The closest known black hole to Earth is just 1,560 light-years away https://t.co/pkX7A9cjNZ
— Curiosity (@MAstronomers) Apr 30, 2024
from Twitter https://twitter.com/MAstronomers
April 30, 2024 at 03:05AM
via IFTTT
The closest known black hole to Earth is just 1,560 light-years away https://t.co/pkX7A9cjNZ
— Curiosity (@MAstronomers) Apr 30, 2024
A marvelous country in Mediterranean Croatia 🇭🇷, a thread 🧵 1. Split https://t.co/l9iyU77TW6
— gisipen (@gisipen) Apr 29, 2024
Another media partnership announcement from OpenAI, this time with the Financial Times. The FT was already an Enterprise customer, now they are a partner. The NYT increasingly stands alone. https://t.co/x7XHcmEtDW https://t.co/knYGtAECQU
— Andrew Curran (@AndrewCurran_) Apr 29, 2024
This is the Hotel Splendido in Portofino, where they serve a dish called “spaghetti alla Liz Taylor”, with a simple but exquisite tomato sauce. In fact, she was often a guest at the hotel, and Richard Burton proposed here. Recipe in the first comment 👇 https://t.co/hCDsyvuMxc
— Mambo Italiano (@mamboitaliano__) Apr 28, 2024
Do not go gentle into that good night but rage, rage against the dying of the light. https://t.co/zkkbBtE3Eb
— Unravel Virtue ※ (@unravelvirtue) Apr 27, 2024
Pretty amazing progress in voice cloning and LLMs in the last year. Here’s a super low latency demo from @play_ht (YC w23) where you can talk to AI me. https://t.co/Odfy1Q8FKZ
— David Lieb (@dflieb) Apr 27, 2024
Steve Jobs explains why computers are a bicycle for the mind: “We humans are tool builders” Steve recalls reading an article in Scientific American when he was 12 years old. The authors measured the efficiency of locomotion for various species (e.g. how many kilocalories did… https://t.co/CcfJwLGxCj https://t.co/VOaeV3onkz
— Startup Archive (@StartupArchive_) Apr 28, 2024
Direct Nash Optimization (DNO) better than DPO for RLHF/RLAIF? 🤔 DNO is a new RLHF method from @Microsoft Research using a batched on-policy algorithm that conducts self-improvement iteratively via contrastive learning. 👀 Implementation: 0️⃣ Start with an SFT LLM (☸️), e.g.,… https://t.co/2sSe6mTyHI https://t.co/NLSS6YuFOA
— Philipp Schmid (@_philschmid) Apr 28, 2024
The beautiful islands of Greece you haven’t seen before - a thread🧵👇 1. Zakynthos island https://t.co/2PndIYbNAU
— OThingstodo (@othingstodo_com) Apr 27, 2024
3 podcasts I re-listen to often (because of the huge impact they had on my thinking at the time)… https://t.co/k87FLL977o https://t.co/lsBf3liSj8
— Dickie Bush 🚢 (@dickiebush) Apr 27, 2024
Still remains the simplest, most lucid explanation of Large Langauge Models by one of the sharpest (and kindest) teachers (and engineers) of our time: Andrej Karpathy Give it a watch: https://t.co/PwIuESZ0BZ
— Electrik Dreams (@electrik_dreams) Apr 27, 2024
Added two new chapters to the advanced DSPy tutorial 🔥 In the second chapter, I explain how to improve the knowledge graph extractor we built in the first chapter, aiming to reduce duplicate entities. In the third chapter, I explain what coreference resolution is and why it is… https://t.co/w8jb37tFG8 https://t.co/6pcHWA7iW4 https://t.co/4ZcuDGI8hK
— Franck SN (@ndzfs) Apr 27, 2024
“So I asked Ilya Sutskever, OpenAI’s chief scientist, for a reading list. He gave me a list of like 40 research papers and said, ‘If you really learn all of these, you’ll know 90% of what matters today.’ And I did. I plowed through all those things and it all started sorting out… https://t.co/k4R1eyTRMd https://t.co/kQE9SvqJ2k
— NIK (@ns123abc) Apr 27, 2024
Totally agree! https://t.co/9hyYnOA7r8
— Bill Gurley (@bgurley) Apr 26, 2024
Steve Jobs: "Don't be a career." https://t.co/WmCXpN80Pl
— Z Fellows (@ZFellows_) Apr 26, 2024
How to Maximize LLM Performance is a well-written blog focusing on a session from @OpenAI Devday in 2023. 👀 With the rise of open LLMs, it is a good refresher, but perspectives can change, like OpenAI's evolving position on custom models. 💡🥊 👉 https://t.co/PkKD0vlIvH https://t.co/OF3Q9ks9AN
— Philipp Schmid (@_philschmid) Apr 26, 2024
From unstructured to structured data using DSPy Full tutorial on Lycee AI (link in second tweet) https://t.co/ghIJwIEMJx https://t.co/4ZcuDGI8hK
— Franck SN (@ndzfs) Apr 26, 2024
🧠 Microsoft introduces FILM: Helps an LLM Fully Utilize the Context! 🚀32K-context 🦾Trained from Mistral-7B-Instruct-v0.2 💪Overcomes LLM's lost-in-the-middle problem 🆕Information-Intensive (In2) Training 💯Performs well in long-context tasks 🚀 26.9 F1 score on NarrativeQA https://t.co/hte38jhDcT
— Gradio (@Gradio) Apr 26, 2024
