Loading video...

Video Failed to Load

Go Home

5/ Synthesis NLP data pipeline + classifier and LLM summarization tool to collate news from biased sources and distill truthful information Basically, Ground News, but AI 🏅 Editor’s (me) choice- for most sophisticated architecture Ron Nachum

225,984 views • 2 years ago •via X (Twitter)

15 Comments

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

Stanford just hosted a hackathon. Over 1000 students from around the world came to build for 36 hours straight. The reward? $100k+ in prizes. Here are the winners and crowd standouts we saw at TreeHacks ‘24 @hackwithtrees (🧵):

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

1/ Baymax Robot arm controlled by natural language and speech to provide medicine to the elderly/physically disabled. Every single robot control, including inverse kinematics, coded from scratch. 🏆Grand prize winner- $10,000 🏆Elderly citizens award- Free trip to Japan

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

2/ Recollect Automatic OCR scanning + page turning machine to digitize books and print media 🏅Hardware award @LawtonSkaling @kaien_yang @jclin22009

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

3/ Good Samaritan Combining Apple Vision Pro with machine learning to help first responders administer first aid 🏆Healthcare impact grand prize @yashdulla @rayhotate @ShloakR @blu3mo

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

4/ Smart bike lock RFID bike lock jerry-rigged together with 9 volt batteries, electrical tape, a 3d printed chassis, and an arruino @SamDuong1

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

6/ Luna AI storyteller that live-generates photos while you’re telling a story Great for kids + fun novelty party game @Kellyyyeh @sfkunals

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

7/ Lingua Franca VR language learning game where you can learn new language by speaking with NPCs 🏅Prize winner

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

8/ Compost Chaos Overcooked-inspired Unity game to promote recycling @LuizAhumada

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

9/ Auto switch Cheap door switch with built in Bluetooth and motion sensors. Retrofits to lights by physically toggling the switch

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

10/ Breathe easy “Breathing” pillow that expands and contracts to match your own breathing pattern. Research suggests this leads to more restful sleep Inside: mechanical accordion. No need for wires or pumps

Alex Reibman 🖇️'s profile picture
Alex Reibman 🖇️2 years ago

And that’s all I could film! There were wayyy too many projects (at least 400) This was the first student hackathon I’ve ever been to. I’m seriously impressed Huge shoutout to @MLHacks @hackwithtrees @Stanford Want to see more hacks, AI agents, and demos? Follow @AlexReibman

Pranav Ramesh's profile picture
Pranav Ramesh2 years ago

@Ground_app @RonNachum Thanks, Alex! This was a cause that's close to each of us, and Ground News was a huge inspiration for us. We're excited to pursue this further!

Ground News's profile picture
Ground News2 years ago

@RonNachum Glad we could be an inspiration! We've recently launched GAIA, a feature very similair to what you've described. It prompt's readers with questions to ask about the story so readers can dive deeper and learn more. Try it out here:

Nándor Hulverscheidt's profile picture
Nándor Hulverscheidt2 years ago

@Ground_app @RonNachum Thanks a lot for the thread, great work by the students! But the example for the LLM news synthesizer is unfortunately well chosen to show the risks of such a system, as Trump was NOT fined for 350 AND 450 million. Afaik it should state 350 for fraud and 80 for Carrolls case.

blockboy's profile picture
blockboy2 years ago

@Ground_app @RonNachum @q

Related Videos

Agentic AI will transform every enterprise–but only if agents are trusted experts. The key: Evaluation & tuning on specialized, expert data. I’m excited to announce two new products to support this–Snorkel AI Evaluate & Expert Data-as-a-Service–along w/ our $100M Series D! --- Snorkel Evaluate is our new data-centric agentic AI evaluation platform for specialized, mission-critical enterprise settings where vibe checks and out-of-the-box metrics driven by simple LLM prompts are not enough. Snorkel Expert Data-as-a-Service is our white glove service for expert-level AI datasets, powering frontier LLM developers in areas like expert knowledge, reasoning, agentic action and tool use, and more! Both built on top of Snorkel AI’s Data Development Platform, using our programmatic technology to drive higher-quality expert data, faster– for getting specialized AI to real production value. If you’re building enterprise AI and want to partner around the key ingredient in AI today–the data–book a demo and let's talk! Finally, see thread for details on 🧵👇 - 📽️ A walkthrough of Snorkel Evaluate and Expert Data-as-a-Service on an agentic AI enterprise task - 📅 An upcoming event on Enterprise Agentic AI with innovators from Accenture @BNY Comcast Stanford University QBE & others - 📊 An upcoming series of benchmark datasets and model artifact releases 👀 Want early access to the full agentic AI dataset? Retweet this post and we'll send you the link!

