Bipin Suresh

🌉 San Francisco's Surprising Comeback: From 'Doom Loop' to Renaissance? Recent data suggests that San Francisco, once trapped in a pandemic-induced 'doom loop', is showing strong signs of recovery. Here's what's changing: 📈 Positive Trends: - Population Growth: Net gains over the past two years, with domestic in-migration positive for the first time since 2012. - Housing Market: Median home sale prices up 1.2% YoY to $1,427,500 in June. - Job Growth: The Bay Area added 7,400 jobs between January and June. Downtown Revival: Foot traffic beyond typical work hours has recovered to over 90% of pre-pandemic levels. - Crime Reduction: Property crime down 32% and violent crime down 14% in Q1 2024 compared to Q1 2023. - AI Boom: The city is positioning itself as a hub for the AI revolution. 🚧 Ongoing Challenges: - High office vacancy rates - Homelessness and drug crisis - Lagging tourism and business travel 🔑 Key Initiatives: - "Vacant to Vibrant" program offering free rent to businesses in empty storefronts - Increased police patrols and encampment sweeping - Crackdown on open-air drug dealing 💡 Expert Insight: "The rising AI empire is giving the city a new edge. Funding, talents, and culture related to the AI revolution will help the city turn the corner." - Lu Chen, Senior Economist at Moody's Analytics 🤔 What's Your Take? Are these positive trends sustainable, or is it too early to declare a comeback? How can other cities learn from San Francisco's recovery strategies? What role will AI play in reshaping urban economies, particularly in tech hubs? https://lnkd.in/gCnPCBRU

14

1 Comment

AI memory is fundamental for conversational AI and AI agents. It is critical for building personalized experiences and improving reasoning. Check out Letta’s course to understand how to use AI memory to enhance the customization and quality of your AI app.

15

NVIDIA's Jensen said "ChatGPT democratized computing, Llama2 democratized generative AI" ...and now Llama3 is the next level 🔥 It's fascinating to watch Meta's strategic moves. With Llama3, they prioritized training on a massive 15T token dataset to pack all into a lean 70B param model, instead of building a massive model. This allows Llama 3 to match trillion+ parameter models like GPT-4, but at 1/10th the compute, storage and inference costs! 💰 This technique was published by Google DeepMind a few years ago https://lnkd.in/gx7VU8aA Meta is not a dark horse in AI anymore. They might be the top dog. https://lnkd.in/gRRJYuxt #llama2 #llama3 #llm #chatgpt #gpt4 #ai #ml

16

5 Comments

$10B to train an SOTA LLM model. I've been asking fellow Product, Engineering and GTM leaders if SOTA LLMs are delivering positive business outcomes. Answer - > without a value chain of other AI models like ETL, RPA, ASR, TTS, w/ lots of fine tuning and non AI pre-post processing, enterprise customers aren't see enough value to pay more. In fact, they want to pay way LESS to get the same jobs done. This reminds me that not much has changed for product and engineering leaders. We still have to listen to the voice of our customers and translate that into a product that has positive unit economics for the companies we work for. Its very expensive to build and maintain AI solutions. Very expensive to run AI solutions. The right question isn't should I leverage AI in my product, rather if I use AI will it improve the net profit margin. $$$

15

2 Comments

Grammarly just acquired startup Coda (last valued at 1.2bn) with a goal of a possible IPO. This could be a big move in productivity, especially for product teams. Coda CEO Shishir Mehrotra was recently on Lenny's podcast (first comment below). https://lnkd.in/gmgfcm2P

