马斯克关于机器人手术的荒谬且可能有害的预测

认为机器人会很快成为地球上最好的外科医生的想法荒谬，我在X上对此进行了质疑：

我为什么还要打这个赌？这里有三个原因。

埃隆的时间线预测以我看来显然很糟糕，但却常被当作真理报道。我用赌注来让人们不那么当真。我上次给他开出百万美元赌注：http：//MyBillingRep@bge.com;（由达米恩·汉克杰提高至1000万美元的报价）是在2024年4月回归，当时马斯克声称人工智能将在2025年底超越所有人的智力。当然，这并没有发生。事实上，在最近接受《经济学人》采访时，马斯克本人悄悄地退回到2031年，这与他对自动驾驶汽车的常规预测有些相似。手术预测也肯定会在稍后以新日期重新包装。埃隆·马斯克，像许多大型语言模型一样，经常错且从不怀疑。够了。

关于机器人手术的时间线，马斯克的错误如此严重，令人难以理解。我问了著名机器人学家罗德尼·布鲁克斯对埃隆·马斯克主张的看法，他五分钟后坚定地反驳，称马斯克的主张是“完全疯狂的说法”，并详细说明如下：

即使有一双手臂牢固固定在地面上，而不是放在两足平衡平台上，也没有实验室演示能接近对活体试验动物进行手术。仅仅是外科医生如何观察发生的情况，甚至没有实验室证明。随着时间推移，我们将看到更多非侵入性手术，未来几年将由AI和几何传感器团队与专家外科医生组成，只有这样，这些技术才足够好用在猪身上，只有经过多年完善，我们才能进入FDA（或其他机构）允许在人类身上试验这些方法。看看我周末在 rodneybrooks.com/blog 的博客文章：https://rodneybrooks.com/blog/，讲述大家时间比例都错了。

当我询问布鲁克斯关于现有机器人手术试验的最新进展时，他补充道

所有这些仍然有人在环节中参与。我在2000年前后是第一个投资他们的风险投资者，Stryker 的 Mako 手术用于膝关节置换，是最接近的案例，因为它处理的不是软组织，而是可以在手术中通过X光准确成像的骨骼，而且主要是几何问题，但随后还有许多由人类关注和监控的事项，以确保不会意外损伤血管或神经。

到本十年末，这一切根本不可能解决。马斯克告诉 Diamandis 的事情根本不会发生。

正如 Hinton 那句著名且著名地错误的 2016 年声明——不再需要培训放射科医生——可能吓跑了潜在的放射科医生一样，马斯克有可能吓跑潜在的外科医生。这非常糟糕，尤其是他的前提在最好的情况下至少错了十年，很可能还要更久。

如果这些科技公司 CEO 对他们所说的话能更负责任就好了。这才是我提出这些赌注的真正原因。而这些 CEO 永远不接受它们的真正原因。

附言：All-In 播客的 Jason Calacanis 表示有兴趣将这场辩论变成 Polymarket 上的一个赌注：https://x.com/jason/status/2083900792109424961?s=61。如果他确认参与，我会报告回来。关于标准，我提供了十项复杂手术的列表作为可能的指标：https://x.com/garymarcus/status/2083935879454953613?s=61，到目前为止还没有回复。

以他糟糕的记录来看，仍有人听马斯克的预测令人难以置信。话虽如此，我们当前这个时代有很多事情也令人难以置信。

错了十年？在这种情况下，他很可能是错了一百年。

The idea that robots will be the best surgeons on earth anytime soon is absurd, and I challenged him about it on X:

Why did I bother to offer such a bet? Here are three reasons.

Elon’s timeline predictions are notoriously, and to my mind, obviously, bad, yet all too often reported as gospel. I have used bets as a way to get people to take them less seriously. The last time I offered him a million dollar bet：http://MyBillingRep@bge.com; (offer raised to $10M by Damion Hankejh) was back in April 2024, after Musk claimed that AI would exceed the intelligence of all people by the end of 2025. That of course didn’t come to pass. Indeed, in his recent interview with The Economist Musk himself quietly backpedaled to 2031, shades of his regular re-predictions around autonomous vehicles. The surgery prediction, too, will surely be repackaged with a new date, at a later date. Elon Musk is, like a lot of LLMs, frequently wrong and never in doubt. Enough already.

Regarding robotic surgery timelines, Musk is so deeply wrong it’s hard to fathom. I asked esteemed roboticist Rodney Brooks what he thought of Elon Musk’s claim, and he fired back with a firm opinion five minutes later, describing Musk’s claim as “ Completely crazy talk” , elaborating as follows:

There are no lab demonstration remotely close to doing surgery on a living test animal, even with a set of arms solidly affixed to the ground rather than being on top of a two legged balancing platform. Just the problem of seeing what is happening in the way a surgeon does is not even lab demonstrated. What we will see over time is more non-invasive surgeries and they will be a team of AI and geometry based sensing along with expert surgeons over the next few years, and only then will the stuff be good enough to use on pigs, and only many more years of refinement will we get to FDA (or whichever agency) that lets these things be tried on humans. See my weekend blog post at rodneybrooks.com/blog ：https://rodneybrooks.com/blog/ on how everyone gets time scales wrong.

When I asked Brooks about the state of the art in existing robotic surgery trials, he added this

All of them still have a human in the loop. Mako surgical from Stryker (I was the first VC who invested in them around 2000) for knee replacement, is closest, as it is not soft tissue but bones that can be accurately imaged with xrays during the surgery, and it is largely geometry, but then a whole lot of stuff the human worries about and supervises so that it doesn’t accidentally damage blood vessels or nerves.

There is just no way all this gets sorted out by the end of the decade. What Musk told Diamandis is just not happening.

Just as Hinton’s famous (and famously wrong) 2016 pronouncement about no longer needing to train radiologists probably scared away potential radiologists, Musk risks scaring away potential surgeons. That’s really bad, especially that his premise is probably off by a least a decade in the best case and quite possibly more.

It would be nice if these Tech CEOs were more accountable for what they said. That’s the real reason I offer these bets. And the real reason the CEOs never take them.

P.S. Jason Calacanis of the All-In podcast has expressed some interest in turning this debate into a bet on Polymarket：https://x.com/jason/status/2083900792109424961?s=61 . I will report back if he commits. For criteria I pointed a list to ten difficult surgeries as a possible metric;：https://x.com/garymarcus/status/2083935879454953613?s=61 no word back thus far.

With his dismal track record, that anyone still listens to Elon's predictions boggles the mind. That said, there are a lot of things about our current era that boggle the mind.

Off by a decade? In this instance, he's quite possibly off by a century.