This week China released the third iteration of its AI safety governance framework
– a document that functions as a “mapping of the AI risk landscape and roughly what one should do about it”,
said Gabriel Wagner, a researcher at Concordia, a Beijing-based AI safety consultancy.
Included in the updates are the risks posed by agents,
the cybersecurity threats from frontier models
and the issue of recursive self-improvement
– an as-yet hypothetical scenario w…
The Steam Frame is released. At a shockingly high price range, but that's RAM for you in these post-ai days.
This means improvement on Linux VR in general with all the latest patches. The home-room is updated, the menus all looking fancy and new. Even if you got an old Vive.
Even more exciting than that, Steam have agreed to serve my place in the wait-queue so one might even arrive here in the coming weeks and thank goodnes it's after payday. If it was a week ago I might have had to give up my place.
Sounds like the Alyx updates intend for that game to run on the machine stand-alone, which would be nice. Able to take that big old complex game on the road to demo at parties.
Anyway. I'm poor again and dunno how long the wait for the package will be, but looking forward to seeing it.
God I hope it's not shit and rescues us from the Facebook-dominated market since they bought Oculus.
#vr #linux
Five years after the Taliban took power, an increasing number of countries are renewing ties or scaling up diplomatic contacts with the hardline Islamists, who seized power in Kabul in 2021 after 20 years of brutal insurgent warfare.
Campaigners fear that the individual contacts by different countries risk undermining “coherent, consistent” demands for an improvement in human rights in return for any diplomatic concessions.
The United Nations described a recent “cascade of [Talib…
How MIT thinks about AI and education. An interesting report. Recommendations:
1. Adapt educational processes for an AI-aware world
2. Center people, community, and the residential experience
3. Build processes, teams and tools for continuous reflection, iteration, and improvement
#AI #Education #MIT
https://aiandeducation.mit.edu/report/
Why Do Prefetchers Fail? Let Agents Answer
Xiangfeng Sun, Ceyu Xu, Ningzhi Ai, Zeyu Zhu, Yiyang Yuan, Yuan Xie
https://arxiv.org/abs/2608.13027 https://arxiv.org/pdf/2608.13027 https://arxiv.org/html/2608.13027
arXiv:2608.13027v1 Announce Type: new
Abstract: Hardware prefetchers are crucial to processor performance, yet their design remains labor-intensive and expert-driven. Architects inspect execution and memory-access traces, identify patterns, translate them into online hardware heuristics, and evaluate them in simulation, often with no guarantee of improvement. Human experts cannot systematically inspect billion-instruction traces across diverse real-world workloads.
We present a performance-anomaly-driven autoresearch flow that repeatedly asks why a deployed prefetcher fails and uses the diagnoses to construct the Mixture of Prefetchers (MoP). Each iteration localizes high-impact unexplained misses to program counters, gives agents hardware logs, source code, and sliced traces, validates diagnoses through runnable minimal cases, and synthesizes specialized sub-prefetchers for recurring pattern families. Measured performance and remaining anomalies feed subsequent iterations, enabling simulator-in-the-loop discovery beyond model priors.
The campaign consumes 1.91 billion DeepSeek V4 Pro tokens. On SPEC CPU2006 and SPEC CPU2017, MoP achieves a 61.1% geomean IPC speedup over no prefetching, outperforming the human-designed Alecto, Berti, and Pythia prefetchers by 14.5%, 21.6%, and 23.6%, respectively. RTL synthesis in a 6nm library reports 110 KB of on-chip storage and 0.0347 mm^2 area. To our knowledge, this is the first empirical demonstration that an agent-driven hardware-design process can produce an RTL-practical prefetcher that outperforms state-of-the-art human designs on unseen workloads.
toXiv_bot_toot
@cdarwin@c.im
