🎧 New: @pmarca joins @JordanHarbinger to talk AI, its role in creative industries and how it may affect the U.S.'s relationship with China. Listen now: https://t.co/0ZQsPYJPVB
— a16z (@a16z) Aug 31, 2023
LangChain x Xata New LangChain integrations from @xata–both vector and memory stores. a neat walkthrough of building a Q&A chatbot with them, new on the blog https://t.co/AHdbilW9uM
— LangChain (@LangChainAI) Aug 29, 2023
3 curious facts about @huggingface's history, mission, and most famous products: 1. Hugging Face started as an AI chatbot in 2015 by three leaders: @ClementDelangue, @julien_c, and @Thom_Wolf 2. In 2017, their mission crystallized from being mere chatbot developers to… https://t.co/fQTCQjAGvj https://t.co/jPeX8Fm2uN
— TuringPost (@TheTuringPost) Aug 29, 2023
Oh shit, new github co-pilot updates! They stole cursor's UI!! Now includes the terminal in context! And 8k context window! Only in Visual Studio though... (i think???) Not VScode :( https://t.co/arznYNWiA3
— Nick Dobos (@NickADobos) Aug 28, 2023
Getting LLaMA to produce structured outputs (e.g., JSON) is a challenge. @evanqjones + @GrantSlatton work on grammar-based sampling is a cool approach: supply a grammar file to guide / constrain sampling. Thanks @deepsense_ai for adding to @LangChainAI llamacpp integration ... https://t.co/xrKRuwijKo
— Lance Martin (@RLanceMartin) Aug 28, 2023
Request for startup — we'd pay a lot for this https://t.co/Dml70nzhyq
— Flo Crivello (@Altimor) Aug 28, 2023
Introducing ChatGPT Enterprise: enterprise-grade security, unlimited high-speed GPT-4 access, extended context windows, and much more. We’ll be onboarding as many enterprises as possible over the next few weeks. Learn more: https://t.co/lU3ErLtBmT https://t.co/lhixa4IXlc
— OpenAI (@OpenAI) Aug 28, 2023
https://t.co/5SuP9rYa2y paints a dire picture for AI startups and compute: @huggingface - Too "GPU-poor" given their reputation, need more investment in models and infrastructure to compete. Leaderboards are misleading open source work in wrong direction. @databricks - Acquired… https://t.co/POk7OYWCsB
— /MachineLearning (@slashML) Aug 28, 2023
Claude is crushing ChatGPT at copywriting. I've done a direct comparison in 6 copywriting tasks. Here are the results: https://t.co/ebwckM7gE9
— Moritz Kremb (@moritzkremb) Aug 27, 2023
You can get nice observability w/ LLM+knowledge graph queries with `ipython-ngql` (@NebulaGraph). Simply run text-to-cypher, and then render the subgraph that the cypher query ended up selecting 🎨 One of my fav parts of our @llama_index KG workshop: https://t.co/xGwxKvhEHz https://t.co/0vb83Auawf
— Jerry Liu (@jerryjliu0) Aug 27, 2023
Very impressive. WizardCoder 34B outperforms GPT-4, Claude, and Bard on HumanEval. It also surpasses all open-source Code LLMs by a substantial margin. WizardCoder 34B is fine-tuned version of Code Llama model using Evol-Instruct, a synthetic dataset. The code, model weights,… https://t.co/LRio9LQIkz https://t.co/kq7mpenrv1
— Lior⚡ (@AlphaSignalAI) Aug 26, 2023
ChatGPT and Canva are a powerful combo. You can create dozens of visuals automatically in seconds. I'll show you how to do it in just 4 simple steps: https://t.co/p8A6wq2WhP https://t.co/9pOzMwtjEj
— Paul Couvert (@itsPaulAi) Aug 26, 2023
Complex and large documents often have many links and references. "the start date is 2020-01-1" only to reference it 9 pages later as "the termination date is 45 days after the start date". Using gpt-4 + instructor I created and resolved entities as a graph. https://t.co/Wn2m7HLdT3
— jason (@jxnlco) Aug 25, 2023
One of the best ways to reduce hallucinations with LLMs is by retrieving useful, factual information and injecting it into the LLM’s prompt as added context. Although this might sound complicated, it’s actually quite easy to implement with standard vector search functionality…… https://t.co/sxUN0eDrEl https://t.co/ySB4nNGgeA
— Cameron R. Wolfe, Ph.D. (@cwolferesearch) Aug 25, 2023