We tested @OpenAI's GPT 3.5 & 4, @Meta's Llama 3, @Google's Gemini 1.5 Pro, @cohere's Command R+ and 130+ other LLMs on writing code. Deep dive blog post soon 🏇 In the meantime: a screenshot of the results (more points is better) https://t.co/p2WJAQvFof
— Markus Zimmermann (@zimmskal) Apr 25, 2024
Impressive. Chips Act funding got basically every major chipmaker to build in the US. “That means the US will become the only country in the world with facilities run by all of the top manufacturers.” https://t.co/yMSqfTzg5c
— Kyle Chan (@kyleichan) Apr 25, 2024
Yesterday, we open sourced the Cohere Toolkit. We think this will be a major accelerant for getting LLMs into production within enterprise. https://t.co/uN4zohd0Xe
— Aidan Gomez (@aidangomez) Apr 25, 2024
Snowflake announces Arctic, an LLM optimized for enterprise tasks such as SQL generation, coding, and instruction following, with an Apache 2.0 license (@mr_bumss / VentureBeat) https://t.co/eL3XwXMjkX 📫 Subscribe: https://t.co/OyWeKSRpIM https://t.co/24QoGQO0fO
— Techmeme (@Techmeme) Apr 25, 2024
Anduril's engineers beat out Lockheed Martin, Northrop Grumman, and Boeing for their largest ever contract (surely at or over a billion) to build an autonomous warplane America still has so much greatness in its people https://t.co/w1klXhz68J
— ib (@Indian_Bronson) Apr 24, 2024
Vapor cone and condensation during an F-22 maneuver [📷Mark and Taj / point_mugu_skies] https://t.co/qUOw77BUty
— Massimo (@Rainmaker1973) Apr 24, 2024
Took notes furiously at @stripe sessions today when Jensen Huang (@nvidia CEO) talked about love and care being essential to making extraordinary technology companies https://t.co/hvQ3O6YU9J
— Nix 🕊 (@startingfromnix) Apr 25, 2024
4-25-2024 An Interview with Walmart CEO Doug McMillon About Omnichannel Retail https://t.co/JM8YXm9M65
— Stratechery (@stratechery) Apr 25, 2024
Why did I deepfake myself? To see if conversing with an AI-generated version of myself can lead to self-reflection, new insights into my thought patterns, and deep truths. https://t.co/DWODoZ9lXL
— Reid Hoffman (@reidhoffman) Apr 24, 2024
what in the good lord is this!!! Phi-3-mini-128k?!! https://t.co/cfPhahCGma
— Maziyar PANAHI (@MaziyarPanahi) Apr 23, 2024
A gentle introduction to #DSPy... I started kicking the tires on DSPy a couple of weeks ago. I struggled. It was complicated and... I have grug brain. This video is a GENTLE introduction to DSPy. Now, don't get me wrong, @lateinteraction and @stanfordnlp are responsive on… https://t.co/AvuN1uf59e https://t.co/NlF8O8XSPy
— Bill Chambers (@bllchmbrs) Apr 23, 2024
Sam Altman explains the two strategies for startups building on AI “I think fundamentally there are two strategies to build on AI right now. There’s one strategy which is assume the model is not going to get better and build all of these little things on top of it. Then there’s… https://t.co/MBHt9FG7th https://t.co/dsZOoWsMnl
— Startup Archive (@StartupArchive_) Apr 23, 2024
AI agents have the ability to fundamentally change the business model of enterprise software. Today, when you build a SaaS product, the primary business model is to sell seats that are tied to the end-users of your service. For "same store" sales, you can essentially grow at the… https://t.co/bd0NnCbIWI
— Aaron Levie (@levie) Apr 23, 2024
How a book written in 1910 could teach you calculus better than several books of today [Full text: https://t.co/55dkZAfec3] https://t.co/HjTVqtI5oM
— Massimo (@Rainmaker1973) Apr 23, 2024
There aren't nearly enough thoughtful papers on evals so I'm excited to go through @sh_reya's paper in detail https://t.co/Cv1Pb2AEfi https://t.co/cQYRS2RE3f
— Eugene Yan (@eugeneyan) Apr 22, 2024
cool experimentations with generative ui https://t.co/2UO6jCm81P
— LangChain (@LangChainAI) Apr 22, 2024
Teenager Christopher Slayton built the ‘entire known universe’ in mine craft https://t.co/ewnirHgDW8
— Science girl (@gunsnrosesgirl3) Apr 23, 2024
AutoCrawler Combines LLMs with crawlers with the goal to help crawlers handle diverse and changing web environments more efficiently. LLMs are powerful at information extraction so it's possible to apply them in many problems that require web automation. Capabilities like… https://t.co/SSqVavImjd https://t.co/WEIBnQkTjF
— elvis (@omarsar0) Apr 22, 2024
The Emergence of Agentic AI: Embracing Collective Intelligence Introduction As we explore the realm of artificial intelligence (AI), it is crucial to challenge our preconceived notions of intelligence and recognize the true nature of general intelligence. The concept of… https://t.co/QjXHJV00jB
— Carlos E. Perez (@IntuitMachine) Apr 23, 2024
