北京时间昨晚，X用户扬·布鲁尔森（Jan Broersen）发帖质疑马斯克的太空算力方案。全文如下：“我对埃隆·马斯克的敬意已经消失的差不多了，但我仍以为他至少是个不错的工程师（特斯拉是好车）。可听到他说在太空里‘把东西冷却下来’更容易之后，连这最后一点敬意也消散殆尽了。事实正好相反。真空是最好的隔热体。在太空里，你只能靠辐射把热量带走。这些太空数据中心极不可能成真（当然也不会在我们有生之年成真）。”
布鲁尔森的个人简介表明，他是荷兰乌得勒支大学哲学、逻辑与AI教授。
这条帖子在十几个小时内迅速冲到约320万浏览，“友好”评论也很多。
SpaceX早期技术核心汤姆·穆勒（Tom Mueller）只充满讽刺的回了一句：“哲学教授”——再配一个白眼。
詹姆斯·斯蒂芬森（James Stephenson）怼道：“那你对桑达尔·皮查伊的敬意怎么样？人家已经有一台在太空里、靠真空里辐射散热的数据中心了。”他指的是Google“Project Suncatcher”原型卫星已在轨点亮TPU并回传结果。
马斯克更是直接怒喷。他毫不客气的回复布鲁尔森：“你完全是个智障。”
X上的特色账号Retard Finder立刻跟帖：“确认”。这个账号专干一件事：把网上离谱发言揪出来示众，粉丝近百万，风格粗暴。
红杉资本合伙人肖恩·马奎尔（Shaun Maguire，加州理工物理博士。没错，本号今早刚发了他写的深度分析文章《人们并不理解SpaceX》）则在布鲁尔森的帖子下写了一段长回复。大意是：
太空散热的总难度确实远高于地面，但其中99.9999%的难点其实是星舰——一旦能便宜地把大量质量送上天，太空散热就真的更轻松。冷却系统可以是被动的：一块又轻、导热又好的大金属（铝是好候选），里面循环工质；氨是太空散热器的历史工业标准。工质在芯片附近由液变气，密度差推动热气在回路里循环；铝把热量从工质里抽走，再靠黑体辐射把热散进太空。黑体辐射功率则随温度四次方上升，芯片和工质越热，散热越痛快。他说，坦白讲这套系统比地面简单得多。
马斯克随后在X上赞同肖恩：
“是的，在太空里散热要容易得多。
每平米1300瓦对于一个远离太阳指向的铝制散热器来说很容易实现，因为它主要看到的只是宇宙背景辐射，大约在绝对零度以上3度左右！
太阳能电池板每平米产生250瓦电，所以你应该预期它们的面积是散热器的约5倍。卫星的大部分质量都是太阳能电池板。
SpaceX披露的AI1卫星规格：中心是算力核心，连接算力核心的狭长条带为双侧液冷辐射散热器，卫星两翼是巨大的太阳能电板
4.
争议的核心是SpaceX公开的轨道AI卫星概念AI1（Starmind星座里的一代机）。
公开材料里，AI1大约展开翼展70米、高度20米量级，峰值算力约150千瓦、持续约120千瓦；散热靠可展开液冷散热器，面积大约110平方米，带冗余泵回路。
太空没有空气对流，热只能变成红外辐射散出去——所以哲学教授说的“只能靠辐射”没错；但马斯克、肖恩强调的是另一半：只要运力便宜、辐射面够大、工质和朝向对了，辐射散热可以很高效，不必像地面机房那样堆风扇、冷却塔和电老虎空调。
散热器往往刀刃立面朝向太阳、两面一起辐射，尽量少接受日照、多朝向接近绝对零度的宇宙背景，这样散热效率最高。
地面数据中心缺电、缺水、缺地，而轨道上太阳常在，冷源是整个宇宙。AI1这套散热设计，说白了就是像射击一样，把多余热量打进虚空。
哲学教授还停留在真空隔热的直觉上，马斯克们抓住的是工程边界条件：辐射散热不是不能做，是以前太贵、太重，星舰如果能把质量成本打下来，就不是“下出一步好棋”，而是完全新造了一个棋子。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Last night, Beijing time, X-user Jan Broersen issued an invitation to question Mask's space algorithm. The whole text reads: “My respect for Elon Mask has almost disappeared, but I still think he's at least a good engineer. And when he said it was easier to cool things down in space, even that last respect was lost. The opposite is true. The vacuum is the best insulation. In space, you can only take the heat with radiation. These space data centres are highly unlikely to come true (and certainly not in our lifetime).
