OpenAI is amending its hastily arranged deal to supply artificial intelligence to the US Department of War (DoW) after the ChatGPT owner’s chief executive admitted it looked 🔸“opportunistic and sloppy”.
The contract prompted fears the San Francisco startup’s AI could be used for ♦️domestic mass surveillance but its boss, Sam Altman, said on Monday night the startup would explicitly bar its technology from being used for that purpose
or being deployed by defence department intelli…
Still, there are some other things Hypercard did we’d do well to study, even with full-scale tools. Off the top of my head:
- It richly rewarded unguided exploration. Unsuccessful experimentation had a way of leading to paths forward, not just dead ends.
- Much of it worked by direct manipulation: if you want the thing there, you put the thing there. (Unity and Godot both sort of kind of do some descendant of this, but not with the same discoverability and transparency.)
- There was a rich library of good starting points, modifiable examples.
- An empty but functioning new project had essentially zero boilerplate. You didn’t have to have 15 files and hundreds of lines of code to get a blank page.
- Its UI made it easy-ish for newcomers to ask “What can I do with this thing here?” Modern autocomplete and inline docs kind of sort of approximate this, but in practice only for people who already have tool expertise.
- HyperTalk (the programming language) is tricky to write (it’s a p-lang), but it’s remarkably easy to read. You can peer at it with very limited knowledge and make educated guesses about its semantics, and those guesses will be mostly correct. (HyperTalk syntax tends to get the most attention when people talk about this, I think at the expense of the other things above.)
Den of Angels is a forum dedicated to the appreciation of resin ball-jointed dolls and the artisans who make them, started in May 2002 as a Yahoo! Group
https://denofangels.com/
Localized Dynamics-Aware Domain Adaption for Off-Dynamics Offline Reinforcement Learning
Zhangjie Xia, Yu Yang, Pan Xu
https://arxiv.org/abs/2602.21072 https://arxiv.org/pdf/2602.21072 https://arxiv.org/html/2602.21072
arXiv:2602.21072v1 Announce Type: new
Abstract: Off-dynamics offline reinforcement learning (RL) aims to learn a policy for a target domain using limited target data and abundant source data collected under different transition dynamics. Existing methods typically address dynamics mismatch either globally over the state space or via pointwise data filtering; these approaches can miss localized cross-domain similarities or incur high computational cost. We propose Localized Dynamics-Aware Domain Adaptation (LoDADA), which exploits localized dynamics mismatch to better reuse source data. LoDADA clusters transitions from source and target datasets and estimates cluster-level dynamics discrepancy via domain discrimination. Source transitions from clusters with small discrepancy are retained, while those from clusters with large discrepancy are filtered out. This yields a fine-grained and scalable data selection strategy that avoids overly coarse global assumptions and expensive per-sample filtering. We provide theoretical insights and extensive experiments across environments with diverse global and local dynamics shifts. Results show that LoDADA consistently outperforms state-of-the-art off-dynamics offline RL methods by better leveraging localized distribution mismatch.
toXiv_bot_toot
RE: https://social.coop/@cwebber/115922491878315173
This is a fascinating change of thinking.
Ronacher was a very explict all-in AI guy and now comes to many of the points critics have been making for years (without acknowledging any of course). He still…
Das Pentagon wird Musk's AI-Chatbot Grok für strategische und interne Entscheidungen einsetzen, explizit weil selbsternannter Kriegsminister Pete Hegseth eine KI möchte welche nicht woke ist und bereit ist Kriege zu führen.
Der Verteidigungsminister sagte, er wolle verantwortungsvolle KI im Pentagon. Zugleich winkte er Modelle ab, „die Kriegsführung nicht zulassen“.
Seine Vision bedeutet, dass militärische KI „ohne ideologische Einschränkungen arbeitet, die rechtmäßig…
Zooey Deschanel, explaining how Prince ended up on "New Girl". I know she's telling the story very accurately, because she says that Prince's manager's email started with "I'm manager to legendary artist Prince...", which is exactly how her messages did start. https://youtu.be/_pMdUxAztjw
A Colourful Past Podcast
A raw, honest podcast hosted by Danny Shannon, exploring the moments that shape us, the turning points, the setbacks, the breakthroughs, and the courage it takes to change...
Great Australian Pods Podcast Directory: https://www.greataustralianpods.com/a-colo