This remains the highest signal-per-word page on the internet. https://t.co/jWIt6yXu1c https://t.co/V2kZXLcDxf
— Mckay Wrigley (@mckaywrigley) Aug 25, 2023
Code llama was released yesterday! @Gradio demo (13B) parameter model available today: https://t.co/3v3udUoj0c
— Gradio (@Gradio) Aug 25, 2023
Preserved Edo Period (1603-1868) Neighborhood in Japan. Narai-juku is a historical post town located in the Kiso Valley of Nagano Prefecture, Japan. It was one of the sixty-nine post towns along the Nakasendo, an ancient road connecting Kyoto and Edo (present-day Tokyo) during… https://t.co/5K40NQ8f3h https://t.co/u6oD6u7UAz
— Historic Vids (@historyinmemes) Aug 24, 2023
We are excited to announce Lilac (@lilac_ai), a new Open Source product to analyze, structure and clean data with AI. Our mission is to make unstructured data visible, quantifiable, and malleable. 🌸 Announcement blog: https://t.co/6Hp1wa0pqf https://t.co/3SR1FJW7BI
— Nikhil Thorat (@nsthorat) Aug 24, 2023
if you are a systems thinker, these are your 12 Commandments once seen cannot unsee https://t.co/4rKQ2KlZhY https://t.co/fHyXIc9qIw
— swyx (@swyx) Aug 24, 2023
I cannot believe zuck et al just beat gpt 3.5 at humaneval pass@1 and is approaching gpt4 with only 34b params (47 pages, therefore reaction thread - code llama) https://t.co/LJ54F9iUMS https://t.co/zURs3U1Ul0
— kache (yacine) (e/boy ) (@yacineMTB) Aug 24, 2023
Llama-2 was almost at GPT-3.5 level except for coding, which was a real bummer. Now Code Llama finally bridges the gap to GPT-3.5! Coding is by far the most important LLM task. It's the cornerstone of strong reasoning engines and powerful AI agents like Voyager. Today is… https://t.co/I0YlWsyRrQ https://t.co/Jwrq4gPhi1
— Jim Fan (@DrJimFan) Aug 24, 2023
@michael_nielsen We released a model (now called stableBeluga) that hit chatGPT (3.5) levels on evals, I think a few now a bit better. Original LLaMA 13b base beat GPT3 on evals. https://t.co/PuvLGHDFAX
— Emad (@EMostaque) Aug 24, 2023
JUST IN: Meta AI releases Code Llama, a family of LLMs for code based on Llama 2. The models provided as part of this release: - Foundation base models (Code Llama) - Python specializations (Code Llama - Python), - Instruction-following models (Code Llama - Instruct) Model… https://t.co/x0wYaHJ9mO https://t.co/Wgd2fgdsgd
— elvis (@omarsar0) Aug 24, 2023
We measured legal reasoning in lots of LLMs across 162 tasks, w/ a big team of AI & legal researchers! https://t.co/k1X7SMNE3q https://t.co/dcHemcBTDT
— John Nay (@johnjnay) Aug 24, 2023
Huge props to the `trl` team at @huggingface for authoring the best content around doing all kinds of policy optimizations for LLMs. They do it keeping accessibility at the forefront 🤗 Hope to bring some of that to 🧨 diffusers someday. For now, enjoy https://t.co/35rZVa6hlB
— Sayak Paul (@RisingSayak) Aug 23, 2023
If you finetune a model on @OpenAI, it's very easy to use it in LangChain - just pass in the new model name and that's it See the code snippet below https://t.co/RU2q0lOh0z
— LangChain (@LangChainAI) Aug 22, 2023
🧪 Tutorial: Quick Midjourney Inpainting Process While Midjourney is slowly rolling out the new InPainting feature for everyone to use, here's a quick tutorial on what you can expect from it! Let's dive in 👇 https://t.co/G7jw6nmI4s
— Nicolas Neubert (@iamneubert) Aug 21, 2023
Very interesting paper extending CoT to graphs of thoughts (allows loops and arbitrary connections), that increases reasoning capabilities of LLMs Graph of Thoughts: Solving Elaborate Problems with Large Language Models https://t.co/4n33l8C56Y https://t.co/oZLnuutI06 https://t.co/oNRfTqhT2j
— Dimitris Papailiopoulos (@DimitrisPapail) Aug 21, 2023
8-21-2023 ($) •Adyen Earnings •Adyen's European Context •Adyen vs. Stripe https://t.co/uHCOtKU1tr
— Stratechery (@stratechery) Aug 21, 2023
Excited to share that my newest article "Building Custom Q&A Applications Using LangChain and Pinecone Vector Database" on Analytics Vidhya is out now📝🚀 Here's the link to the article and check out the below posts for quick synopsis of this article👇 https://t.co/lDhjL8uw6I
