desertnaut (@desertnaut) profile picture

Formerly Greek Air Force, @UNPeacekeeping in Africa, @intelen, @satshipai, @DataRobot, @MiltechH. Competitions Master @Kaggle, top ML respondent @StackOverflow

Athens, Greece
Joined December 2011
desertnaut retweeted
Τα υπεραντικείμενα είναι εδώ, στον κοινωνικό, βιωματικό μας χώρο και μας κοιτάζουν σαν πρόσωπα κολλημένα στο παράθυρο ✍️Γιώργος Τζιρτζιλάκης athensvoice.gr/politismos/de…
1
268
Ready for Shadow Ticket #PynchonInPublic
1
8
167
desertnaut retweeted
IS THIS THE END OF THE IRAN WAR?, by Swift Centre (@swift_centre) At the beginning of the year, Swift Centre forecasters correctly identified Iran as 2026’s most likely geopolitical danger. What’s next for the war? The researchers, writing exclusively for UnHerd, estimate a 30% likelihood of the US entering a ground war in Iran — and more than double the usual risk of a recession. Read more below ⬇️ buff.ly/09uo5su
2
2
9
6,355
AI doomers’ echo chamber ⤵️ reinforcement loop of doomerism ⤵️ Radicalization ⤵️ Real harm to its members
Another thread from the PauseAI discord. It’s no big deal to disagree with these guys at the surface level. We think each other wrong, whatever. What the deeper problem is, is that this is a cult ideology that is robbing young talented people of their mental health. That’s evil
5
35
3,669
The demo we are showing is an RL agent on a laptop playing an unmodified Atari with a robotic controller moving the joystick and a camera looking at the screen.
172
286
58
3,122
268,116
True for MOBAs, MMOGs, and strategy games. Not necessarily so for FPS. Sorry, its just science.
if you ever meet someone who at one point was globally ranked in a video game hire them immediately
13
24
7
277
35,341
"o3, show me a photo of the most stereotypical X and LinkedIn feeds as seen on a mobile device. Really lean into it."
43
51
13
994
64,411
desertnaut retweeted
The unnamed book EMPIRE OF AI is mine, based on 300+ interviews, 7 yrs covering AI, & my time as the first reporter who got extensive access to OpenAI. I sought OpenAI's perspective throughout. For months they said it was coming. It never did. Pre-order: empireofai.com.
there are some books coming out about openai and me. we only participated in two—one by keach hagey focused on me, and one by ashlee vance on openai (the only author we’ve allowed behind-the-scenes and in meetings). no book will get everything right, especially when some people are so intent on twisting things, but these two authors are trying to. ashlee has spent a lot of time inside openai and will have a lot more insight—should be out next year.
155
395
57
2,040
307,915
Here with the Greek translation of Inherent Vice #pynchoninpublic
2
14
415
I'm glad @levelsio checked this, but sad our contrib has been erased by later big tech co's. Alec Radford said ULMFiT inspired GPT. ULMFiT's first demo predated BERT. Today's 3-stage LLM approach of general corpus pretraining and 2 stages of fine-tuning was pioneered by ULMFiT.
Replying to @jeremyphoward
So, I asked every AI model if this was true Then I checked Wikipedia Then I asked AI researcher ppl I knew to be sure And not a single source confirms that you created the first LLM You had a massive impact creating the ULMFiT algorithm which pioneered transfer learning and fine-tuning techniques though Happy to be corrected if you did create the first LLM, of course, but I can't find the sources anywhere
42
68
21
1,177
304,475
Traitor? Elon, if you don’t understand that defending freedom is a basic tenet of what makes America great and keeps us safe, maybe you should leave it to those of us who do.