1

1 Comment

Anthropic's latest Claude AI model is taking human-AI interaction to a whole new level! 🖥️🤖 The new Claude 3.5 Sonnet can now use computers just like a human would - by looking at the screen, moving the cursor, clicking buttons, and typing text. This is a major leap forward in AI capabilities. Key features: Operates based on screenshots, emulating human computer use Can perform complex multi-step tasks across different applications Self-corrects and retries when encountering obstacles2 Accessible via API for developers to build with Why it's game-changing: This opens up incredible possibilities for AI assistance in data entry, form filling, software testing, and automating repetitive computer tasks13. Imagine having an AI assistant that can seamlessly work alongside you, handling tedious computer work while you focus on higher-level tasks. While still in beta, this technology demonstrates how AI is rapidly advancing toward more human-like interaction with digital interfaces. As it develops further, it could revolutionize how we work with computers and boost productivity across industries. The future of human-AI collaboration just got a lot more exciting! What do you think about this new capability? #AIInnovation #FutureOfWork Here is a video about it: https://lnkd.in/gh8aNYEu

6

So - are LLMs Universal? This guy looked at a lot of proofs that LLMs are universal and concluded that they are all wrong or misleading. I didn't check all the proofs and I didn't check all the errors or misunderstandings that he mentions. But I would say that it is not confirmed that any realistic model of LLMs is universal. Specifically about https://lnkd.in/da4qHFgh I did not make up my mind if the model is realistic or not, but I think it makes improvement by 1) Getting the outer while loop into the LLM by emphasizing auto-regression 2) Getting the ability to skip parts of the input with the bounded context window, whether a realistic aspect of a theoretical model or not.

4

1 Comment

$2 H100s: How the GPU Bubble Burst Last year, H100s were $8/hr if you could get them. Today, there's 7 different resale markets selling them under $2. What happened? The market has flipped from shortage ($8/hr) to oversupplied ($2/hr), because of reserved compute resales, open model finetuning, and decline in new foundation model co’s. Rent instead. https://lnkd.in/gpQEidmC

2

1 Comment

🚨🚨 🚨 Unlocking Innovation: The Power of the Adjacent Possible Innovation programs can stall when the team tries to ideate way beyond their collective perspective and frame of reference. Like a software engineer brainstorming genetic modification of mycelia, or a senior executive ideating process improvements which they haven't been participated in for decades. It's challenging, exhausting, and ultimately self-defeating. Enter the concept of the "Adjacent Possible" - a game-changer for innovation. What is it? The Adjacent Possible, coined by Stuart Kauffman, represents the realm of new ideas within reach based on our current knowledge, resources, and capabilities. Why it matters: ➡️ Realistic Innovation: it focuses on achievable advancements, not far-fetched concepts. ➡️ Incremental Progress: each innovation opens doors to new possibilities. How to leverage it: • Constantly explore the boundaries of your adjacent possible. What's in your line of sight is what you know best. • Combine existing ideas in novel ways. Even small perturbations from status quo can generate large innovation leaps. • Stay curious about developments in adjacent industries. Cross-domain patterns create fresh sources of ideas, while also expanding the adjacent possible. Most of all, it takes the 'genius' pressure off the team. No one knows everything, but each one of us knows a lot about something, and we can all drive plenty of innovation potential by focusing on the spaces we know best. Are you exploring your Adjacent Possible? Share your thoughts below! 👇 #Innovation #innovationsystems #AdjacentPossible

8

2 Comments

THERE ARE THREE R'S IN STRABERRY 🍓🍓🍓 Looking through OpenAI's new o1 launch today, it got me thinking about Daniel Kahneman's book, Thinking, Fast and Slow 📚. In it, Kahneman explains that our brains operate using two systems of thought. System 1 is fast, automatic, and intuitive—it handles everyday decisions without us even realizing it 🤔. System 2, on the other hand, is slow, deliberate, and analytical 🧠. It's what we engage when we're solving complex problems or learning something new. This dual-process theory highlights how we balance efficiency and thoroughness in how we think. By the way, did you catch the typo in "STRABERRY"? If not, that's System 1 at work—glossing over details to get the gist quickly. It takes System 2 to slow down and spot the error 🔎 OpenAI's o1 feels like a significant step towards embracing that System 2 thinking — more deliberate and analytical, targeting high-complexity (or high-risk) scenarios. Meanwhile, the current GPT models are akin to System 1: fast, intuitive, but at times superficial or even inaccurate. Both systems are necessary—in our brains and in our quest towards AGI. In future, it's not hard to imagine that you won't have to pick the right model to use - based on the prompt the AI will pick the right system to use to optimize the energy and time - just like our brain does. Exciting times! 🚀🌐 P.S. – The title (except the typo) and the screenshot come from the o1 launch examples, which I thought was very witty of them given all these AI strawberry memes. #OpenAI #o1 #ThinkingFastAndSlow #AI #AGI #Innovation

