Leo Walker

On why you should use Jupyter Notebooks: "Notebooks are a great way to tell a story, and telling stories is what all fields should be about. Especially computer science." This is from an interview with Doug Blank, Head of Research at @cometml. He continues: "To me, a Jupyter notebook is a blank sheet of paper. You can write a story in it. And if you change the name of one of the characters in paragraph one, you have to change the name of the character throughout the whole story." People often tell me they don't like notebooks because people write bad code in them. Nonsense. Bad developers write bad code. Notebooks have nothing to do with that. "Some educators feel that it's too open-ended and too flexible, but I disagree — it's a new way of doing computing (...)." I always recommend that developers learn how to use notebooks. Not as their primary way of writing software but as an alternative tool they can use to experiment, troubleshoot, and become more effective at their work. The math in my head is simple: A developer who knows how to use notebooks effectively is better than a developer who doesn't. Notebooks aren't a replacement for what you do. They are a boost. The rest of the interview is pretty good: https://lnkd.in/e7ezJsx7

83

8 Comments

🚀 Tackling #MachineLearning challenges! Learn how Phillies are using Airflow to deploy, retrain, and monitor ML models, empowering teams to focus on baseball research. ⚾ Join Mike Hirsch & Sophie Keith at #AirflowSummit Register at: https://airflowsummit.org/ Apache Airflow #AirflowSummit #Airflow #ApacheAirflow #AirflowSummit2024 #dataorchestration #batchprocessing #dataprocessing #datapipelines #data #datasience #opensource

11

Last week, I spoke about AI and regulations at an event at the U.S. Capitol attended by legislative and business leaders. I’m encouraged by the progress the open source community has made fending off regulations that would have stifled innovation. But opponents of open source are continuing to shift their arguments, with the latest worries centering on open source's impact on national security. I hope we’ll all keep protecting open source! I’m encouraged by the progress the U.S. federal government has made getting a realistic grasp of AI’s risks. To be clear, guardrails are needed. But they should be applied to AI applications, not to general-purpose AI technology. Nonetheless, some companies are eager to limit open source, possibly to protect the value of massive investments they’ve made in proprietary models and to deter competitors. It has been fascinating to watch their arguments change over time. For instance, about 12 months ago, the Center For AI Safety’s “Statement on AI Risk” warned AI could cause human extinction and stoked fears of AI taking over. This alarmed leaders in Washington. But many people in AI pointed out that this dystopian science-fiction scenario had little basis in reality. About six months later, when I testified at the U.S. Senate’s AI Insight forum, legislators no longer worried much about an AI takeover. Then the opponents of open source shifted gears. Their leading argument shifted to the risk of AI helping to create bioweapons. Soon afterward, OpenAI and RAND showed that current AI does not significantly increase the ability of malefactors to build bioweapons. This fear of AI-enabled bioweapons has diminished. To be sure, that bad actors could use bioweapons — with or without AI — remains a topic of great international concern. The latest argument for blocking open source AI has shifted to national security. AI is useful for both economic competition and warfare, and open source opponents say the U.S. should make sure its adversaries don’t have access to the latest foundation models. While I don’t want authoritarian governments to use AI, particularly to wage unjust wars, the LLM cat is out of the bag, and authoritarian countries will fill the vacuum if democratic nations limit access. When, some day, a child asks an AI system questions about democracy, the role of a free press, or an independent judiciary, I would like the AI to reflect democratic values rather those of authoritarian leaders. I came away from Washington optimistic about the progress we’ve made defending open source. Looking beyond the U.S. federal government, there are many jurisdictions. Unfortunately, arguments for regulations that would stifle AI continue to proliferate. But talking to regulators does have an impact. If you have a chance to talk to a regulator, I hope you’ll do what you can to help governments better understand AI. [Original text in The Batch (with links): https://lnkd.in/g5SZ6ypa ]

1,615

58 Comments

How often does an opportunity come to work on a greenfield platform like Pydantic Logfire? Build a database that will scale (if we get it right) to petabyte scale and change how develoeprs understand their applications.