— Avi Kumar Talaviya (@avikumart_) Aug 20, 2023
Just found this great video by @1dolinski. He went through my tutorial on how to make consistent coloring book characters, step by step. Awesome stuff, mate! https://t.co/2aBcOhBMzO
— Chase Lean (@chaseleantj) Aug 20, 2023
Finally got a chance to read @eugeneyan's write-up on practical patterns for building LLM-based systems and products. Really good insights and references throughout covering the following topics: - Evaluation - covers why evaluation is needed, how to apply evals, and the… https://t.co/pBzDy8oRK2 https://t.co/Xg1zWrfBzh
— elvis (@omarsar0) Aug 19, 2023
So the fattest country on the list charges 10x what the skinny countries charge for Ozempic and wegovy? https://t.co/mw3mTyOaTk
— @jason (@Jason) Aug 18, 2023
This is a solid and fast long-context model for summarization, Q&A etc. It's better than GPT-3.5x 16K on benchmarks... surprising for a model likely 3-4x smaller! https://t.co/SKdu4MjdUQ
— Vipul Ved Prakash (@vipulved) Aug 18, 2023
Quick, Cheap, and Powerful Refinement of LLMs It presents Platypus, a family of fine-tuned and merged LLMs currently topping the Open LLM Leaderboard. It describes a process of efficiently fine-tuning and merging LoRA modules and also shows the benefits of collecting… https://t.co/lPmWskPKcr https://t.co/pqXFincJeQ
— elvis (@omarsar0) Aug 18, 2023
Use @LangChainAI and @MyScaleDB to reduce hallucination in #LLMs: https://t.co/e6Go6TjR6A
— Qin Liu (@qliu52) Aug 18, 2023
Who are building agents? AutoGPT, MutlON, Minion, Beebot, BabyAGI, AI Legion, Bloop AI, Camel AI, GPT Engineer, localGPT, MiniAGI, Multi-GPT, OpenAGI, Pezzo, ReactAgent... Please tag any missing projects below (with name and with github/website link if available).
— Sasha👩🏻💻 (@hackgoofer) Aug 18, 2023
Llama 2 is an incredible model, but it’s not very good at text-to-SQL. Enter fine-tuning 💡 We’re excited to release an e2e repo/tutorial - fine-tune Llama 2 on text-to-SQL, use it with @llama_index 💫 Best of all you can do it yourself! 🎁👇 https://t.co/ECFoALDnYN https://t.co/g7dKtM4I22
— Jerry Liu (@jerryjliu0) Aug 17, 2023
The recent Bayesian Flow Networks paper (https://t.co/698R3Hl6qm) is quite cool but I haven't seen an accessible explanation yet. I think if you've played the board game Codenames you already mostly know how BFNs work tl;dr: you're teaching a network to listen to a spymaster 1/n https://t.co/Xc6guEiFwI
— Eric Zelikman (@ericzelikman) Aug 17, 2023
There’s now two main techniques emerging for using LLMs to query a knowledge graph 🔥: 1️⃣ Text2cypher 2️⃣ Knowledge graph-based RAG (recently introduced in @llama_index) When should you use one or the other❓ Why should you use knowledge graphs at all? 🤔 See below + video 👇… https://t.co/SJVsHa4ON9 https://t.co/oDshFnYrVr
— Jerry Liu (@jerryjliu0) Aug 17, 2023
🚢Benchmarking Question/Answering Over CSV Data Deep dive on improving an application that does question answering over CSV data: 📜3000 word blog post 🎥30min video 🛌Open sourced eval data 🎬Open sourced code for gathering feedback 🤖Open sourced final agent code 🧵 https://t.co/IPHF8M4imG
— Harrison Chase (@hwchase17) Aug 15, 2023
With the rise of open LLMs, like Llama 2, LLMOps might become a critical component of AI applications💪🏻 I wrote a new blog post on how to use AWS CDK to easily deploy and manage Llama 2🚀 👉 https://t.co/sCj9DDWpYB Help your Ops team and make your LLMs accessible 🤗
— Philipp Schmid (@_philschmid) Aug 15, 2023
Want to guide LLMs into always outputting answers that are valid JSON or satisfy a regex pattern? https://t.co/X5lZWOk1DH is a set of structured LLM APIs that always returns structured data. Three APIs that you can use in production today: • /v1/completion/regex Given a… https://t.co/76bVE5JPBJ
— Matt Rickard (@mattrickard) Aug 14, 2023
If this does not make you fall in ♥️ with @LangChainAI , I don't know what will. Amazing presentation made by @RLanceMartin on some of the most common use cases of generative AI which can be easily implemented by LangChain. If you haven't already, go check it out and I am sure… https://t.co/rz6XNKIZk9 https://t.co/u2wymBrvxN
— Hard Kothari (@HardKothari) Aug 14, 2023
Rank Responses to Align Language Models with Human Feedback (RRHF): A new method to align LLMs from preferences. 🔎🤖 👉https://t.co/KqGsCG6zDg 🧵1/3 https://t.co/5jwodGGdNS
— Philipp Schmid (@_philschmid) Aug 13, 2023
This is what I been saying, fine tuning is gonna continue to be a thing - and task specific models will be very capable. Found this recently published from Roblox https://t.co/8ifASKdmMf
— anton (@abacaj) Aug 12, 2023
Last week @sweep__ai wrote a blogpost on chunking 2M+ files a day for code search that got to the front page of HackerNews We're super excited to have them our webinar next week to talk about what data ingestion looks like in production! https://t.co/YAmf2i0nZZ
— LangChain (@LangChainAI) Aug 11, 2023
PaLM is one of the most advanced LLMs. PaLM achieved remarkable results on a variety of tasks and has inspired other works such as PaLM 2, Med-PaLM, and all sorts of real-world applications, including Bard. There is no doubt about the impact of PaLM so understanding this… https://t.co/rzFT3AZ51h
— DAIR.AI (@dair_ai) Aug 10, 2023
Simulating Multi-Agent Social Networks Excited to share a writeup by the 2nd team at a recent @AnthropicAI hackathon They pushed the limits of prompt engineering to construct a complex multi-agent social network simulation-and used LangSmith to do it https://t.co/sKwzJ2oJTw
— LangChain (@LangChainAI) Aug 10, 2023
Scoop & first time sharing a byline with @steph_palazzolo & @erinkwoo: VCs are eager to back Tome, one of the buzziest AI startups, at a $600M valuation. It *just* started generating revenue. https://t.co/XIM1LyVbnf
— natasha mascarenhas (@nmasc_) Aug 9, 2023
LLMs have a hard time saying "no" or "i don't know" because they've been so fine-tuned to please. Thus, the context we provide to the LLM is crucial. This makes the R in RAG so important, especially with regard to precision. https://t.co/dpiy3NBSNF
— Eugene Yan (@eugeneyan) Aug 9, 2023
JUST IN: Google announces a new browser-based code environment. It will bring the entire full-stack and app development workflow to the cloud. It also includes Generative AI features based on PALM 2: ▸ Code generation ▸ Code completion ▸ Translating code between languages… https://t.co/SAwqyfMOU7 https://t.co/JepX57Vv1D
— Lior⚡ (@AlphaSignalAI) Aug 8, 2023
Humans can’t compete with this: Earlier this week, I invited an artificial agent to my class, and now I'm hooked. The agent was @firefliesai, an AI meeting assistant. I added it to my meeting invite just like any other participant and forgot about it. I finished my class 2… https://t.co/l970neubH4 https://t.co/gTFHMFUNku
— Santiago (@svpino) Aug 8, 2023
AgentBench: Evaluating LLMs as Agents Presents a multi-dimensional evolving benchmark that currently consists of 8 distinct environments to assess LLM as Agent's reasoning and decision-making abilities in a multi-turn open-ended generation setting. repo: https://t.co/82ErWHRj72… https://t.co/bHgGuBg39Y https://t.co/gZhdwragzX
— Aran Komatsuzaki (@arankomatsuzaki) Aug 8, 2023
YouTube Shorts are the easiest way to create a profitable YouTube channel. Now with ChatGPT & Canva you can create hundreds YouTube shorts in minutes. Here's a 4 step guide on how I created 30 YouTube shorts in 3 minutes using AI:
— Mark Gadala-Maria (@markgadala) Aug 6, 2023
Introducing MetaGPT: The Multi-Agent Framework 🥷 MetaGPT takes a one line requirement as input and outputs user stories / competitive analysis / requirements / data structures / APIs / documents, etc. Internally, MetaGPT includes product managers / architects / project… https://t.co/uTgEZ7cUoo https://t.co/LK9WF6dwhI
— Itamar Golan 🤓 (@ItakGol) Aug 6, 2023