Alex Ratner

49,924 views • 1 year ago

New course: MCP: Build Rich-Context AI Apps with Anthropic. Learn to build AI apps that access tools, data, and prompts using the Model Context Protocol in this short course, created in partnership with Anthropic Anthropic and taught by Elie Schoppik Elie Schoppik, its Head of Technical Education. Connecting AI applications to external systems that bring rich context to LLM-based applications has often meant writing custom integrations for each use case. MCP is an open protocol that standardizes how LLMs access tools, data, and prompts from external sources, and simplifies how you provide context to your LLM-based applications. For example, you can provide context via third-party tools that let your LLM make API calls to search the web, access data from local docs, retrieve code from a GitHub repo, and so on. MCP, developed by Anthropic, is based on a client-server architecture that defines the communication details between an MCP client, hosted inside the AI application, and an MCP server that exposes tools, resources, and prompt templates. The server can be a subprocess launched by the client that runs locally or an independent process running remotely. In this hands-on course, you'll learn the core architecture behind MCP. You’ll create an MCP-compatible chatbot, build and deploy an MCP server, and connect the chatbot to your MCP server and other open-source servers. Here’s what you’ll do: - Understand why MCP makes AI development less fragmented and standardizes connections between AI applications and external data sources - Learn the core components of the client-server architecture of MCP and the underlying communication mechanism - Build a chatbot with custom tools for searching academic papers, and transform it into an MCP-compatible application - Build a local MCP server that exposes tools, resources, and prompt templates using FastMCP, and test it using MCP Inspector - Create an MCP client inside your chatbot to dynamically connect to your server - Connect your chatbot to reference servers built by Anthropic’s MCP team, such as filesystem, which implements filesystem operations, and fetch, which extracts contents from the web as markdown - Configure Claude Desktop to connect to your server and others, and explore how it abstracts away the low-level logic of MCP clients - Deploy your MCP server remotely and test it with the Inspector or other MCP-compatible applications - Learn about the roadmap for future MCP development, such as multi-agent architecture, MCP registry API, server discovery, authorization, and authentication MCP is an exciting and important technology that lets you build rich-context AI applications that connect to a growing ecosystem of MCP servers, with minimal integration work. Please sign up here!

Andrew Ng

142,010 views • 1 year ago

Traditional data pipelines don't work for RAG applications. There are 3 issues with them: ​ 1. Traditional data engineering solutions are optimized to handle structured data. RAG applications rely primarily on unstructured data. ​ 2. The connector ecosystem to load data from unstructured data sources is very immature. ​ 3. Traditional solutions do not offer any way to transform unstructured data into an optimized vector search index. ​ The goal of a RAG Pipeline is to solve these problems. ​ The number one objective is to create a reliable vector search index using factual knowledge and relevant context. This sounds easy, but it's one of the biggest challenges we face when building RAG applications. ​ At a high level, there are four different stages in the architecture of a RAG pipeline: ​ 1. Ingestion: Here is where the pipeline loads the information from the data source. ​ 2. Extraction: Where the pipeline processes the input data and decides how to retrieve the text contained inside them. ​ 3. Transform: Where the pipeline chunks the data and generates document embeddings. ​ 4. Load: Where the pipeline creates a search index in a vector database and loads the document embeddings. ​ There are different rabbit holes at each one of these stages. Here are three of them: ​ 1. Ingesting data once is simple. The hard part is refreshing the vector database whenever the original data source changes. ​ 2. Extracting the content of a plain text document is simple. The hard part is to extract content from complex documents containing tables, images, or cross-references. ​ 3. A simple continual chunking strategy with an overlap is simple. The hard part is to find the optimal strategy for your specific knowledge base and the way you are planning to query it. ​ In the attached video, I'll show you how you can build an enterprise-grade RAG Pipeline that solves every one of the above problems. ​ I'll use Vectorize. They partnered with me on this post. You can use them to build RAG pipelines optimized for accurate context retrieval. ​ ​ If you have a few documents lying around, set up a free account and give it a try.

