s0T8I18V4

What Matters in Training a GPT4-Style Language Model with Multimodal Inputs? paper page: https://t.co/9kgHFBQKC7 Recent advancements in Large Language Models (LLMs) such as GPT4 have displayed exceptional multi-modal capabilities in following open-ended instructions given… https://t.co/osp3ZPPj3c https://t.co/CcYWUVxong

Put ChatGPT at a cocktail party🥂. Can it - understand people's conversations, gestures - figure out their relations, - and even chime in with social advice? 🦍Announce KokoMind. 🌟Check out this demo! More at https://t.co/giyU7cyg3D #AI #GPT4 #ChatGPT #OpenAI #Shrinking 🧵 https://t.co/h6BOBDtL4F

Introducing Superalignment by Open AI blog: https://t.co/5nKrhGyHzO Our goal is to build a roughly human-level automated alignment researcher. We can then use vast amounts of compute to scale our efforts, and iteratively align superintelligence. To align the first automated… https://t.co/SStXKp5ZM6 https://t.co/RPCCeHDlRN

Nature paper: DNA can be used as a "living digital camera" to capture images! 96 pixels for now. Instead of synthesizing a new DNA for each image, which is costly & complex, this work (called "BacCam") records the light signal directly via site-specific DNA editing. Highlights:… https://t.co/j57haEug1W https://t.co/3vDyA2xvJq

Google Will Use Your Data to Train Their #AI According to Updated Privacy Policy 😕 https://t.co/188UH0LBEy https://t.co/1jqNH7pHvk

You can now try out 🐟GILL on HuggingFace Spaces! GILL is a model capable of conditioning on interleaved image + text inputs to generate image + text outputs. Try the demo now: https://t.co/VlWCIs0ifQ Project page: https://t.co/w6VgcntGl2 More cool examples below 🧵👇 https://t.co/R9QU0lswnW

Personality Traits in Large Language Models paper page: https://t.co/Jv1dh7VrlL The advent of large language models (LLMs) has revolutionized natural language processing, enabling the generation of coherent and contextually relevant text. As LLMs increasingly power… https://t.co/1sKxcGsj7W https://t.co/aHtiiwGbA7

Huh, it looks like you can use ChatGPT to bypass some paywalls 😲 https://t.co/9wnc8WsRkv

Evaluating complex, implicit human biases in #T2I generative models? Our #ACL2023NLP findings introduce Text-to-Image Association Test (T2IAT), a tool to quantify implicit stereotypes! Paper: https://t.co/VN7yeSJjaP Code: https://t.co/RRmn26qc6N Demo: https://t.co/4x90EKjgXj https://t.co/7hFVIiyyVZ

It feels like we're at a special moment in AI. In fact, progress has been gradual. The ELIZA chatbot is almost 60y old. When an AI tech (esp. general purpose & consumer facing) starts working sufficiently well, it's everywhere at once. Progress feels exponential. AGI feels close. https://t.co/ZpcywbbTu8

This is one of the fastest ways to build a custom ChatGPT-like system on top of your data. It's called ChatLLM (by @abacusai). Here is a demo of how to build a simple custom chat LLM: https://t.co/KBATOGuUIL

Peter Attia is the #1 longevity doctor. He charges $150,000 a year to be his patient. His 13-steps for a long healthy life: https://t.co/k5QRTbxgzh

Jeff Siskind has sent me email with the full account of the development of STALIN and it's every bit as epic as I remember. I would like you to contemplate the absolute beast-mode coding that we are talking about here. I do not joke when I say this should be in a computer museum. https://t.co/h6Z3mhN0Wj https://t.co/BU4qxfsg3l

+2 more

EmbedChain - a new framework to easily create LLM-powered bots over any dataset Building a chatbot with EmbedChain only requires you to: • create an app instance • add a dataset • query https://t.co/4Fz4lNrICN

