Distilling 200+ Hours of NeurIPS: What’s Next for AI // Nikolaos Vasiloglou // #336

Key takeaways

AI plays an increasingly important role in synthesizing complex academic research, particularly in events like NeurIPS.
The conference underscored the importance of open-source AI models and their rapid development in the current landscape.
Caution is advised when adopting AI for decision-making purposes, as these systems are inherently probabilistic and can result in unreliable outcomes.
Knowledge graphs are pivotal for organizations to obtain structured insights from disparate data sources and enhance application development.
Data attribution and modular language models present innovative approaches for future AI development and monetization.

Summary

Introduction to NeurIPS and Personal Journey

Nikolaos Vasiloglou introduces himself, sharing his journey as a marathon runner, researcher, and recent PhD graduate. He discusses the challenges he faced accessing and analyzing an overwhelming amount of data from the NeurIPS conference, which included over 8,000 publications and hundreds of hours of video presentations. His goal was to communicate this complex information effectively to industry leaders.

The Role of AI in Synthesizing Information

Throughout the discussion, Nikolaos shares his evolving view on the potential of AI in assisting with research synthesis. Initially relying on manual processes, he speaks about experimenting with AI models to distill complex academic proceedings into digestible formats while acknowledging some limitations during his experience at NeurIPS.

Insights from NeurIPS: Key Themes and Developments

Nikolaos outlines twelve critical themes observed at the NeurIPS conference, including advancements in language models and agents in AI. He highlights the shift towards open-source models and the significance of capturing knowledge while emphasizing ongoing debates in the AI community, particularly around the effectiveness of smaller AI models versus established frameworks.

The Future of AI in Decision Making

The conversation shifts to the role of AI in decision making. Nikolaos discusses the increasing pressure for AI systems to take on decision-making roles but warns against fully trusting probabilistic models in critical scenarios. He expresses concerns about the limitations and risks attached to AI decisions, urging caution in the ongoing evolution of AI technologies.

Knowledge Graphs and Their Applications

Nikolaos elaborates on the concept of knowledge graphs, which serve as an intermediary between different data sources, helping organizations to derive structured insights effectively. He emphasizes the importance of designing databases that reflect business realities, thus enhancing application development and operational efficiencies.

Data Monetization and Future Innovations

Towards the end of the conversation, Nikolaos discusses potential strategies for data monetization, including data attribution models that honor the original creators of training data. He also considers the idea of creating modular language models that can be combined to perform complex tasks, which may represent the future of AI model development.

Show notes

Distilling 200+ Hours of NeurIPS: What’s Next for AI // MLOps Podcast #336 with Nikolaos Vasiloglou, VP of Research ML at RelationalAI.

Join the Community: https://go.mlops.community/YTJoinIn

Get the newsletter: https://go.mlops.community/YTNewsletter

// Abstract

Nikolaos' widely shared analysis on LinkedIn highlighted key insights across agentic AI, scaling laws, LLM development, and more. Now, he’s exploring how AI itself might be trained to automate this process in the future, offering a glimpse into how researchers could harness LLMs to synthesize conferences like NeurIPS in real-time.

// Bio

Nikolaos Vasiloglou is VP of Research-ML for RelationalAI, the industry's first knowledge graph coprocessor for the data cloud. Nikolaos has over 20 years of experience implementing high-value machine learning and AI solutions across various industries.

// Related Links

Website: https://relational.ai/

~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

Join our Slack community [https://go.mlops.community/slack]

Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

Sign up for the next meetup: [https://go.mlops.community/register]

MLOps Swag/Merch: [https://shop.mlops.community/]

Connect with Demetrios on LinkedIn: /dpbrinkm

Connect with Nikolaos on LinkedIn: /vasiloglou/

Timestamps:

[00:00] Nik's preferred coffee

[01:05] Distilling NeurIPS insights

[06:43] Choosing research papers

[16:49] Agent patterns at NeurIPS

[21:16] Interest in agent-based innovation

[25:54] Time series forecasting models

[28:15] Tabular foundation models

[36:25] Verifier challenges and complexity

[39:36] Knowledge graph

[45:00] Knowledge graph data challenges

[47:14] Worldview in knowledge graphs

[50:30] Self-serve analytics challenges

[56:22] Llama model adaptation comparison

[56:59] Wrap up