This is why we fund science and specifically mRNA research.

🚨 New AG2 Talk Alert! 🚨 Join us on Aug 28, 9 AM PST for "Maris: A Security Controlled Development Paradigm for Multi-Agent Collaboration Systems" by Jian Cui Jian Cui from UIUC & Berkeley AgentX competition winner! 🛡️ #AI #Cybersecurity RSVP now: discord.com/events/1153072…

How about don't listen to anyone who says anything like this because they actually have no clue how it will play out and just because they do something something AI does not mean they have any experience accurately predicting the future of civilization and the arc of tech

Thanks to ideas like this, we can clearly see that Founder of Google's Generative AI Team can be replaced with an LLM. I think GPT-3.5 should be enough.

✔️ AI agents that are accurate, explainable, and ready for real-world use. Google’s MCP Toolbox now supports #Neo4j, making it easier to build GraphRAG-powered agents that query structured data with precision. In this live session, which is worth revisiting, Kurtis Van Gent

Gary Marcus has been proven wrong repeatedly, you say? like his prediction that GPT-5 would be underwhelming? or that it would be late? or that 2025 agents would be underwhelming and unreliable? or that GPT-5 would still hallucinate? or still make stupid errors? or that

The proprietary frontier models of today are ephemeral artifacts. Essentially very expensive sandcastles. Destined to be washed away by the rising tide of open source replication (first) and algorithmic disruption (later).

The high cost required to train large foundation models makes them the fastest depreciating asset in human history.

“Bubble or not, the AI backlash is validating what one researcher and critic has been saying for years” -FORTUNE

Holy shit, even David Sacks - who built White House AI policy based on strong beliefs in scaling - has come around to saying what I had been saying for years. Scaling as we knew it is done.

Continuing the journey of optimal LLM-assisted coding experience. In particular, I find that instead of narrowing in on a perfect one thing my usage is increasingly diversifying across a few workflows that I "stitch up" the pros/cons of: Personally the bread & butter (~75%?) of

Pattern matching: the most straightforward and effective technique for solving real-world data problems.😮 Some use cases where you can apply pattern matching to your data: 👉Personalize Product Recommendations 👉Optimize Supply Chains 👉Detect Fraud (take a look at the fraud

I'll take the other side of this bet...

Want to train a GPT model with your own data and deploy it fast? 🚀 With #HuggingFace Transformers in PyCharm, you can: ✔️ Browse and add models in your IDE ✔️ Fine-tune models with custom datasets ✔️ Deploy models via FastAPI See the step-by-step guide by Cheuk Ting Ho here:

Fine-tuning LLM Agents without Fine-tuning LLMs Catchy title and very cool memory technique to improve deep research agents. Great for continuous, real-time learning without gradient updates. Here are my notes:

I cannot overemphasize how close we are to AGI. Exponential progress is an amazing drug!

Scaling Test-Time Inference with Parallel Graph-Retrieval-Augmented Reasoning Chains Graph-based retrieval is useful in lots of applications with complex data. This paper is a good example of the benefits:

Ready to level up your RAG apps? 🚀 Dive into Tomaz Bratanic's tutorial to see how knowledge graphs with Neo4j and LangChain can make a difference. Explore more: bit.ly/41lMexf #Neo4j #RAG

Build better RAG by letting a team of agents extract and connect your reference materials into a knowledge graph. Our new short course, “Agentic Knowledge Graph Construction,” taught by @Neo4j Innovation Lead Andreas Kollegger, shows you how. Knowledge graphs are an important way to

Nice essay in defense of tokenizers, and on why tokenizer-free approaches are not really tokenizers free.