35

3 Comments

Just over a year in and GPT4 is now 12x cheaper, faster, better! We're most excited about GPT-4o – for complex reasoning tasks, and Gemini 1.5 Flash – for data transformations, especially given its long context length. As chips get faster, model architectures improve, and we have more data, it feel inevitable that the cost of intelligence (as a service) will keep going down -- just like what happened for compute and storage costs in the last decades.

77

2 Comments

Gemini 2.0 handles complex math and planning tasks previously out of reach of LLMs. You can turn on "code generation" in AI Studio and in the APIs. For context, let's talk about function calling and voice AI agents. (I spend a lot of time working on voice AI.) Voice AI agents use function calls for: - Interacting with external systems specific to particular use cases and deployment contexts. - A wide range of RAG-like information lookup tasks. - Conversation persistence, including complex actions like saving a privacy-redacted version of a conversation. - Script following. Some of these uses were obvious before we had built and deployed production voice AI systems. But some of them weren't. (Script following, for example.) We had to build experiments, and then production systems, to figure out what features and tooling these new systems really need. I'm convinced that code generation will be as important as function calling, once we've accumulated some experience building with Gemini 2.0 and similarly capable new LLM versions. Credit for this insight goes entirely to Shrestha Basu Mallick, who shepherds work at Google on the Gemini APIs and AI Studio. When Shrestha first told me that really good code generation was going to make a lot of new things possible, I didn't really get it. But after playing with code generation for a few weeks ... I still don't fully get it, but in a good way, now. It's clear that there are some obvious things to explore: - Context-aware math on demand driven by natural language. - Generative UI. - Better "binding" behavior with function calling. But I'm also convinced now that there is a lot of unexplored, non-obvious (at least to me, yet) territory. Let's take that last one — improved function calling performance. Function calling and code generation go together like peanut butter and jelly. Functions often take arguments that are fairly complicated. With code generation enabled, Gemini has access to this new pattern: natural language ➡️ code that creates argument structs for functions ➡️ multiple function calls Think about how powerful this is. The ability to use even very basic code structures — loops, for example — massively expands what LLMs can do. In fact, my impression is that function calling in Gemini 2.0 is now trained as a subset of the model's code generation capabilities. This feels quite different than how function calling works in other models that I have access to. In other models, function calling feels like a close cousin to structured output. Which is great. But this new "modality" is even better. (More on the prompt in this post, and the meta-prompt for this post, here: https://lnkd.in/gycKtnNX)

49

10 Comments

How are the best Large Language Models ranked? Large Language Models (LLMs) have opened up incredible new possibilities, but ensuring they align with human preferences remains a challenge. That's where Chatbot Arena comes in. Managed by a team at UC Berkeley, this crowdsourced open platform evaluates LLMs based on human preferences. Chatbot Arena ranks different language models using ELO ratings, similar to the ranking system in chess. Users can visit the platform, enter a question, and receive responses from two anonymous models. By selecting the preferred response, ELO scores are calculated based on win rates, with higher scores indicating better performance. Currently, the proprietary models, like the GPT models from OpenAI, and the Claude series from Anthropic and Gemini from Google, dominate the top rankings. This highlights that open-source models have yet to catch up to their closed counterparts. Chatbot Arena has already become one of the most widely cited LLM leaderboards, referenced by top LLM developers and companies. Chatbot arena is dependent on community participation, so you can always contribute by casting your vote.