If the rise of LLMs caught you by surprise, here's your chance to get a preview of what's likely to be the next monumental jump in AI capabilities: LLM-backed agents that use software tools In this video, I'll walk you through the concepts and code of building an LLM-backed… https://t.co/7LHa6BqTUH https://t.co/sZL18Br8lw
— Jay Alammar (@JayAlammar) Apr 22, 2024
i took everything i read about @AIatMeta llama 3 release and wrote a 2300 word newsletter about it. lots of implications from the release about scaling laws, sizes of models etc also covered @finkd on @dwarkesh_sp podcast and how its becoming clear that energy is the next big… https://t.co/NpksqRrqC8
— Nofil (@lostlifon) Apr 23, 2024
The classic hourglass structure of paper writing https://t.co/9oRk9g9cS9
— Prof Lennart Nacke, PhD (@acagamic) Apr 22, 2024
"But something slowly began to dawn on me — I still loved what I did. The turn of events at Apple had not changed that one bit. I had been rejected, but I was still in love. And so I decided to start over." — Steve Jobs
— Steve Jobs Stories (@SJobs_Stories) Apr 21, 2024
If you're looking to learn about how Transformers "ate" other architectures here's a great overview by Lucas Beyer (@giffmana) from @GoogleDeepMind. The slides are available at https://t.co/qi81G26pM6. https://t.co/QLPu8RE79u
— Jarek Wilkiewicz (@wjarek) Apr 22, 2024
CrewAI LLaMa 3: How I created AI Agents to Automate News Agency and Integrated it with a Database 🔧 API Setup 📊 Database Integration 🛠️ Tool Build Custom Tool 🤖 Build AI Agents 🎯 Assign Tasks 📝 Using @crewAIInc (by @joaomdmoura ) @GroqInc Subscribe:… https://t.co/KdpYrWce3b https://t.co/lJV11YPY9t
— Mervin Praison (@MervinPraison) Apr 22, 2024
What an amazing demo by Tim He used Claude Opus LLM to extract data from image to convert into a form using AI How cool is that? Imagine what GPT-5 might do.. https://t.co/UKlVBv7Kwl
— Prashant | AI (@Prashant_1722) Apr 21, 2024
The next big breakthrough in AI is AI Agents. This is when AI goes from being used as an assistant to chat with, to using AI to accomplish complete tasks that a human might otherwise have to perform. This moves AI from being a "read-only" operation to fundamentally a "read/write"… https://t.co/jgikASu1YZ
— Aaron Levie (@levie) Apr 21, 2024
Kick-ass online machine learning courses: • Coursera - Machine Learning Specialization • MIT 6.S191 Introduction to Deep Learning • DS-GA 1008 Deep Learning • UC Berkeley Full Stack Deep Learning • Cornell Tech CS 5787 Applied Machine Learning It's now up to you.
— Santiago (@svpino) Apr 21, 2024
The color of a star is a function of its surface temperature https://t.co/FCQQ22QYxF
— Historic Vids (@historyinmemes) Apr 21, 2024
Google's Griffin paper is a PERFECT demonstration of scaling laws. Roughly 10% improvement in performance on tasks as they scaled parameters by 7x. All trained on the same 300 billion tokens data. Models get more sample efficient or extrapolate better as you scale across… https://t.co/1eNZL0ZHhD https://t.co/gmBL9Od0fL
— Rohan Paul (@rohanpaul_ai) Apr 21, 2024
“We are what we repeatedly do. Excellence, then, is not an act, but a habit.” —Aristotle
— David Senra (@FoundersPodcast) Apr 21, 2024
✂️But, How is Chunking Done? To create Better RAG applications, you need to know how to split or chunk the documents so you preserve the content while asking questions. Data Science Basics shows how to do this with LangChain https://t.co/zyelE0wZh7 https://t.co/cIBiLrsFV7
— LangChain (@LangChainAI) Apr 20, 2024
I'm learning about quantization. I learn best when I teach, so here are my notes which over time should morph into a "zero to hero" for quanting. https://t.co/IdsS4ATqno I'm excited - let's goo! https://t.co/w9XWkkuMdZ https://t.co/GifzAlDolb
— Umer Adil (@UmerHAdil) Apr 19, 2024
90% of my searches have moved from Google to https://t.co/3dP0wFuSot
— Michael Arrington 🏴☠️ (@arrington) Apr 20, 2024
Oh. My. God. Can someone please port this to PyTorch? https://t.co/x1pqZxDW8g
— Miles Cranmer (@MilesCranmer) Apr 19, 2024
Peter Thiel: Meaning is found in doing things that are important, that otherwise wouldn't get done. Do the work that only you can do. https://t.co/kPcnVkDoI6 https://t.co/YnJr979a33
— BUILD OR DIE (@BUILD_OR_DIE) Apr 20, 2024
#Bitcoin is 10% of the way through its reward era. This is still the first innings. https://t.co/REpzIeJftb
— Thomas | heyapollo.com (@thomas_fahrer) Apr 19, 2024
We now have an open-source model that is beating Claude 3 Opus... being served at nearly **300 tokens per second** on @GroqInc. The applications built off of this tech will be nothing short of revolutionary. https://t.co/NQuPho0IVL https://t.co/v934g0rwU5