22,062
33,932
3,223
257,505
16,944,893
desertnaut retweeted
ANNOUNCEMENT: Read our interview with @nature, one of the largest scientific journals in the world. We talk about how @yesnoerror and $YNE are using AI agents to audit scientific research. --- “Identifying minor errors in research papers may appear trivial, however as we saw with the recent paper claiming high levels of toxicity in black cooking utensils, even one extra zero can have a big impact on the world.” - @MattPRD The systems first extract information, including tables and images, from the papers. They then craft a set of complex instructions, known as a prompt, which tells a ‘reasoning’ model — a specialist type of LLM — what it is looking at and what kinds of error to hunt for. The model might analyse a paper multiple times, either scanning for different types of error each time, or to cross-check results. The cost of analysing each paper ranges from 15 cents to a few dollars, depending on the length of the paper and the series of prompts used. --- Audit science. Seek truth. Accelerate humanity.
Projects like @yesnoerror & @snewmanpv's Black Spatula Project are using LLMs to quickly & cheaply find errors in scientific papers. The impact could be huge. But the methods create false positives & come with risks. My story on the v different approaches: nature.com/articles/d41586-0…
61
68
14
234
62,728
Get your elan vital in gear: here's my review of a new Henri Bergson biography for the Society of Psychical Research. spr.ac.uk/node/19561
3
10
3
34
1,968
I'll get straight to the point. We trained 2 new models. Like BERT, but modern. ModernBERT. Not some hypey GenAI thing, but a proper workhorse model, for retrieval, classification, etc. Real practical stuff. It's much faster, more accurate, longer context, and more useful. 🧵
126
649
116
4,644
437,843
desertnaut retweeted
Everything you love about generative models — now powered by real physics! Announcing the Genesis project — after a 24-month large-scale research collaboration involving over 20 research labs — a generative physics engine able to generate 4D dynamical worlds powered by a physics simulation platform designed for general-purpose robotics and physical AI applications. Genesis's physics engine is developed in pure Python, while being 10-80x faster than existing GPU-accelerated stacks like Isaac Gym and MJX. It delivers a simulation speed ~430,000 faster than in real-time, and takes only 26 seconds to train a robotic locomotion policy transferrable to the real world on a single RTX4090 (see tutorial: genesis-world.readthedocs.io…). The Genesis physics engine and simulation platform is fully open source at github.com/Genesis-Embodied-…. We'll gradually roll out access to our generative framework in the near future. Genesis implements a unified simulation framework all from scratch, integrating a wide spectrum of state-of-the-art physics solvers, allowing simulation of the whole physical world in a virtual realm with the highest realism. We aim to build a universal data engine that leverages an upper-level generative framework to autonomously create physical worlds, together with various modes of data, including environments, camera motions, robotic task proposals, reward functions, robot policies, character motions, fully interactive 3D scenes, open-world articulated assets, and more, aiming towards fully automated data generation for robotics, physical AI and other applications. Open Source Code: github.com/Genesis-Embodied-… Project webpage: genesis-embodied-ai.github.i… Documentation: genesis-world.readthedocs.io… 1/n
552
2,940
1,607
15,740
3,826,089
desertnaut retweeted
Announcing new open-source Python package: aisuite! This makes it easy for developers to use large language models from multiple providers. When building applications I found it a hassle to integrate with multiple providers. Aisuite lets you pick a "provider:model" just by changing one string, like openai:gpt-4o, anthropic:claude-3-5-sonnet-20241022, ollama:llama3.1:8b, etc. pip install aisuite Open-source code with instructions: github.com/andrewyng/aisuite Thanks to Rohit Prsad, Kevin Solorio, @standsleeping, Jeff Tang and @Johnsanterre for helping build this!
143
1,033
108
5,625
439,065
Just 6 years ago. This NLP reviewer was 100% certain that prompt engineering to unify all NLP problems in a single neural network and just ask it any question was completely misguided and rejected the paper. It would then be cited by the GPT2 and 3 papers. Thanks @arxiv!
19
48
13
612
58,183
Just watched Cursor’s AI coding agent struggle to fix a missing parenthesis for 15 minutes. Guess I still have a job
2
4
1
41
2,200