Santiago

40,441 views • 1 year ago

Debunking Coffeezilla’s prediction market claims In his recent video “prediction markets aren’t just gambling,” Coffeezilla made the three untrue or misleading claims below: 1) “The only way you get the news early, by the way, is if it’s insider trading.” 2) “All these prediction markets are doing is aggregating sentiment on the news.” 3) “The only way you can get something not in the news from these markets is if someone with non-news information trades, which is AKA inside information.” I want to start by saying that I’ve really enjoyed Coffeezilla’s content in the past, and I appreciate him blacking out my name in the tweet he screenshared. I also watched the full video and agreed with parts of it (and disagreed with other parts). That said, these three claims are egregiously incorrect, and I want to correct the record. I hope Coffeezilla reads this and reconsiders these points 1) “The only way you get the news early, by the way, is if it’s insider trading.” Merriam-Webster defines insider trading as “the illegal use of information available only to insiders in order to make a profit in financial trading.” This claim is wrong because it is clearly possible to get information early in entirely legal ways. For example, in CPI inflation markets, someone might notice prices rising on goods they regularly buy, or a sophisticated trader might aggregate pricing data across many products and form a forecast before the CPI release. Journalists may later report on inflation, but the information existed beforehand. Markets also react faster to sudden events, like a Trump Truth Social post or an earthquake, than journalists do. Traders are financially incentivized to react in seconds; journalists are not. Recent high-profile examples include Nobel Peace Prize, Spotify, and Time POTY markets, where traders had information before the news broke. In the Nobel case, there was disinformation claiming insider trading, but as far as most observers can tell, the information was obtained legally via web-scraping. Even if the Nobel Committee disliked it, legally obtained information is fair game. 2) “All these prediction markets are doing is aggregating sentiment on the news.” This is easy to debunk. Prediction markets do aggregate information, but not merely sentiment or headlines. That’s likely why CNN and CNBC partnered with Kalshi. News is filtered through editors, incentives, and bias. Taking headlines at face value is not a winning trading strategy. Savvy traders treat news as one input among many variables. If a headline says “Poll X shows Clinton up 10 points,” markets may adjust, but they don’t blindly price the headline. They factor in other variables. I’d argue markets are often smarter than the news. Domer❤️‍🔥 has even argued that Fed markets on Kalshi are more accurate than CME due to traders like himself making them more efficient, and I think he’s right. 3) “The only way you can get something not in the news from these markets is if someone with non-news information trades, which is AKA inside information.” Merriam-Webster defines insider information as “information not known to the public that one has obtained by virtue of being an insider.” You can obtain non-news information without being an insider. This overlaps with point one, but here’s a concrete example. For the recent TN-07 special election, I traveled to TN-07 and spoke with voters leaving early-voting sites and with everyday residents. I learned how little awareness there was that a special election was even happening, and how voters were thinking about the race. That information wasn’t in the news, but it informed how I traded. I’m not an insider. This was “alpha hunting” through firsthand observation. Almost every serious prediction market trader has similar stories. This is certainly not "insider trading." I’m genuinely curious to hear your thoughts, Coffeezilla, and hope for a good-faith dialogue. I hope you are doing well!

Benjamin Freeman

80,021 views • 7 months ago

Announcing a new Coursera course: Retrieval Augmented Generation (RAG) You'll learn to build high performance, production-ready RAG systems in this hands-on, in-depth course created by and taught by Zain, experienced AI and ML engineer, researcher, and educator. RAG is a critical component today of many LLM-based applications in customer support, internal company Q&A systems, even many of the leading chatbots that use web search to answer your questions. This course teaches you in-depth how to make RAG work well. LLMs can produce generic or outdated responses, especially when asked specialized questions not covered in its training data. RAG is the most widely used technique for addressing this. It brings in data from new data sources, such as internal documents or recent news, to give the LLM the relevant context to private, recent, or specialized information. This lets it generate more grounded and accurate responses. In this course, you’ll learn to design and implement every part of a RAG system, from retrievers to vector databases to generation to evals. You’ll learn about the fundamental principles behind RAG and how to optimize it at both the component and whole-system levels. As AI evolves, RAG is evolving too. New models can handle longer context windows, reason more effectively, and can be parts of complex agentic workflows. One exciting growth area is Agentic RAG, in which an AI agent at runtime (rather than it being hardcoded at development time) autonomously decides what data to retrieve, and when/how to go deeper. Even with this evolution, access to high-quality data at runtime is essential, which is why RAG is a key part of so many applications. You'll learn via hands-on experiences to: - Build a RAG system with retrieval and prompt augmentation - Compare retrieval methods like BM25, semantic search, and Reciprocal Rank Fusion - Chunk, index, and retrieve documents using a Weaviate vector database and a news dataset - Develop a chatbot, using open-source LLMs hosted by Together AI, for a fictional store that answers product and FAQ questions - Use evals to drive improving reliability, and incorporate multi-modal data RAG is an important foundational technique. Become good at it through this course! Please sign up here:

Andrew Ng

124,458 views • 1 year ago

Meiselas: As we hit 5 million subscribers, I want to say this: this is one small step for the MeidasTouch Network, but I think one giant leap for independent journalism and democracy generally. What we set out to build was an unapologetically pro-democracy, independent news network—to get out accurate, timely news that doesn’t “both sides” what’s taking place, but instead confronts head-on the threats we face and offers solutions. And now, the little engine that could—the one people said could never happen—the MeidasTouch Network has hit 5 million subscribers, 33 million views every 48 hours, more views than Fox News and any cable news network. Fueled by the movement, the Meidas Mighty, we are so grateful to share this moment with you all. Five million subscribers—and from here, we continue to grow. We grow the MeidasTouch Network. We help grow the other networks affiliated with MeidasTouch. And we use our platform to help others entering the independent news space. We use our platform to support organizers on the ground—like the organizers who made the No Kings protest possible. And we use our platform for good: to always shine a light on the truth, to expose the injustices taking place, and to confront the authoritarian Trump regime head-on, where corporate news has utterly failed and betrayed us all. Meidas Mighty, we thank you from the bottom of our hearts for this shared accomplishment of reaching 5 million subscribers.

Acyn

218,033 views • 1 year ago