— Matt Shumer (@mattshumer_) Apr 19, 2024
1/n Agentic AI: Can Machines Learn to Cooperate Like Humans? Imagine a world where AI agents seamlessly collaborate with humans and each other, tackling complex tasks with efficiency and precision. From disaster response teams coordinating rescue efforts to robots working… https://t.co/BVueBa807u https://t.co/WJrM1mQCpe
— Carlos E. Perez (@IntuitMachine) Apr 19, 2024
$META just released some new things: - Two small Llama 3 models (the larger, multimodal one being released in the coming months) - $META AI integrated into the search box of all of their assets (WA, IG, FB & Messenger) - Standalone website where you can get $META AI - $META AI… https://t.co/91zqmqNFNc
— Rihard Jarc (@RihardJarc) Apr 18, 2024
Meta'a Llama 3 model is released and we have day 0 support! @ravithejads and @LoganMarkewich put together this cookbook showing how to use Llama 3 directly from @huggingface, from running basic prompts to a full RAG pipeline: https://t.co/RMB7MhXIOA https://t.co/OIe3w9Cm21
— LlamaIndex 🦙 (@llama_index) Apr 18, 2024
Hello. We receive so many questions about agents in DSPy. Did you know you that ~15 lines of DSPy can turn an agent that scores 30% to a prompt-optimized, multi-agent aggregation system that scores 60% EM on a HotPotQA sample? I released a notebook showing how to do this 🧵⤵️ https://t.co/xO8Xel0OCh
— Omar Khattab (@lateinteraction) Apr 18, 2024
I'm begging you @OpenAIDevs, please tell us how your chunking works! It's a small detail, but it makes a huge difference in helping me make decisions about how to effectively use your RAG implement https://t.co/D9wgWkotmP
— Simon Willison (@simonw) Apr 17, 2024
The next generation of models seem to mostly target infinite context and adaptive compute per token. Basically, these two papers: Google: Mixture of Depths Google: Infini-Attention
— Casper Hansen (@casper_hansen_) Apr 17, 2024
The slowest of Kurzweil’s horsemen of the Singularity: nanobots https://t.co/rD23wFeD0x
— Ate-a-Pi (@8teAPi) Apr 17, 2024
We are open sourcing AIDE, a machine learning code generation agent. Simply describe the problem in natural language; the agent will begin experimenting on your local machine and then deliver the solution code. https://t.co/CjxWDTMXMb https://t.co/0b7iJ82EgT
— Weco AI (@WecoAI) Apr 17, 2024
Ultra realistic AI-video from a photo This is VASA-1 from Microsoft research The improvements in quality we’re getting between each new release is incredible. Links below https://t.co/wtIxsJkaYM
— Linus ●ᴗ● Ekenstam (@LinusEkenstam) Apr 18, 2024
Seriously, should this be made illegal? https://t.co/UcQzkKUhRz https://t.co/RMVjjLLHGl
— Carlos E. Perez (@IntuitMachine) Apr 17, 2024
This blog by Terrance Tao is extremely good, so many gold advice there. https://t.co/F1QlXb1yV6 https://t.co/Vn2dV9eAj0
— TK (@wordsofteekay) Apr 17, 2024
This is a nice improvement that makes iterating on evaluation easier You can define an LLM evaluator in the UI and attach it to a dataset. It will then run on all evaluation runs on that dataset https://t.co/p4Xr2WxWL5
— Harrison Chase (@hwchase17) Apr 16, 2024
@francoisfleuret Yeah graphics in Fukushima papers were something else. Also notice he casually introduces the ReLU back in 1969. Based. https://t.co/21W4XpY8cH https://t.co/MY7K9ovDWm
— milton (@miltonlopez_) Apr 17, 2024
My excellent Conversation with Peter Thiel: https://t.co/5pjoxbzOje, #Bible, #Calvinism, #Schmitt, #katechon, #Cowenconvos
— tylercowen (@tylercowen) Apr 17, 2024
OMG! I wish I had this technology when I was a kid! https://t.co/HafyiRMyFx
— Carlos E. Perez (@IntuitMachine) Apr 16, 2024
Check out this 7-part tutorial series by @ravithejads on an introduction to agents. Agents can be really complicated, but if you're trying to get a basic handle of what agents are and how to use them, this is a great intro. ✅ Learn how basic agent implementations work… https://t.co/d7wvdfMInM https://t.co/JNtzgb8IXL https://t.co/3VE09eUZn9
— Jerry Liu (@jerryjliu0) Apr 16, 2024
One of my favorite insights from this paper: knowledge graphs not only help with retrieval, but also with scientific discovery of new ideas/connections that were previously unexplored. You can do this by feeding an LLM an existing knowledge graph, and it will overlay the KG… https://t.co/hyTra27cTO https://t.co/xTwehQowOm https://t.co/no9IyX5WSr
— Jerry Liu (@jerryjliu0) Apr 16, 2024
Brainstorm ideas, analyze images, and process long documents with a conversational AI that is state-of-the-art.