115

3 Comments

“𝐈 𝐭𝐡𝐢𝐧𝐤 𝐭𝐡𝐞 𝐩𝐫𝐨𝐜𝐞𝐬𝐬 𝐨𝐟 𝐜𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐧𝐠 𝐝𝐚𝐭𝐚 𝐭𝐨 𝐀𝐈 𝐢𝐬 𝐨𝐧𝐞 𝐨𝐟 𝐭𝐡𝐞 𝐡𝐢𝐠𝐡𝐞𝐬𝐭 𝐥𝐞𝐯𝐞𝐫𝐚𝐠𝐞 𝐣𝐨𝐛𝐬 𝐭𝐡𝐚𝐭 𝐡𝐮𝐦𝐚𝐧𝐬 𝐜𝐚𝐧 𝐡𝐚𝐯𝐞.” 𝐀𝐥𝐞𝐱 𝐖𝐚𝐧𝐠, 𝐅𝐨𝐮𝐧𝐝𝐞𝐫 𝐒𝐜𝐚𝐥𝐞.𝐀𝐈 “As a human expert you have the ability to have society wide impact by producing data to help improve these models.” Had a few minutes to catch Alex Wang on 20VC today. Alex highlighted a critical bottleneck in AI model improvement. He explained that any data easily accessible from the open internet has already been utilized to enhance existing AI models, making them excellent at emulating the internet. However, truly powerful AI systems need to go beyond this. We want AI to perform complex reasoning and collaborate effectively with humans. But we can't achieve this by just crawling internet data. Why? The complex thinking and reasoning that humans employ when tackling intricate tasks aren’t captured on the internet. The intellectual processes driving our economy aren't documented online. So, how do we capture data that isn’t readily available in datasets? We need frontier data, which includes complex reasoning chains and agent interactions. This data is often locked within enterprises and must be mined, tailored to each customer's proprietary information and desired output. Even a decade ago, when enterprises began showing interest in deploying intelligence on aerial devices like drones, this was apparent. It was also a key inspiration behind what we built at Cloneable.ai How can we enable internal teams and data scientists to leverage their organization's proprietary information to perform complex tasks that augment the human workforce.... and rapidly deploy these solutions in the real world; while keeping their proprietary data and processes safe and secure. There’s other great stuff in here too. Give it a listen. https://lnkd.in/etzH-xWE

8

AI agents are reshaping what's possible in tech, but there's a crucial piece often overlooked: Systems Engineering. We tend to see a great deal of attention on advanced models, while the work of the foundational systems that actually brings these agents to life often goes unnoticed. I think this survey paper does an excellent job translating how effective AI agent advancements depend on the robustness of systems design. Consider the capabilities of an AI voice agent. Its performance doesn't depend solely on its ability to parse language but also on how it’s engineered to reason, plan, and interact across various tools and environments. Without careful engineering, the agent’s decision loops or tool-calling functionalities may fall short, leading to interruptions or mismatches with user needs. By emphasizing system architecture, defining clear roles, managing inter-agent communication, and ensuring scalable tool access—engineers can prevent these bottlenecks, enhancing the agent’s reliability and accuracy in complex use-cases. The paper also highlights how current benchmarks, like AgentBench and WildBench, highlight another critical factor: real-world variability. Many agent benchmarks still rely on static datasets, which struggle to keep pace with evolving LLM capabilities. Check out this paper for some interesting insights into AI agent architecture. #AI #AIagents #datascience #systemsengineering https://lnkd.in/ei5A7v9S

12

2 Comments

🚀 New Article on Towards Data Science which explores how Topic Modeling can better enhance machine recommendations to users. 🧠This directly impacts the limits in Semantic Reasoning of recommendation systems and Large Language Models (LLM), and could overtime lead to more human perception in these systems! ❓ While many businesses are leveraging Natural Language Processing (NLP) systems, particularly for recommendation engines and LLMs, extracting deep value from these technologies can be computationally expensive. In my work, I aim to address their limitations, specifically their inability to provide context-aware recommendations based on nuanced, user-specific information—while keeping computational costs low. 🔴 As humans, we naturally deduce connections and topics across different pieces of information. To advance AI tools like LLMs and recommendation systems, we must work to build this same capability—enabling them to understand and align context in a way that mirrors human reasoning. 🔴 While recommendation systems and LLMs have made incredible strides in understanding language at scale, we’re still seeing gaps in their ability to grasp deep semantic meaning, leading to challenges like: • Misinterpreting Context • Struggling with Complex Logic • Falling Short in Topic Alignment I break down these issues and discuss what’s next for overcoming these barriers. Check it out and let me know your thoughts! #NLP #AI #RecommenderSystems #SemanticReasoning #MachineLearning #AIResearch #TowardsDataScience #DataScience #AIinBusiness #TopicModeling

