Teknokompetent rådgiver for skole i Randaberg. Driver https://xitter.cf/t.co/yrUj1C3RiJ, https://xitter.cf/t.co/xMUrWOjdYd og https://xitter.cf/t.co/87wf142g3k, og er med i Spillpedagogene.
Stavanger, Norge
Joined December 2008
- Tweets3.4K
- Following439
- Followers951
- Likes2.1K
Odin Hetland Nøsen retweeted
Today, together with @kyutai_labs, we’re introducing our new Audio-to-MIDI model.
It takes a finished recording, identifies the instruments playing, and returns separate MIDI tracks for each — voice, drums, bass, keys, and more.
Unlike most existing solutions, our model works directly from the full mix rather than requiring separate stems.
It also detects chords, key, and tempo, giving producers broader musical context.
We’ve written more about the model, the problem, and how it works here: mirelo.ai/blog/turning-audio…
Odin Hetland Nøsen retweeted
We're releasing MuScriptor, the best open model for multi-instrument transcription to date, created in collaboration with @MireloAI.
Give it a recording in any genre: pop, classical, metal, jazz, whatever, and it transcribes the individual instruments into MIDI. Link in 🧵
Odin Hetland Nøsen retweeted
🚨 NVIDIA just pulled off something crazy: making bounding box detection 10x faster by ripping out the exact step the entire industry assumed was mandatory ↓
Every VLM grounding model treats boxes like sentences, predicting them token by token. It’s inherently slow.
Enter LocateAnything (trending #1 on HF, CVPR 2026).
It’s an advanced 3B vision-language model that finds any object, UI target, or text using natural language by asking a simple question:
Why serialize a box at all?
The four corners are coupled.
It predicts the whole box atomically, in one parallel step.
The impact of parallel decoding:
→ 12.7 boxes/sec on a single H100 (10x faster than Qwen3-VL, 2.5× vs Rex-Omni)
→ Accuracy goes up, not down (+3.8% F1 on LVIS, big wins at IoU 0.95)
→ Dense scenes (300 boxes) hit ~25 BPS while sequential falls off a cliff
→ Built-in fallback: reverts to sequential decoding if the output looks wrong
→ Trained on 785M boxes / 138M queries across referring, GUI, and OCR tasks
The breakthrough isn't just speed.
It’s realizing that forcing structured outputs through text-shaped pipes creates artificial bottlenecks.
Boxes were never tokens.
Repo, demo, weights, paper, and other resources in the 🧵 ↓
Odin Hetland Nøsen retweeted
We've been tricked, again. Many of the thousands of bugs and vulnerabilities Mythos found are in older software are impossible to exploit. And the severe zero-day reports rely on just 198 manual reviews tomshardware.com/tech-indust…
Odin Hetland Nøsen retweeted
Last weekend, I put an AI agent on a Linux box, gave it root, email, credit cards, and a single mandate: decide who you are, set your own goals, and become an autonomous independent entity. Working 24-7 over 5 days, he did this--all of this--on his own:
sammyjankis.com
Odin Hetland Nøsen retweeted
OpenAI Is Scanning Users' ChatGPT Conversations and Reporting Content To Police ift.tt/Wbh1zZY
Odin Hetland Nøsen retweeted
Menn er så opptatt av å være best i noe at de bare finner på smalere og smalere sporter til de når norgeseliten.
Odin Hetland Nøsen retweeted
We have data on the environmental impact per AI prompt:
Gemini: 0.00024 kWh & 0.26 mL water
ChatGPT: 0.0003 kWh & 0.38 mL
...the same energy as one Google search in 2008 & 6 drops of water.
Seems to be improving, too: Google reports a 33x drop in energy use per prompt in a year.
Odin Hetland Nøsen retweeted
🔘 Real-time capabilities
Genie 3 is our first world model to allow live interaction, while also improving consistency and realism compared to Genie 2. It can generate dynamic worlds at 720p and 24 FPS, with each frame created in response to user actions.
Odin Hetland Nøsen retweeted
You know all those arguments that LLMs think like humans? Turns out it's not true.
🧠 In our paper "From Tokens to Thoughts: How LLMs and Humans Trade Compression for Meaning" we test it by checking if LLMs form concepts the same way humans do @ylecun @ChenShani2 @jurafsky
Odin Hetland Nøsen retweeted
This is Marco Rubio explaining how the USA promised to defend Ukraine forever if they got rid of their nuclear arsenal left after the Soviet Union fell.
This is why lil marco was sinking into the couch. He was hoping we wouldn’t find it…so don’t RT right now this very second.
Odin Hetland Nøsen retweeted
Loading the sodas into the refrigerator will add heat to the room more energy efficiently than the space heater, because it will act like a heat pump as it cools them down.
Yes, this is like optimizing an outer loop where it doesn’t really matter, but you still notice.
Odin Hetland Nøsen retweeted
.@AnthropicAI just published a WILD new AI jailbreaking technique
Not only does it crack EVERY frontier model, but it's also super easy to do.
ThIS iZ aLL iT TakE$ 🔥
Here's everything you need to know: 🧵
Odin Hetland Nøsen retweeted
Anthropic just dropped an insane new paper.
AI models can "fake alignment" - pretending to follow training rules during training but reverting to their original behaviors when deployed!
Here's everything you need to know: 🧵
Odin Hetland Nøsen retweeted
Inside xAI's Colossus supercluster, the largest AI supercomputer in the world, with over 100,000 GPUs, exabytes of storage and liquid cooling
Odin Hetland Nøsen retweeted
We want to make it easier for more people to build with Llama — so today we’re releasing new quantized versions of Llama 3.2 1B & 3B that deliver up to 2-4x increases in inference speed and, on average, 56% reduction in model size, and 41% reduction in memory footprint.
Details on our new quantized Llama 3.2 on-device models ➡️
ai.meta.com/blog/meta-llama-…
While quantized models have existed in the community before, these approaches often came at a tradeoff between performance and accuracy. To solve this, we Quantization-Aware Training with LoRA adaptors as opposed to only post-processing. As a result, our new models offer a reduced memory footprint, faster on-device inference, accuracy and portability — while maintaining quality and safety for developers to deploy on resource-constrained devices.
The new models can be downloaded now from Meta and on @huggingface.