— Anthropic (@AnthropicAI) Apr 11, 2024
Which are the most interesting companies working on one of these four things? 👇 “He pointed to a quartet of cognitive challenges: reasoning, planning, persistent memory, and understanding the physical world. “Those are four essential characteristics of human intelligence —… https://t.co/2ymn0ZibYl
— Patrick OShaughnessy (@patrick_oshag) Apr 15, 2024
Happy to share our survey preprint on using generative models for recommender systems. Awesome collaboration across industry and academia! This is my first paper after GDM. :) Paper: https://t.co/S8z5fgOPDy https://t.co/PGE0Icd96G https://t.co/zTBHJOvQjw
— Mahesh Sathiamoorthy (@madiator) Apr 15, 2024
Google is building a GPT store! Vertex ai agent builder, 1st look Exact same builder Ui, straight up clone Actions via openAPI schema Code interpreter! Notable Improvements: -Call tools & other agents via system prompt -example conversations. Mini training data sets https://t.co/Bply8zE6Fz
— Nick Dobos (@NickADobos) Apr 15, 2024
CosXL Edit: The Image Editing AI We've Been Waiting For The CosXL model from @StabilityAI might be THE image editing AI we've been waiting for. Just tell it how to change an image with a prompt. Wrote a 1 click launcher for this magical gradio app from @multimodalart. Enjoy! https://t.co/2mduiWADTq https://t.co/YvN5lGMLbz https://t.co/ldyig8kUeO
— cocktail peanut (@cocktailpeanut) Apr 13, 2024
This scene pretty much sums up this generation https://t.co/bUoxs3mvxb
— Historic Vids (@historyinmemes) Apr 14, 2024
I see a lot of debate on this site, so two things about the current state of AI agents in the real world 1) Agents remain limited by GPT-4 brains, it is early 2) Both academic testing 👇 & my own subjective experience (https://t.co/ASeRLJd73s) suggest agents have huge potential https://t.co/UDvNnqLblC
— Ethan Mollick (@emollick) Apr 13, 2024
Google may have multimodal AI that is head and shoulders above everyone else! https://t.co/F3yWjZ2wd2
— Carlos E. Perez (@IntuitMachine) Apr 14, 2024
Sunday morning read: Your AI Product Needs Evals by @HamelHusain ☕️ We are all building LLM applications, but how many of us have set up robust testing? In this article, Hamel begins with a case study on Rechat’s real estate AI assistant, Lucy. I loved the quote, “Addressing… https://t.co/CaGIg5VOMg https://t.co/FJdqrSANb1
— Erika Cardenas (@ecardenas300) Apr 14, 2024
1/n The AI That Grades Itself: Meet RankPrompt, the Self-Critiquing Language Model Imagine having an intelligent assistant that can flawlessly reason through complex problems, offering step-by-step solutions that are logical, coherent, and accurate every single time. As alluring… https://t.co/oIsfazigrc https://t.co/kvWVbSbVok
— Carlos E. Perez (@IntuitMachine) Mar 21, 2024
One of the most common concerns about AI is the risk that it takes a meaningful portion of jobs that humans currently do, leading to major economic dislocation. Often these headlines come out of economic studies that look at various job functions and estimate the impact that AI… https://t.co/eEtBL0j6u9
— Aaron Levie (@levie) Apr 6, 2024
How does Docker work? The diagram below shows the architecture of Docker and how it works when we run “docker build”, “docker pull” and “docker run”. There are 3 components in Docker architecture: 🔹 Docker client The docker client talks to the Docker daemon. 🔹… https://t.co/NTHO9zyTaU https://t.co/bE5cB9sUmK
— Alex Xu (@alexxubyte) Apr 13, 2024
How to fold an effectively flying paper fighter jet https://t.co/VvWlSgF0pc
— Massimo (@Rainmaker1973) Apr 13, 2024
Focus on what’s never going to change. - Warren Buffett https://t.co/uFaSXN8O1f
— Finding Compounders (@F_Compounders) Apr 13, 2024
The Paradox of Time according to Aristotle https://t.co/BPggcDMRW6
— Historic Vids (@historyinmemes) Apr 13, 2024
Honored and proud of our designers! https://t.co/pU3CLbZNiE
— Aravind Srinivas (@AravSrinivas) Apr 12, 2024
A spectacular sight 1225m (4019 ft) beneath the waves off Baja California as EVNautilus encounter the amazing Halitrephes maasi jelly. https://t.co/PcpF3hDxyn
— Wonder of Science (@wonderofscience) Apr 12, 2024
Steve Jobs on Life: “Remembering that you are going to die is the best way I know to avoid the trap of thinking you have something to lose. You are already naked. There is no reason not to follow your heart.” “You appear, have a chance to blaze in the sky, then you disappear.” https://t.co/hUqdXeoksx
— Z Fellows (@ZFellows_) Apr 12, 2024
There is a perception that the dominant companies of today are more powerful than those in the past. In fact, their market share is pretty similar to that of past global stars, but they contribute much less to global employment, productivity & growth than previous top companies. https://t.co/rO3hgDHm61 https://t.co/lxcYHJlQMw
— Ethan Mollick (@emollick) Apr 11, 2024
