Eunji Noh

Looking forward to my second #ProductTO, I am excited to go back as a facilitator this time, discussing one of my favourite topics: #AI and #ProductManagement, obviously we will scope it down to a specific topic! Here is my experience from last month's event: https://lnkd.in/g_V7kb3Y I highly recommend joining this community! The event format is different, allowing meaningful discussions and connection building! #ProductManagement #Canada #Toronto #Learning

8

Unleashing the Power of Geospatial Analytics with Snowflake! 🗺️ A must-see demo from Volodymyr Bilonenko 🇺🇦🇪🇺 on how Snowflake's geospatial features are transforming how Vay visualizes and analyzes geographic data. Whether you're working in GIS, logistics, or retail, Snowflake's scalable, cloud-native platform is a game-changer for processing complex geospatial data at lightning speed. From mapping to real-time data visualization, Snowflake is helping businesses make smarter, location-driven decisions. 💡

53

2 Comments

LLMs are not vast, infallible databases. As models grow larger so does their tendency to confidently state falsehoods (hallucinations). Matei Zaharia, co-founder & CTO of Databricks, warns against treating LLMs as internet-scale encyclopedias and proposes we focus on what they excel at:  → Reasoning → Interpreting language → Understanding context For founders building AI applications, treat LLMs as powerful reasoning engines and not fact repositories. Build your knowledge base separately, feeding the LLM with trusted up-to-date information through retrieval mechanisms and tool-calling. This approach lets you harness the LLM's strengths in language analysis and generation while sidestepping the risks of its unreliable memory. Building with AI? Share your approach to balancing LLM reasoning power with factual accuracy in the comments. Let's hear it!

137

22 Comments

👉 Howdy LinkedIn! Tomorrow, on "Building Products with Generative AI" - I will be hosting 🚘 Nitin Kartik at 11:30AM CST/12:30 PM EST! We're going to be digging into the intersection of Product Marketing & Generative AI, as well as other topics that will branch from Nitin's experience in the space. I'm excited for another interesting episode with a great thought leader in the space - we hope to see you join for tomorrow's episode! #GenerativeAI #AI #ProductMarketing

15

2 Comments

Entering year three of the post-AI era, and here's my prediction: Enterprises are about to jump headfirst into custom model training and fine-tuning. Generic AI models are great for consumers and some prosumers. But thinking they are suited for industry-specific use cases is a bit.. optimistic. These models weren't built for that and definitely haven't been trained on your vast, private enterprise data. There's a growing gulf between data inside enterprise silos and the data generic AI models are trained on—a gap that's only widening. It's basically the public internet vs. your company's private data. Companies aren't keen on giving up their moat and all their data. So, custom model training and fine-tuning on hardware **you** control is set to become the new norm. Simply deploy the trained, proprietary model to your secure cloud or even on-premises.

18

2 Comments

Part II of our blog is out! In this installment Abde Tambawala and I explored usage metrics and pricing models across different AI modalities and discussed how traditional User-based subscriptions will need to be supplemented with usage metrics. Take a look and let us know what you think! #pricingstrategy #ai #monetization

27

2 Comments

There’s a lot of value to be unlocked by teaching AI to reason within the enterprise context and doing the last mile work. The ‘wrapper on top of gpt’equivalent critique for saas was ‘ui layer on top of databases’. We know how that ended. “Application layer AI companies are not just UIs on top of a foundation model. Far from it. They have sophisticated cognitive architectures that typically include multiple foundation models with some sort of routing mechanism on top, vector and/or graph databases for RAG, guardrails to ensure compliance, and application logic that mimics the way a human might think about reasoning through a workflow.” #ai #aiagents

14

I wrote a blog post on a thorny but underappreciated problem in ML-based marketing personalization: event attribution. It's how you go from the actions you take and customer behavior you observe to creating training experiences for your model. Good attribution makes a huge difference in driving uplift, there are a lot of degrees of freedom in configuring it, and at the same time, it's rarely discussed in ML/bandits/RL literature.

45

4 Comments

Signal boosting this post from my colleague Arun Shankar. My snarky summary "It's like caching, but semantic" :P Oddly, I think it's a reasonable summary... semantically similar inputs hit the same cache, otherwise dissimilar input miss.... ... but how do you do that on non-exact matches and how to you manage such a system? I guess you'll have to read: https://lnkd.in/gPgtM8T5

13

1 Comment

More re: ChatGPT: During the last week, I've fed thousands of texts into ChatGPT with the same prompt and browsed through at least hundreds of results. Fun fact I've noticed: if you zoom waaaay out of a single response and start looking at patterns across multiple responses, you'll notice all the responses falling into a small handful of structures. For example, I'd see some documents using 4 bullet points and some others using, say, 3 paragraphs. It's as if the distribution of the responses has a handful of modes, and all the responses just fall into one of those modes. Which gets me thinking: it's not just a "next word predictor" as many people say, it's more than that, but not too much more. It's as if it has memorized a few templates, and for each response it just picks a template and fills it out word by word. That also helps answer another question I was wondering about recently: if an LLM is just a "next word predictor", how can it generate songs that rhyme? If you think about this, it should be pretty hard to do when you're just spitting out one word at a time. But if you have a certain structure in mind ahead of time, then it gets much easier. And here's another thought: I have a huge suspicion that those response structures - 4 bullet points, 3 paragraphs, yada yada - occur a lot in the training data fed into RLHF. Basically data either written or curated by those huge crowds of low wage workers "training" the next GPT. And if so, here's the thing: this is what makes LLMs ultimately not creative. They just memorize structures and regurgitate them. It's just the ability to memorize a handful of structures that suddenly makes people ascribe all sorts of "intelligence" to those things.

