wrhn

muxamilian

Joined 6 years ago

Karma
240
Submissions
63
Member since
Apr 15
Maximilian Bachl https://github.com/muxamilian https://scholar.google.at/citations?user=llysaaMAAAAJ&hl=en [email protected] https://www.reddit.com/user/muxamilian/ https://x.com/muxamilian All views and opinions are my own
View on Hacker News

Recent submissions

Comment · 4 months ago
Yes but in my experience, LLMs are prone to cheating and implementing "shortcuts" that look functional but are in fact not really if you look closely. At the moment, you still need a human in the loop to make sure the LLM actually did what it promised and didn't cheat.
Comment · 5 months ago
1. Camera captures video 2. Neural network recognizes pigeon 3. Watergun turns toward pigeon 4. Spray pigeon with watergun

Components:

* Electric battery-driven water gun (disassembled, orange) * USB camera * Orange Pi 5 * 2 servo motors (SG90 or MG90S) * Resistors and a transistor for turning on the watergun (e.g. IRLZ44N) It uses an open vocabulary object detection neural network (yolo_world_v2l), so any target can be programmed, not just pigeons. Runs on the Rockchip 3588's Neural Processing Unit.

See also https://x.com/sciencegirl/status/2054945932400087109

Comment · 1 year ago
They get this number by simply adding up all CPUs' processing speed. There is no NPU or similar: https://medium.com/@zlodeibaal/orange-pi-rv2-ai-board-scam-7...
Comment · 1 year ago
I'm rather puzzled by how bad the COCO ground truth is. This is the benchmark dataset for object detection? Wow. I would say Gemini's output is better than the ground truth in most of the example images.

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton