Colombo AI Conclave ?

Any thoughts of hosting an MCP server that allows the intelligent world to connct to your legacy applications?

Interesting! I don’t mean to dampen the humor, but here’s my take: The meme depicts AI agents heavily reliant on data to a similar extent as legacy ML systems. But I’d argue the story’s more nuanced. AI agents, especially those powered by generative AI, aren’t primarily reliant on massive, structured datasets like legacy ML systems. Instead, they thrive on unorganized inputs and human-like tools, adapting on the fly as models grow more intelligent. They can thrive on unstructured inputs… Unstructured data isn’t the enemy; it’s the fuel…. The real power lies in AI agents leveraging data dynamically (think real-time analysis or external queries) rather than being limited by its absence. Enterprise success with Gen AI isn’t about being constrained by unstructured data; it’s about mastering it through the flexibility of intelligence embedded in products.

Imagine combining NuExtract with Mistral OCR ! https://lnkd.in/eCrAFDxf

We at the AI Center (in Mountain View CA) in Samsung Research are calling for proposals through START program (https://lnkd.in/gimcsa5t) to collaborate with brilliant minds in academia. The program calls for proposals in many topics. Specifically, our team, which is knowledge and natural language processing lab, is interested in using knowledge representation, reasoning, retrieval, and QA to provide our Samsung Mobile users with exceptional user experience with their personal data. This will be a ground breaking addition to our AI experiences on edge devices to support millions of users globally with privacy in mind (data never leave their phones). Our team is working on question answering on user personal data using knowledge graphs (RDF/S, OWL) and LLMs. Our approaches completely run on the phones. There are many challenges to solve to make this a reality. Attached is the call for proposals on this particular topic. Please submit your novel ideas and proposals, so that we can work on this problem together. You can send a note here ([email protected]) if you submit a proposal for this particular topic.

Mistral AI gets ahead of the pack on OCR! Mistral's new OCR API — parses 1000-2000 pages for $1 — achieves state of the art results on tables, multilingual — supports structure: images, bounding boxes, scans, equations Let's start feeding the PDFs ! https://lnkd.in/e6g__zfS

?? BREAKING: Amazon Nova Micro just got INSANELY faster! ?? Take a look at the graph! Seriously! ?? - check out this incredible performance jump! ?? What you're seeing is Nova Micro's output speed absolutely SKYROCKETING. When choosing a model, there are lots of things you might look for, with speed and cost often at the top of the list. Amazon Nova has always been optimised for both of these, and now it just got faster! What you need to know about the speed boost for Nova Micro: ?? You don’t need to do anything. Just keep using Nova and see the speed boost So if your apps need near instantaneous response, give this a go, and let me know. -> Link here: https://lnkd.in/g_8xEv9X FYI: In Python, if you want to time something, such as a function call to an LLM, you can use time.perf_counter() something like this: ----------------- import time start = time.perf_counter() function_to_time() end = time.perf_counter() print(f"Execution time: {end - start:.6f} seconds") ----------------- Let me know how you go! ?? P.S. When you call the Bedrock invoke_model() or the converse() API on Nova, add the ‘us.’ or ‘eu.’ for the cross-region inference profile! Amazon Web Services (AWS) #AmazonNova #MachineLearning #AWS #AIPerformance #LLM #GenerativeAI #AIPlatforms #DeveloperTools #TechNews #CloudComputing #AmazonBedrock #LLMOptimization #AILatency #SoftwareDevelopment #ProductivityBoost #TechInnovation

3D game starter kit for @cursor_ai. ?? Latest React 19 + React Three Fiber ?? Predefined .cursor/rules ?? Typescript https://lnkd.in/g_zh7dhS https://lnkd.in/gTcUWsPe

The future of AI is “stateful agents” - agents that can learn from experience. Large language models possess vast knowledge, but they're trapped in an eternal present moment. While they can draw from the collected wisdom of the internet, they can't form new memories or learn from experience: beyond their weights, they are completely stateless. Every interaction starts anew, bound by the static knowledge captured in their weights. As a result, most “agents” are more akin to LLM-based workflows, rather than agents in the traditional sense. The next major advancement in AI won't come from larger models or more training data, but from LLM-driven agents that can actually learn from experience. At Letta, we are calling these systems “stateful agents”: AI systems that maintain persistent memory and actually learn during deployment, not just during training. Most LLM APIs and agentic frameworks that are built around the assumption of statelessness. State is assumed to be limited to the duration of ephemeral sessions and threads, baking in the assumption that agents are and always be stateless. A stateful agent has an inherent concept of experience. Its state represents the accumulation of all past interactions, processed into meaningful memories that persist and evolve over time. This goes far beyond just having access to a message history or a knowledge base via RAG. Key characteristics include: - A persistent identity providing continuity across interactions - Active formation and updating of memories based on experiences - Learning via accumulating state that influences future behavior The next generation of AI applications won't just access static knowledge - they'll learn continuously, form meaningful memories, and develop deeper understanding through experience. This represents a fundamental shift from treating LLMs as a component of a stateless workflow, to building agentic systems that truly learn from experience. The term "agent" has strong roots in reinforcement learning (RL) but recently has started to lose all meaning - "stateful agents" adds an important qualifier to the term to clearly distinguish it from an "LLM-driven workflow". Next time someone tells you about the agent they're building, try asking them if it's a stateful agent - if not, why? Full blog post on stateful agents in comments. ??

GPT-4.5 is released!!! ?? OpenAI has released a research preview of GPT-4.5, an improved version of its AI model. Building on GPT-4o, it expands pre-training and is designed to handle a wider range of tasks beyond STEM-focused reasoning. What’s New? ?? Better understanding and responsiveness: GPT-4.5 improves how it interprets and follows user intent, making interactions feel more natural. ?? Fewer errors and hallucinations: It’s more reliable when providing information and reasoning through complex topics. ?? Improved alignment: New training techniques help the model better grasp context, nuance, and human needs. ?? Stronger safety measures: Extensive evaluations found no significant increase in risks compared to previous versions. #ai #openai #chatgpt #gpt45

Veracity Group - Come Accelerate Change with us :)