12

1 Comment

💥 Finnt is now publicly available! 🌎 Building an AI-native software is a whole new beast - actually that’s why there’s so many great demos but so few amazing experiences in production. The nature of Generative AI makes it difficult to reproduce identical, deterministic outcomes. Where a traditional feature would have two statuses “functional” vs. “bugged”; in the GenAI world, a slightly different prompt or setup can drastically change the output. But this is also what makes AI so powerful and why we are so excited to be building in this space. 🧠 As a reminder, Finnt specializes in extracting crucial data from large, unstructured, and complex documents — including OMs, PPMs, 10-Q/10-K, as well as prospectuses, reports, leases, loans, etc. It then integrates these findings into AI-generated and templated memos, enriching them with real-time market data. 🔑 Key facts: - Ingest any large, complex, and unstructured files - Design custom frameworks tailored towards your specific use cases - Select your preferred LLMs, from closed to open-source options - Customize the goals, structures, and formats of your output reports - Augment your reports with third-party sources and live market data - Ask anything to your files and transform responses into insights - Seamlessly integrate Finnt into your existing workflows 👉 Check the comments to get the link to signup. PS: you might need to get a custom onboarding but we are nice people 😅

179

47 Comments

Last week, I posted about how AGI is on the brink of transforming everything, but Yann LeCun, Chief AI Scientist at Meta, has another point of view. In a recent interview, he explained why today’s AI, while impressive, is nowhere near human intelligence and likely won’t be for decades. Even cats have a mental model of the physical world, persistent memory, and some reasoning and planning abilities — none of which the most advanced AI models possess. To create truly intelligent systems, we must build AI that behaves more like a living assistant – navigating the world and making decisions independently. But we’re not there yet; getting there will take decades, not years. LeCun also challenges the belief that scaling up current AI models is the solution. Adding more GPUs and making models larger won’t magically lead to human-like intelligence. The problem isn’t size; it’s architecture. At Meta, researchers are exploring alternative approaches, like training AI on visual data to help it build an understanding of the world – similar to how animals learn. This shift in focus could be the key to creating AI that doesn’t just process information but truly “understands” it. LeCun is skeptical about large language models like ChatGPT. While undeniably impressive, he doesn’t see them as a path to true intelligence. He said, “We’re used to equating language skills with intelligence, but that’s not true. You can manipulate language without being smart – and that’s exactly what LLMs do.” What do you think? Are we overestimating the potential of today’s AI?

18

10 Comments

LLMs are terrible at basic math. Why? Here's why: Unlike people, LLMs don’t conceptualize mathematics. They simply recognize patterns in the training data (text) they’ve learned from. When asked what 2 + 2 is, an LLM recalls instances of seeing ‘2’, ‘+’, ‘2’ and so on, allowing it to approximate solutions but not match the precision of a calculator. That’s how we end up with LLMs asserting that 2 + 2 = 5. But for analytical teams, like finance, that makes LLMs really hard to adopt in day-to-day workflows. The productivity gains of having AI do calculations are negated by needing to fact-check and audit every calculation and number. LLMs are also not deterministic, so no amount of prompting is going to guarantee you’ll get the correct answer every time. And that’s just for basic math – the outputs I’ve seen from LLMs on more complex calculations and analyses can get very wonky. Add thousands of rows of data and you’re squarely in hallucination land. So what’s the solution? Calculate things manually? No. The good news is that this is changing quickly and there are options: 1 - AI is getting better at math. Researchers continue to train LLMs with more specialized datasets that include a much wider array of mathematical problems and solutions. The newest models, like OpenAI’s latest family of models, are incorporating things like symbolic reasoning and “chain of thought” methods to break complex problems into smaller, logical steps the LLM can more easily process. Results have been really promising and we’ve seen each new iteration of models continue to beat previous benchmarks. 2 - Leverage AI agents For intensive workflows, and to really make sure your outputs are correct, working with companies that are building high-performing AI agents is likely your best bet. Agentic workflows allow LLMs to be used for what they are great at (inferring user intent, making decisions, etc.) while outsourcing the things they struggle with (complex calculations) to other parts of a broader system. Text-to-SQL, “tool-use”, few-shot prompting, and other strategies can be combined with real-time data integrations to dramatically improve performance on analytical operations. At Concourse we’ve spent the better half of 2024 getting our AI Agents to be really good at math, and I’m so excited to continue growing our capabilities to tackle even more complex mathematical problems for our customers in 2025. Reach out if you’re interested in learning more about how we’ve built Concourse and the impact we’re already delivering for finance teams.