Adobe used images created by tools like Midjourney and uploaded to its stock marketplace by users, to train Firefly; Adobe says ~5% of images were AI-generated (Bloomberg) https://t.co/GTWv7A2ja7 📫 Subscribe: https://t.co/OyWeKSRpIM https://t.co/PckN7fzFP0
— Techmeme (@Techmeme) Apr 12, 2024
This is an excellent read. https://t.co/zcI4fezZJ3
— Ram Parameswaran (@_ram_) Apr 12, 2024
🚀Excited to introduce ResearchAgent🤖, your new partner in crafting the next best papers!📰 ResearchAgent is designed to revolutionize your research process, which automatically identifies problems, develops methods, and designs experiments. Paper: https://t.co/dRKncpXcMq https://t.co/V1Vk5WuFHZ
— Jinheon Baek (@jinheonbaek) Apr 12, 2024
🤖Palantir $PLTR analysis🔎 A one pager with all key information of a company that is increasingly becoming an A.I. platform for other businesses👇 Main takeaways📢 • $PLTR is well positioned to benefit from the A.I. megatrend with its A.I. Platform (AIP) gaining a lot of… https://t.co/nLGcpMHY0F https://t.co/vrh39TCo8c
— Investing visuals (@ZeevyInvesting) Apr 12, 2024
The ‘Palantir for AI’ thesis. Emergence - The Death of Deloitte: AI-Enabled Services Are Opening a Whole New Market https://t.co/NvdX16WU8I
— Bilal Zuberi (@bznotes) Apr 12, 2024
BREAKING: In Sudden Reversal, Harvard To Require Standardized Testing for Next Admissions Cycle @e1ysegoncalves and @matanjosephy report. https://t.co/PwcQUkFWCi
— The Harvard Crimson (@thecrimson) Apr 11, 2024
👀 Graph Chain-of-Thought - Graph-CoT : iterative reasoning on graph (with a 3 steps approach : LLM reasoning > LLM-graph interaction > graph execution) -GRBench : graph reasoning benchmark dataset Bravo to the team ! 👏 https://t.co/aSBUzSj1mV https://t.co/yWyBmf6UlC
— Michel (@Moi39017963) Apr 11, 2024
How do you start a startup that's building a supersonic jet? @bscholl did, and his company @boomaero just did their first flight. Here's a breakdown of 8 hard-tech companies with crazy ambitious ideas, and how they got started. https://t.co/yqQyzNcO3D
— Jared Friedman (@snowmaker) Apr 11, 2024
No surprise here, but Morgan Housel's new essay is great. "You want to try to emulate skills that are repeatable ... You can learn a lot from Warren Buffett's patience. But you can't replicate the market environment he had in the 1950s, so be careful copying the specific… https://t.co/Ybse2z3bZO
— Kevin Carpenter (@kejca) Apr 10, 2024
The weirdest/most interesting ongoing GenAI art project is probably the Infinite Backrooms: two instances of Claude 3 engage in conversation prompted to act like neo-Lovecraftian esoteric computers simulating reality. Strange & fascinating AI role-playing. https://t.co/o91mJ8OOOy https://t.co/ukbTCLMSpF
— Ethan Mollick (@emollick) Apr 11, 2024
1/n Griffin: DeepMind's Soaring Leap Beyond the Transformer Paradigm Imagine having a conversation with an AI assistant that can engage with you for hours on end, seamlessly incorporating context from your entire dialogue history. Or envision AI models that can analyze books,… https://t.co/aSwbx3lhIJ https://t.co/Uf6mgu6EaZ
— Carlos E. Perez (@IntuitMachine) Apr 11, 2024
Lucky vs. repeatable https://t.co/MjTwj2aYky
— Morgan Housel (@morganhousel) Apr 10, 2024
Langone is a business legend https://t.co/HQ79DB83Zc
— Elad Gil (@eladgil) Apr 9, 2024
Gemini 1.5 and Google's Nature Google Cloud Next 2024 was Google's most impressive assertion yet that it has the AI scale advantage and is determined to use it. https://t.co/YfcQCJ5G7u
— Stratechery (@stratechery) Apr 10, 2024
Discover the cutting-edge world of spoken LLMs with this comprehensive survey! link: https://t.co/vikqrRT04i https://t.co/h5jyJ6ekRr
— Hung-yi Lee (李宏毅) (@HungyiLee2) Apr 9, 2024
Few people shared this today and this made me think a lot https://t.co/GuiiShVNNf
— Kris Kashtanova (@icreatelife) Apr 10, 2024
MetaGPT (Hong et al.) is an awesome multi-agent framework (ICLR 2024) that models agents as a software company following a structured SOP - PMs, architects, engineers, and more communicate with each other to solve the task at hand. We’re excited to feature RAG-enhanced MetaGPT,… https://t.co/sAF41j0uL4 https://t.co/fcGFofq7gK
— LlamaIndex 🦙 (@llama_index) Apr 10, 2024
ComfyUI LLM Vision & Text Nodes Claude 3 & GPT-4 nodes in ComfyUI! Allowing for both image and text-based interactions within the ComfyUI workflow Setup link in post below: https://t.co/BmVfj6xXup https://t.co/y27C898XnA
— Allen T (@Mr_AllenT) Apr 8, 2024
Corollary of all this: To make the most of DSPy, you will need to think pretty differently about your problem. Don't think of the prompt you should be writing. Focus on the right pipeline/program and the right metrics you want to maximize. https://t.co/LQyTOju0QL
— Omar Khattab (@lateinteraction) Feb 17, 2024
Now that DSPy is popular, some nuance. DSPy is *not* a prompting framework. It proposes a new paradigm: a framework for expressing & optimizing Language Programs. Think neural networks but with LM "layers", i.e. abstract modules that *learn* prompts/finetunes given a metric.