14

Check out my new tutorial video "Build Hierarchy in Table with Recursive Metachain Scope Query." This video demonstrates the powerful capabilities of using the scope query with the "Build Hierarchy" option in Cameo Systems Modeler. Perfect for anyone looking to leverage recursive metachains to visualize expandable decompositions and other repeating patterns directly within a table. https://lnkd.in/esZi3wqY #SystemsEngineering #ModelBasedSystemsEngineering #MBSE #CameoSystemsModeler #MagicDraw #Engineering #Tutorial #TechCommunity

60

13 Comments

Join us on May 29, 2024, for a webinar covering deep learning frameworks on Argonne National Laboratory's Aurora exascale supercomputer. In this developer session, ALCF's Corey Adams will discuss deep learning frameworks optimized for Aurora, including TensorFlow, PyTorch, and the experimental release of JAX. Other topics include how to achieve optimal single device performance, as well as scaling up and scaling out models. The session will include examples and demonstrations that can run on Sunspot or Aurora. https://lnkd.in/gY2WMZ3x

7

Data Veterans are too humble. Don't let a Model Moddler take your next job. You're not bragging about your unique selling points. Maybe you think that being humble is a virtue. It's not. Humility won't land you your next job. If you are the perfect fit for the role you need to telegraph it at each step of the selection process. That's what the recruiter needs. That's what your future employer needs. Go back to being humble once you start your new job. Don't let your next opportunity slip away because you're not explicit enough about your abilities. It's time to start bragging. P.S. Repost this ♻️ if you think it's time to stop the humble mumble.

10

4 Comments

The atmospheric river flooded San Francisco, so we flooded our codebase with a downpour of new updates. 🌊 ⛈️ 🌊 ⛈️ 🌊 ⛈️   💨 Underscore expressions now support more  chalk.functions for working with arrays and Dataframes, mathematical operations, encoding, formatting datetime, and strings.  😶‍🌫️ You can now choose whether to cache nulls or default values in the online store with the cache_nulls and cache_defaults parameters. Customers with Redis or DynamoDB online stores can also select to evict null/default feature values for any null/default feature value that would have existed in the online store.  🗺️ You can now define Chalk features as map types, for example user_preferences: dict[str, bool]  🎣 In addition, you can now retrieve Map document types from DynamoDB data sources as either dicts or strings. As always, more detail and much more linked in the full changelog in comments. From the Chalk team, have a wonderful holiday! 🦃

9

1 Comment

Here's the recording of yesterday's live stream on building durable #Python backends with #opensource DBOS and DBOS Cloud: https://lnkd.in/gXXKWGVK DBOS architects Qian Li and Peter Kraft explained how: - DBOS wraps Python workflows, operations, and steps with functionality that makes them durable and observable by default. - #RAG workflows are made reliable  - Durability solves problems in #AI apps like #LLM wait time cost, time outs, and rate limiting,  Check it out and let us know what you think!

9

What's the difference between model explainability, interpretability, and observability? Jason Zhong's debut TDS article outlines the nuanced distinctions you should be aware of.

15

1 Comment

For our current defending LLMs project I’ve been doing a bit of ye olde model training. This is the output from a recent model for detecting SQL injections, fine tuning on top of DistilBERT (https://lnkd.in/g_Dh8KDJ). Happy with those results. I really do believe that more of us need to be exploring how to properly secure our LLMs as opposed to just treating them like cheap labour. #splunk #splunksecurity #AI #ML #Steve #pirates

45

3 Comments

Videos from the recent Apache Iceberg Meetup in SF have been posted! Rewatch Kevin W.’s talk on “Lessons From Building Iceberg Capabilities in Daft, A Distributed Query Engine” where he walked through how we adapted PyIceberg for distributed workloads and introduced workarounds we implemented for challenges with using existing Python/Rust Iceberg tooling. Learn about what it means for an Iceberg library to provide useful abstractions while giving the query engine proper control over execution. Watch now:

4

In this podcast, we speak with Mosaic ML's Hagay Lupesko about Mixture of Experts and their new DBRX LLM. #DataScience #AI #ArtificialIntelligence https://hubs.li/Q02y7pD00

2

In this podcast, we speak with Mosaic ML's Hagay Lupesko about Mixture of Experts and their new DBRX LLM. #DataScience #AI #ArtificialIntelligence https://hubs.li/Q02ywxGY0

1