Bing ChatGPT makes mistake, doubles down and then quits https://t.co/MjiytFhJto

Textbooks Are All You Need paper page: https://t.co/D0AEOSLfGM introduce phi-1, a new large language model for code, with significantly smaller size than competing models: phi-1 is a Transformer-based model with 1.3B parameters, trained for 4 days on 8 A100s, using a selection… https://t.co/zCP0YvcvuC https://t.co/OySgnXAk6v

Where is AI going? How could it transform life sciences and healthcare? Why would this happen now? Check out the @a16zBioHealth AI thesis for my take: https://t.co/FglnOD6jNn https://t.co/SsVWmP90Yg

Demystifying GPT Self-Repair for Code Generation We've seen a couple of papers showing the promise of self-repair in code generation. This paper finds that modest performance gains are seen when using GPT-4 for textual feedback. Another interesting finding: significant… https://t.co/kYLyeEkhJW https://t.co/C4ZIRF5QIu

Don’t hate the playa - when asked to name the top 30 AI startups currently, Claude (Anthropic) ranks itself as #1, 5, 8, 13, 14, 16, 18, 19, 21, 23 and 24. https://t.co/cFTYHLVKhg

Fascinating & seems reproducible! Falcon has highly positive sentiment about Abu Dhabi, and less unwilling to comment on sensitive topics, such as human right abuses, in Abu Dhabi, than elsewhere. Could have various causes, but it's an important reminder that open source-model… https://t.co/BS7CmkRjiL https://t.co/kWtUqU55fN

It’s quite impressive how rapidly language models have evolved. We have seen three main phases over the last few years: - Pre-trained LMs - Instruction-tuned LMs - RLHF LLMs Let’s take a look a closer look: https://t.co/xz8XwwPbt4

Apple's Object Capture now runs *on-device*! 💥 😳 This is a game changer as as any app can now integrate 3D scanning *for free*, and you can see where this is going at next WWDCs... https://t.co/UrHymmlJbH https://t.co/2f2lsQwxMZ

BOOM! The astonishing v3-13b-hermes-q5_1 LLM AI model is absolutely amazing. This is a local academic file of ~61,000 and it generated a summary that bests anything ChatGPT can do. AND THIS COMPUTER HAS NO INTERNET. Join us for FREE and own your own AI so it don’t own you. https://t.co/MbeANLRxuA https://t.co/muZKgTcQAb

Introducing humans on @flairAI_.🔥 Generate thousands of composable human models in various poses to sell your products! Target different demographics, and represent your audience in marketing campaigns. RT and comment to try this feature in beta. Serving access codes today:) https://t.co/EjTtVB0cJ6

🚨 Attention data scientists: We're very excited to announce that Giskard's new beta release is out, hooray! 🤩 What's new in this version? 👇 [1/8] https://t.co/xdGS0EGwMH

Excited to announce a big step toward organizing the internet! Metaphor now enables: 1. Neural search over the internet's newest content 2. Filter by any number of domains 3. Filter by date So you get all the power of Metaphor, but over a subset of the internet that you define. https://t.co/2GYVrPWdYL

Announcing RedPajama 7B trained on 1T tokens! 🚀 • Instruct, chat, base, and interim checkpoints on @huggingface • The instruct model outperforms all open 7B models on HELM benchmarks • The 5TB dataset has been used to train over 100 models Details👇 https://t.co/oUNKqYBmlS https://t.co/bBTYf51M28

Transformer Taxonomy (the last lit review) This list is nicely put together to catch up with papers on notable Transformer models. https://t.co/TT7hR8dL0j https://t.co/bjzzlRZ9fy

The entire motion capture industry is about to change. 4DHumans utilizes a fully "transformerized" network, HMR 2.0, for human mesh recovery. This network is capable of analyzing a single image or video and creating detailed 3D models of the people in it. These 3D models are… https://t.co/f6He1OVaKu https://t.co/HjM4lqNqOp