Hume AI’s Post

Your voice can control your computer. At first glance, this is a groundbreaking accessibility feature; then you realize that, in a matter of months, this will often be preferable over a mouse and keyboard - for anyone, in a big fraction of their interactions with technology. It's the fastest and most efficient way we'll convey our preferences to AI in real time.

Voice technology is a leap forward 10x further than keyboard to touchscreen.

It still blows my mind how Figure 01 uses a singular E2E model for converting instructions + environment into action trajectory. How does it work? This probably means there's an encoder for vision+audio, followed by an exposed representation layer in a language model, with prop further going into a speech network + a brand new model they've devised for reasoning out the optimal trajectory of actions. And this entire net is then end-to-end fine-tuned on real world environment data. No wonder it's as fast as it is. (This is just my speculation; Figure hasn't actually spoken about the algo behind 01).

Spatial context and semantic context are two different aspects in computer vision that play distinct roles in understanding visual data. The differences between spatial context and semantic context are described concisely in the following short video: https://lnkd.in/dZCGAVzx

In the ever-evolving landscape of computer vision, the You Only Look Once (YOLO) algorithm has continued to push boundaries since its debut. Just seven months after the release of its pioneering first version, YOLO v2 emerged, refining and expanding upon its predecessor’s capabilities. This iteration introduced a host of innovations aimed at enhancing both accuracy and efficiency in object detection. YOLO v2 represents a significant leap forward, addressing key limitations observed in the original version while introducing novel techniques to improve localization, recall, and adaptability across diverse datasets. From advancements in network architecture and training methodologies to strategic improvements in handling small objects and multi-scale detection, YOLO v2 showcases a meticulous approach to refining real-time object detection systems. Join us as we delve into the details of YOLO v2, exploring its innovative features, performance improvements, and the methodology that propelled it beyond its inaugural version. This summary highlights the pivotal changes that have cemented YOLO v2 as a cornerstone in the field of object detection, setting a benchmark for subsequent iterations and inspiring further advancements in the realm of computer vision.

Part 2/3 w/ Alex Hurtado covering how Monte Copilot can take simple natural language and bring your detection & query logic to life. ⚡ Wanna check out how Monte can help you glow up your SOC? Check out the full guide here: https://lnkd.in/dVuU6Paf

This 30-min-read blog post on how to craft and generate a 25B+ tokens synthetic text dataset distills more information than most NeurIPS best papers! Enjoy it here: - https://lnkd.in/eGArdjwY - https://lnkd.in/ey2cdTap

Notebook with implementation of ML pipeline with Apache Beam that extracts audio from files, classifies texts, gets answer from an LLM and turns the text back into speech. Join Beam College for the walkthrough.

Even with stringent controls like lowering GPT-4's temperature to zero, setting a fixed seed, and adjusting the Top P parameter, the model still produces significantly different outputs. This inherent randomness, influenced by minute variations in software, hardware, and timing, makes it challenging to achieve consistent results from LLMs. (Example from paper: https://lnkd.in/eHpivd_B, found by Lennart Meincke)

Why is "False" the answer to the test question on algorithm time complexity?   #TestQuestion #AlgorithmComplexity #TimeComplexity Hey everyone! 🌟 So, I recently came across a tricky test question that got me scratching my head. The question was about two algorithms, A₁ and A₂, with time complexities T₁ = O(n²) and T₂ = O(n log n), respectively. The question asked if A₂ must be faster than A₁ - True or False? I initially chose "True" but the correct answer was "False". I... Source: https://lnkd.in/gCeRkVs7 #mymetric360