12

3 Comments

🚀 We've open-sourced TimesFM – a decoder-only foundation model for time-series forecasting 📈 Key Highlights: ✅ Large-Scale Training Data: Trained on a corpus of over 100 billion real-world time points, including Google Trends and Wikipedia pageviews. ✅ Zero-Shot Forecasting: Offers high-quality forecasts even on unseen datasets, without requiring fine-tuning or additional training. ✅ Outstanding Benchmark Performance: Achieved competitive results across a comprehensive suite of benchmarks. 🤗 Explore the model on Hugging Face and GitHub. I'm encouraged by the early feedback and look forward to seeing how #TimesFM will be used in real-world applications for forecasting and anomaly detection. 🙏 A special thanks to the talented research team behind this achievement: Abhimanyu Das Rajat Sen Yichen Zhou Weihao Kong

44

1 Comment

In the next few weeks, Snowflake and Databricks will host their annual summits - back-to-back - in San Francisco. The burden on the larger tech community is real: relationships will be tested. Dogs will go unwalked. Extended childcare will be arranged. Strict diets will be neglected, exercise foregone. Why are we doing this to ourselves? Because in 2024, these ecosystems are THE most important growth drivers. These data platforms largely dictate the fortunes of anyone in their orbit. No one is more trusted than the data platform. No technology is more central to more mission-critical workloads. As a result, no one is more influential in driving technology decisions. A primary data platform implies a primary ecosystem. And a primary ecosystem’s primary event is a big deal. At these events, seismic announcements will be made. Feature announcements that reshape reality for partners. Joint announcements with partners that drive critical awareness and pipeline. In parallel, decision makers will be exposed to new apps. New data providers. New changes in the SI landscape. New preferred SaaS tools for each workload (e.g. analytics, ETL, machine learning, dev ops, martech). For partners, these events are a pivotal time. They’ve spent the year building up the two critical aspects of ecosystem credibility: quality and volume of work. Now is the time to reap what has been sewn, and to plant for the forthcoming season. Year over year, outcomes from these events can and should improve. For those pursuing ecosystem-led growth - and, in 2024, it should be EVERYONE - navigating these events is an instructive microcosm for broader ecosystem participation: look for gaps within the platform’s ecosystem. Win with initial projects. Continue to establish yourself in the ecosystem with more marquis projects. Promote successes effectively within the ecosystem. And, importantly, don't bite off more than you can chew. Be disciplined and measured. Credibility within the ecosystem is built over years, and can be lost in a day. Personally, I’m very much looking forward to seeing all of you at both of them. Maybe even at the dingy hotel gym this year. I swear I’m gonna make it this time… #snowflake #databricks #ecosystem #data

55

7 Comments

🚨📈 Data nerds – your mission, should you choose to accept it… …it to find a reliable, reasonably exhaustive source for *literally rates* across countries. I thought I was reasonably versed in the art of combing through supranational organizations data, but I was successively let down by the World Bank, the IMF, ILO, UNESCO, OCDE, WEF. We cannot fix what we cannot measure. It’s concerning that none of the major institutions concerned with human and economic development seem to gather data to track something as basic as literacy, a cornerstone of economic prosperity and human progress. How do we fix this?

5

2 Comments

Large Language Models (LLMs) as super lossy compression. I have been thinking a lot about demystifying what LLMs do because lots of y'all's are very mystified. In the linked article, I try to reframe them for you, hoping it will help you understand what they good for and what they are not good for. Side note: I am still looking for my next adventure. If communicating complicated concepts so that regular people can grasp them is something you need, you should message me. I write code and stuff too and am usually fun to be around. #WrittenByMe

6

3 Comments

It's been a busy week for new model releases! OpenAI and Google had big announcements, and we saw some new open source models released as well. Trends The top models are starting to converge, asymptotically approaching an upper limit on core capabilities. This is driving predictable product decisions in top tier proprietary orgs. Pricing But prices for a given unit of 'cognitive output' are declining relatively rapidly. OpenAI's newest model, GPT-4o, is at least as good as Anthropic's best model Opus. But at ~1/4 the price of Opus, it's priced competitively with Sonnet, Anthropic's second tier model. Differentiation Not a lot... Among the top models, switching costs are relatively low and for many use cases the outputs are relatively interchangeable. Open source Still second tier at best, but capable enough for a lot of use cases, and significantly cheaper. In six to twelve months, will we see much cheaper open source models competitive with first tier? This will create incentives for 'regulatory capture' if so.

15

1 Comment

Heads up to all my friends working with AI/LLMs, no more tearing your hair out trying to debug issues, or squinting your eyes at charts trying to figure out which model is best for your use case. 🎉 @New Relic AI monitoring, the industry’s first APM for AI, is GA! Troubleshoot faster with full stack visibility. Get performance data on every single request. Compare model behavior and performance. Hit me up if you have any questions Get started here:

37

"AI is powerful, but without the right data foundation, it’s just noise." In the latest Tech Teams Today podcast episode, I had a great conversation with Yongsheng Wu, VP of Engineering at Granica, about why a solid data foundation is crucial for AI success. Here are the top insights from the episode: 1️⃣ Data Foundations Matter: Before diving into AI, ensure your data systems are robust. Solid pipelines & data integrity are critical for meaningful insights. 2️⃣ AI as an Enhancer: AI and ML should optimize existing processes, not solve undefined problems. 3️⃣ Data Quality is King: Curating and labeling data accurately is vital for effective AI models. The better the data, the better the outcomes. 4️⃣ Efficiency & Cost-Effectiveness: Successful AI solutions balance effectiveness with cost efficiency, especially as models scale. Catch the full episode on all major podcast platforms or watch it on YouTube: https://lnkd.in/dbkE__JC #AI #EngineeringLeadership #DataScience #MachineLearning #TechTeamsToday

30

4 Comments