— Omar Khattab (@lateinteraction) Feb 17, 2024
The beauty of the Milky Way captured in Western Australia is truly out of this world! https://t.co/HZErZ2FXBc
— Figen (@TheFigen_) Apr 8, 2024
Visualizing Attention, a Transformer's Heart https://t.co/XXsUxpFryX
— Suzana Ilić (@suzatweet) Apr 9, 2024
This still blows my mind. Look at what she does when there are no lyrics playing. Now that AI music is almost indistinguishable from real music, what do you think future music videos might look like? https://t.co/RbYpw94ELI
— Halim Alrasihi (@HalimAlrasihi) Apr 9, 2024
The paper "More Agents Is All You Need" When the ensemble size scales up to 15, Llama2-13B achieves comparable accuracy with Llama2-70B 🔥 "The two-phase process begins by feeding the task query, either alone or combined with prompt engineering methods, into LLM agents to… https://t.co/F5Y688cOjA https://t.co/x8H8STLgLE
— Rohan Paul (@rohanpaul_ai) Apr 8, 2024
In our latest episode, @caro_milanesi and I hit on some important angles of the DOJ v. Apple case. Proving consumer harm will be extremely difficult and many don't seem to see the game the DOJ is playing that would set a dangerous precedent. https://t.co/SOmhrMQLkc
— Ben Bajarin (@BenBajarin) Apr 8, 2024
The key message to the W24 @ycombinator batch was: Velocity, velocity, velocity 🚀 But what does velocity really mean, and how can you move as fast as possible? Here are the 10 commandments for shipping product at lightning speed: 🧵 https://t.co/ZKWVIwRMyK
— Chrys Bader (@chrysb) Apr 8, 2024
Mount Fuji Japan 🇬🇱 https://t.co/7p5Hp2x7rp
— Earth (@earthcurated) Apr 7, 2024
Apple's 3B ReALM language model Outperforms GPT-4 on some specific use-cases ✨ 📌 They also found that the performance of ReaLM and GPT-4 are very similar for the unseen domain. Ofcourse ReALM is for a very specific use-case and task and not a generalized model like GPT4 and… https://t.co/vRfPxFwsGL https://t.co/DyuLrcfHyW
— Rohan Paul (@rohanpaul_ai) Apr 8, 2024
How it started: How it's going: https://t.co/s997MtDJGI
— Andrew Curran (@AndrewCurran_) Apr 8, 2024
Just a reminder that a German scientist named Wernher von Braun wrote a book in 1953 called ‘Project Mars’ in which the title of the leader of the Martian government would be ‘Elon’ https://t.co/LOvwvRmwW3
— ALX 🇺🇸 (@alx) Apr 7, 2024
I Built a Stable Diffusion Pipeline to Create Artistic QR Codes using LangChain, DeepLake and ControlNet with Stable Diffusion https://t.co/4vEHb7NxH0
— /MachineLearning (@slashML) Apr 7, 2024
8 MUST read books for anyone interested in #Bitcoin 📚 1) The Creature from Jekyll Island Where does money come from? Where does it go? Who makes it? Start here. https://t.co/LZkwTqqTaJ
— Thomas | heyapollo.com (@thomas_fahrer) Apr 7, 2024
The Top ML Papers of the Week (April 1 - April 7): - SWE-Agent - Mixture-of-Depths - Many-shot Jailbreaking - Visualization-of-Thought - Advancing LLM Reasoning - Representation Finetuning for LMs ...
— DAIR.AI (@dair_ai) Apr 7, 2024
On this day in 1969: a 25-year old grad student published a document about ARPANET. He stayed up all night quietly writing it in a bathroom so he wouldn’t wake people up where he was staying. It helped pave the way for the Internet. https://t.co/ZCsspDqsbB
— Jon Erlichman (@JonErlichman) Apr 7, 2024
Google DeepMind Presents Mixture-of-Depths: Optimizing Transformer Models for Dynamic Resource Allocation and Enhanced Computational Sustainability Quick read: https://t.co/KExZSWZrZ5 Researchers from Google DeepMind, McGill University, and Mila have introduced a groundbreaking… https://t.co/fjFFoM8YiI
— Marktechpost AI Research News ⚡ (@Marktechpost) Apr 6, 2024
Some more details from @EricSteinb about @magiclabs (see why @karpathy @danielgross @natfriedman @polynoamial all invested) - LONG context - more than TENS OF MILLIONS tokens support - Inference time compute idea - Agents with 99.9% accuracy - AGI within a decade thanks… https://t.co/LYGeIoyKRi https://t.co/N9SucIqXr4 https://t.co/58ezLKAmvq
— Alex Volkov (Thursd/AI) (@altryne) Apr 5, 2024
Thanks to @alexalbert__ and the Anthropic team for helping me resolve the rate limit issues that were preventing me from finishing this! https://t.co/q1Tcfa5aDo
— Matt Shumer (@mattshumer_) Apr 5, 2024
Optimization of LLM Systems with DSPy and LangChain/LangSmith The most interesting LLM applications are less a single LLM call and more a complex system that relies on multiple LLM calls. https://t.co/Evm0TBshS2
— Le Dang Trung (@trung_rta) Apr 5, 2024
Adobe Firefly has API for enterprise clients. My colleague at Adobe is writing all kind of scrips / apps with it now. You can read more here, a lot of you have been asking me, so here it goes: https://t.co/tP6nTDyfIk https://t.co/xPoevBvcbF
— Kris Kashtanova (@icreatelife) Apr 5, 2024
Yesterday @AnthropicAI launched a tool use beta! Here's an example of something I built with it: a customer service bot that can actually resolve your issue! Looking forward to seeing what else people build! https://t.co/Xmi7pnwouS https://t.co/T9QDTaqphC https://t.co/T5bE4peexR