42

5 Comments

Reimagining Healthcare Workforce Platfforms: Turning Capital Into Labor The future isn’t just about filling shifts. It’s about scaling labor intelligently, driving better outcomes for systems, staff, and patients alike. Andreessen Horowitz and Alastair (Alex) Rampell reinforced this Vision in a recent post and it applies to the next phase of healthcare workforce evolution. Healthcare workforce marketplaces plus AI are at the tipping point of transformation, evolving into systems of record that don’t just manage labor—they optimize it. Much like EHRs centralized patient care data, these marketplaces can centralize workforce operations, aligning talent supply with clinical demand. But there’s a bigger opportunity here: turning capital into labor. Here’s what I mean. Historically, industries (including healthcare) have focused on enterprise computing tasks—streamlining data entry, compliance, and back-office functions. While valuable, this approach often stops short of transforming the real-world, repetitive processes that make up the bulk of work in healthcare. Think beyond enterprise software. The next frontier lies in leveraging embodied AI and workforce marketplaces to automate and enhance labor-driven processes, transforming capital investments into productive labor outputs. For healthcare, this means turning staffing costs into scalable, dynamic systems that deliver value far beyond filling shifts: 1. Dynamic Labor Matching: Capital invested in marketplace platforms enables real-time matching of labor supply with patient demand, replacing outdated scheduling models with predictive, automated systems. 2. Workflow Automation: From credentialing to compliance and shift bidding, embodied AI can streamline traditionally manual tasks, freeing human resources for higher-value work. 3. Labor as a Strategic Asset: Workforce marketplaces shift the focus from static staffing budgets to proactive labor deployment, ensuring that every dollar invested in labor drives outcomes and efficiency. It is about the billions of repetitive, planned tasks that AI can improve. In healthcare, that means automating everything from staffing logistics to clinical task assignments, creating a horizontal system of engagement for labor (think Zendesk) that integrates seamlessly with vertical, workflow-specific platforms (think Toast for staffing operations). The result? Hundreds of billions in untapped market potential—and a healthcare workforce model that transforms how care is delivered. By turning capital into labor through smarter marketplaces and automation, we don’t just solve staffing challenges; we redefine what’s possible in healthcare operations and ShiftMed is leading the way with health systems moving beyond tired models. #capitalintolabor #systemsofrecord #healthworkforce #aiworkforce #healthsystems #hospitals #cno #hospitalceo #healthcareinnovation

11

2 Comments

PSA: Stop using the original OpenAI CLIP weights. I see a lot of people using CLIP for multi-modal retrieval. What many people may not know is that there are 100’s of CLIP models now since OpenAI’s original release. These newer models offer significantly better performance across many benchmarks. For product search related tasks we recommend the following. + Fastest inference: ViT-B-32 [laion2b_s34b_b79k] + Best balanced: ViT-L-14 [laion2b_s32b_b82k] + Best-all round: xlm-roberta-large-ViT-H-14 [frozen_laion5b_s13b_b90k] You might notice these are not the ones that perform the best on the benchmarks. Why? Because the benchmark only test for very specific tasks which are often quite different from search and recommendations. These are ones we have found to be the best for product related search and retrieval tasks and generally very good first choices if you are unsure where to start (even for non-product related search). Here is a hugging face space with all the data: https://lnkd.in/gxu5UYkt Read more: https://lnkd.in/g5_D4ccs

293

22 Comments