A personal profile of Brewerson indicates that he is a professor of philosophy, logic and AI at the University of Utrecht, the Netherlands.
The post quickly reached about 3.2 million viewers in a dozen hours, and there were many “friendly” comments.
"Professor of Philosophy" -- another white eye.
James Stephenson said, "What about your respect for Sandra Pichay? There's already a data centre in space, in a vacuum, where radiation is spreading." He referred to the Google “Project Sunchacher” prototype satellite that had lighted TPU in orbit and returned the results.
Mask is even more directly angry. And he said, "You're completely retarded."
The special account on X, Retard Finder, immediately posted "Confirm". This account is dedicated to one thing: drawing out ridiculous online speeches, fans of almost a million, harsh style.
Shaun Maguire, Doctor of Technology, California. Yes, this number just this morning sent out an in-depth analysis of his article " People don't understand SpaceX " ) and wrote a long reply under Brewerson's post. To the effect that:
It is true that the total difficulty of space dispersion is much greater than that of the ground, but 99,99999 per cent of it is actually a starship — it is really easier to disperse space once a large amount of mass can be delivered cheaply. The cooling system can be passive: a piece of large metal that is light, hot and good (aluminium is a good candidate), which is recycled; ammonia is the historical industrial standard for space radiators. The mass of work is transformed by liquid gas near the chip, and the density gap drives the heat through the circuit; the aluminium removes the heat from the mass of work and spreads the heat into space with bold radiation. The irradiation power in bold increases four times with temperature, and the heat of chips and works increases and heat spreads. He said that, frankly, the system was much simpler than the ground.
Mask then endorsed Sean on X:
"Yes, it's much easier to heat up in space.
1300 watts per square metre is easy to achieve for an aluminum radiator far from the sun, because it sees only cosmic background radiation, about 3 degrees above absolute zero!
Solar panels generate 250 watts per square metre, so you should expect them to be about five times larger than the radiator. Most of the mass of the satellite is solar panels.
A1 satellite specification disclosed by SpaceX: the centre is the core of computing power, with a narrow strip connecting to the core of computing power being a two-sided liquid cold radiation radiator, and satellite wings being huge solar panels.
4.
At the heart of the dispute was SpaceX ' s open orbital AISAT concept AI1 (a generation machine in the Starmind constellation).
Of the publicly available materials, AI1 has a wingspan of approximately 70 metres and an altitude of 20 metres, with a peak calculus of about 150 kilowatts and a duration of approximately 120 kilowatts; the heat is based on an extended liquid cold radiator, which covers an area of approximately 110 square metres and carries a redundant pump back road.
Space is free of air currents, and the heat can only spread out into infrared radiation — so the professor of philosophy is right about “radiation only”; but Mask and Sean stress the other half: radiation dispersion can be efficient, without faning, cooling towers, and electric tiger air conditioning, as long as the power of transport is cheap, the radiation surface is large, the work is high and the direction is right.
The radiator tends to be the most efficient in the form of a bladed face towards the sun, with both sides irradiated, with as little as possible exposure to sunlight, and more towards an absolute zero-degree universe.
Ground-based data centres lack electricity, water and ground, while the sun is often present in orbit and the cold source is the entire universe. AI1's distilled design, which is white, is like shooting, filling the excess heat into the void.
Professors of Philosophy are still stuck in a vacuum of insulation, and the Mask have captured the engineering boundary conditions: radiation dispersion is not impossible, it was too expensive and too heavy, and if a starship can hit down its mass costs, it is not a “good move” but a completely new piece.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.