— Erik Schluntz (@ErikSchluntz) Apr 5, 2024
RAG for Sales Outreach 🧑💼 In our latest webinar, we highlight a unique sales use case for RAG - instead of writing hard-coded email templates for sales outreach, write a prompt template that can be used by an LLM to generate personalized sales emails, while filling in the input… https://t.co/kV7MGJ6PqS https://t.co/kEAICHJ080
— LlamaIndex 🦙 (@llama_index) Apr 5, 2024
🎉 I updated the LLM Course: - New HuggingChat assistant (cc @victormustar) - New notebooks (AutoQuant, ZeroChat) - New DSPy module - Course flow rework (LLM fundamentals hidden by default) 📚 LLM Course: https://t.co/hsdjIpgJ0R https://t.co/KxhHQGIGHf
— Maxime Labonne (@maximelabonne) Apr 5, 2024
Larry Page on having a healthy disregard for the impossible. https://t.co/OoXWPLsphU
— Jon Erlichman (@JonErlichman) Apr 5, 2024
doesn't speculative decoding basically solve this problem already? (I think we can already run models 3-4x faster in practice by using a draft model and doing speculative decoding) https://t.co/vR0dbEPhtA
— jack morris (@jxmnop) Apr 4, 2024
Adaptive RAG 🧠🔎 Not all questions are created equal. RAG techniques that do well on simple questions will fail on complex multi-step ones; fancy agentic techniques that address the latter will be way slower for simple questions. Adaptive RAG is a cool paper by @SoyeongJeong97… https://t.co/SZQppddC95 https://t.co/hFu7agKPCB
— LlamaIndex 🦙 (@llama_index) Apr 4, 2024
“3 new open source Devin’s” Lmao🤣 love this has become a term for an agent design style https://t.co/0Z1JIPtIbu
— Nick Dobos (@NickADobos) Apr 4, 2024
Function calling support in Claude means the structured data extraction trick I wrote about here now works with the Claude 3 models - including their vision models, notable because GPT-4 Vision doesn't support functions yet https://t.co/ivmOZy1g7O https://t.co/qMrFHWG5yq
— Simon Willison (@simonw) Apr 4, 2024
Kinetic Sand is made from fine sand and polydimethylsiloxane (also called dimethicone). It's an unusual substance in that its viscosity increases under stress. You can form it into fun shapes and slice it. [📹 sand.tagious] https://t.co/Lcec3Ddnts
— Massimo (@Rainmaker1973) Apr 4, 2024
1/n An ontology of Large Language Model (LLM) powered Multi-Agents - Single LLM-based agents have shown promising capabilities such as planning, tool use, memory, and decision making. This has motivated research into multi-agent systems. - LLM-multi agent (LLM-MA) systems aim to… https://t.co/dlYtAiEAzB https://t.co/L60UCmg30V
— Carlos E. Perez (@IntuitMachine) Feb 5, 2024
Looked at all pitches of @ycombinator companies from W24 batch. My observations : - Most founders are accomplished. Be it pedigree, experience or a history of building. YC bets on extraordinary founders. Stanford, MIT, FAANG is a common background. - Many repeat founders + YC… https://t.co/OS07dwNOAe
— Pritika Mehta (@pritika_mehta) Apr 3, 2024
A stable qubit needing only 30 physical qubits? @Microsoft and @QuantinuumQC achieve an 800x improvement in logical vs physical error rate, with both entanglement and error corrections. A real landmark in quantum computing. https://t.co/Ut6GdIKYVK
— Peter Lee (@peteratmsr) Apr 3, 2024
At @cohere, we've been hearing from our customers that Command-R is a game-changer for RAG-optimized workloads. 💪 But to truly maximize the potential of your models, among other things you need to focus on data parsing. 🎯
— vivek raju muppalla (@vim1up) Apr 3, 2024
"I think we’re systems architects and know how to build systems in a very efficient way. So I think this is a product we’ve been training for, for the last decade." — Steve Jobs, about the iPad, 2010
— Steve Jobs Stories (@SJobs_Stories) Apr 3, 2024
Introducing `Claude-Author` 📕✍️ One prompt -> an entire novel! Just describe the high-level details, and a chain of AI systems will write an entire book for you in minutes. - complete w/ cover art - packages your book as a real e-book And it's open-source! https://t.co/rBIpFma3gE https://t.co/56jJzCusEf
— Matt Shumer (@mattshumer_) Apr 3, 2024
I created a Claude 3 AI agent. It learns your writing style, can write anything mimicking it, and improves itself based on a similarity score: https://t.co/RTRVioERT7
— Alvaro Cintas (@dr_cintas) Apr 2, 2024
The long-awaited video on Transformers by @3blue1brown is finally here! Even for those well-versed in Transformers, the visualizations in this video are simply outstanding. https://t.co/qLj35FPgRD https://t.co/VjOxDir1Q0
— Parul Pandey (@pandeyparul) Apr 2, 2024
The famous @ycombinator dropped a very cool article to showcase their 25+ companies who have trained their own AI models, rather than just use someone else's closed model through an API like a black box. ⤵️ https://t.co/LDIZA4AWW5
— Julien Chaumond (@julien_c) Apr 2, 2024