Charlie Stross visits the Advanced Gas-cooled Reactors at Torness nuclear power station:
The AGRs at Torness [in the UK] are not ordinary civil [nuclear] power reactors. Designed in the 1970’s, they were the UK’s bid to build an export-earning civil nuclear power system. They’re sensitive thoroughbreds, able to reach a peak conversion efficiency of 43% — that is, able to turn up to 43% of their energy output into electricity. By comparison, a PWR peaks at 31-32%. However, the PWRs have won the race for commercial success: they’re much, much, simpler. AGRs are like Concorde — technological marvels, extremely sophisticated and efficient, and just too damned expensive and complex for their own good. (You want complexity? Torness was opened in 1989. For many years thereafter, its roughly fifty thousand kilometres of aluminium plumbing made it the most complex and demanding piece of pipework in Europe. You want size? The multi-thousand ton reactor core of an AGR is bigger than the entire plant at some PWR installations.) It’s a weird experience, crawling over the guts of one of the marvels of the atomic age, smelling the thing (mostly machine oil and steam, and a hint of ozone near the transformers), all the while knowing that although it’s one of the safest and most energy-efficient civilian power reactors ever built it’s a a technological dead-end, that there won’t be any more of them, and that when it shuts down in thirty or forty years’ time this colossal collision between space age physics and victorian plumbing will be relegated to a footnote in the history books. “Energy too cheap to meter” it ain’t, but as a symbol of what we can achieve through engineering it’s hard to beat.
“This plaque was commemorated on October 10, 2018, commemorate its own commemoration. Plaques like this one are an integral part of the campaign to support more plaques like this one. By reading this plaque, you have made a valuable addition to the number of people who have read this plaque. To this day and up to the end of this sentence, this plaque continues to be read by people like yourself. Heritage Toronto 2018”
This is a really atrocious idea:
Palantir also isn’t selling a military-specific AI or large language model (LLM) here, it’s offering to integrate existing systems into a controlled environment. The AIP demo shows the software supporting different open-source LLMs, including FLAN-T5 XL, a fine-tuned version of GPT-NeoX-20B, and Dolly-v2-12b, as well as several custom plug-ins. Even fine-tuned AI systems off the shelf have plenty of known issues that could make asking them what to do in a warzone a nightmare. For example, they’re prone to simply making things up, or “hallucinating.” GPT-NeoX-20B in particular is an open-source alternative to GPT-3, a previous version of OpenAI’s language model, created by a startup called EleutherAI. One of EleutherAI’s open-source models — fine-tuned by another startup called Chai — recently convinced a Belgian man who spoke to it for six weeks to kill himself. What Palantir is offering is the illusion of safety and control for the Pentagon as it begins to adopt AI. […] What AIP does not do is walk through how it plans to deal with the various pernicious problems of LLMs and what the consequences might be in a military context. AIP does not appear to offer solutions to those problems beyond “frameworks” and “guardrails” it promises will make the use of military AI “ethical” and “legal.”