If you want to learn math by yourself, here is a beautiful book list 👇 30 Best Math Books to Learn Advanced Mathematics for Self-Learners https://t.co/7jnktM9UxK
— Math Lady Hazel 🇦🇷 (@mathladyhazel) Aug 5, 2023
🚨 Exciting News! Andrew NG has just released a new short courses on Generative AI with Hugging Face. The best part? It's absolutely FREE! https://t.co/N466ftKldS
— Shubham Saboo (@Saboo_Shubham_) Aug 5, 2023
Startups finally have to admit an ugly truth: Most of them are not worth close to what they thought. Inside recent cram-down deals, filings & emotional reckonings in private tech land, w/ @KateClarkTweets https://t.co/0YFheU3ca0 https://t.co/h6eJZ60RLj
— Cory Weinberg (@coryweinberg) Aug 4, 2023
The most detailed and practical write up on applying LLMs! 🙏 This reads like a survey paper but written for the industry and applications @eugeneyan is known as the best NLP writer for a reason. It’s the most comprehensive overview of patterns on building Large Language Models… https://t.co/Uc2QLSk9po https://t.co/8etXVExSAd
— Sanyam Bhutani (@bhutanisanyam1) Aug 4, 2023
1/15 While Adam Posen correctly recognizes the problems the Chinese economy currently faces, his explanation of what has gone wrong, and what Beijing must do to revive the economy is, in my opinion, completely off the mark. https://t.co/CHxcqE2wqH via @ForeignAffairs
— Michael Pettis (@michaelxpettis) Aug 4, 2023
Function calling in Llama: https://t.co/Mbh9l8hjGv
— Teknium (e/λ) (@Teknium1) Aug 4, 2023
“Affordable RLHF for all” ❤️ It’s almost like an openly rebellious group at MSR have decided to subvert Microsoft’s investments in ClosedAI. https://t.co/znbMymeZiN https://t.co/a5P1OegXv9
— Delip Rao e/σ (@deliprao) Aug 4, 2023
A great (new) guide and overview on securing LLM systems against prompt injection by @nvidia We did a webinar on prompt injection a few months and the main takeaway was more awareness was needed around this. Great to see posts like this doing that https://t.co/334tYjB5Ss
— Harrison Chase (@hwchase17) Aug 3, 2023
Apple Q3 revenue: iPhone down 2.4% YoY to $39.7B, Mac down 7.3% YoY to $6.8B, iPad down 19.8% to $5.8B, and Wearables, Home, and Accessories up 2.4% to $8.3B (@aatilley / Wall Street Journal) https://t.co/v17jmmW3nn https://t.co/0rT0RBxAjw
— Techmeme (@Techmeme) Aug 3, 2023
An underrated aspect of LlamaIndex is that we have FULL support for your custom retrieval/RAG workflows. Don’t like our out of the box retrievers? No problem, easily write your own. @ravithejads shows you how to write a custom retriever in our tutorial/notebook combo 👇 https://t.co/XKWo1nJLew
— LlamaIndex 🦙 (GPT Index) (@llama_index) Aug 3, 2023
AI has long delivered results to incumbent vendors like Google and Meta. But the economics and behavior shifts that come with generative AI provide a huge opportunity, now for startups and the overall economy, say @martin_casado and @sarahdingwang: https://t.co/xNhVD47EX2
— a16z (@a16z) Aug 3, 2023
“Potential cancer breakthrough as 'groundbreaking' pill annihilates ALL types of solid tumors in early study” https://t.co/K0NtHOQU4M
— Brian Roemmele (@BrianRoemmele) Aug 2, 2023
StableBeluga 7b and 13b models now top performing 7b and 13b models on the @huggingface 🤗 OpenLLM leaderboard try them out https://t.co/2HFnezuHUh https://t.co/uPd1oB4RuQ https://t.co/UWHsd6JgCn https://t.co/TPOXvCUqUK
— Emad (@EMostaque) Aug 2, 2023
AlphaFold was just the start of our journey in using AI to understand proteins, which power every biological process. 🧬 Scientists are now harnessing the tool to create brand new proteins - a potential game changer for medical science and drug discovery. https://t.co/5HcMIWusJT
— Google DeepMind (@GoogleDeepMind) Aug 1, 2023
The book "Interpreting Machine Learning Models With SHAP" is published 🥳 A deep dive into SHAP, the popular method to interpret machine learning models. Theory, tips, code examples, limitations -- all covered. Get your copy with a discount: https://t.co/yfxMoDJkqz
— Christoph Molnar (@ChristophMolnar) Aug 1, 2023
