120 人对 6 万人。
2020 年底,Neuralink 的员工规模大约 120 人,特斯拉超过 6 万人。前者被它的创始人称为“很小的公司”。
2020 年 12 月,马斯克到柏林领 Axel Springer Award。颁奖礼上主持人问他,同时管着好几家公司,精力怎么分配。他先说了上面那句,接着提到 Boring Company,说那家也很小。
Neura Pod 这期节目的主播对这句话的评价是“安静自信又过分谦虚”。评价挺准。99% 的人不敢去创办一家 Neuralink 这种体量的公司,但在马斯克自己的坐标系里,它确实还处在实验室规模。
这个奖的分量
Axel Springer 官网对这个奖的定义是:颁给“特别具有创新性、创造并改变市场、影响文化,同时直面自身社会责任”的杰出人物。之前的获奖者里有扎克伯格和贝索斯。同一年,马斯克还被《财富》杂志评为 2020 年度商业人物。
主播在节目里说,看到马斯克拿到更多他应得的认可,挺高兴。这句话带着明显的支持者视角。这期节目的立场是粉丝向的,但信息密度不低,值得认真拆一遍。
颁奖礼当天的访谈覆盖了马斯克名下几乎所有公司。Neuralink 只占其中一个片段,但那个片段里有今年最实在的一句话:第一批植入人体的设备,打算给谁用。
第一批植入者:四肢瘫痪的病人
马斯克的原话是这样的:
“短期到中期,Neuralink 主要是帮助治愈脑损伤。我认为第一批植入人体的设备会给四肢瘫痪的人、截瘫的人,让他们只靠思想就能控制电脑或手机。”(马斯克)
这句话里有两个信息,很多人只注意到第一个。
第一个是适应症。四肢瘫痪,包括颈髓损伤导致的四肢功能丧失。这类病人数量不大,但需求极端明确,而且现有方案(眼动仪、头戴式开关、语音识别)都有明显天花板。用思想直接控制光标,等于绕过了所有外围设备。
第二个是“短期到中期”。这个词放在 2020 年底,含义是人体临床试验已经在射程之内。2020 年 8 月那场发布会上,Neuralink 展示过植入设备的猪,活蹦乱跳,实时读取神经信号。从猪到人,剩下的是审批和安全性验证。
主播在节目里描述了马斯克常说的一句话:人应该有一些值得醒来、值得期待的东西。
Neuralink 能在大脑之间收发足够多的神经信息,帮瘫痪病人驱动假肢,这个未来让他兴奋。不止如此,还能帮助消除阿尔茨海默症、帕金森病的主要负面效应。
他写了一段很个人的话:一个孩子面对认不出自己的祖父母,那种挫败感会把人击垮。一个这种挫败感不再存在的未来,让人振奋。
这段话不技术,但它解释了为什么有人愿意为这家 120 人的公司熬夜。技术叙事最终要落到具体的人身上,否则就是 PPT。
颁奖礼变成了特斯拉发布会
整场访谈的重心其实在特斯拉。柏林超级工厂当时在推进,主持人问的都是产能、欧洲市场、监管这类问题。
Neura Pod 是个只讲 Neuralink 的频道,主播的处理方式很干脆:特斯拉那段建议听众去听 Rob Maurer 的 Tesla Daily。垂直播客各管一段,这个分工在 2020 年已经很成熟了。
不过他还是把其他几条业务线的进展捋了一遍,因为这几条线跟“马斯克把精力放在哪”直接相关。
SpaceX 那边,Starlink 开始产生更多头条。节目给出的判断是,未来几年内用高速稳定的网络连接全球人口,可能性越来越大。Starship 是 SpaceX 运力最强的飞行器,首发时间就在录制前后几天。
Boring Company 的隧道和砖块,主播猜可能在为后面更大的建设项目打地基。这句话节目里也标明了是猜测,不是消息。
访谈最后,马斯克按惯例致谢。这段值得抄下来:
“我不想只是笼统地感谢特斯拉、SpaceX、Neuralink、Boring Company 的优秀同事。我真的代表你们接受这个奖,为了你们做的所有了不起的事。非常感谢。但我特别想强调,是公司里那些人让这一切发生。”(马斯克)
一个同时管六家公司的人,在领个人奖项时说“代表你们接受”,这套话术练了很多年。但放在 2020 年底的语境里,它也是实话:特斯拉产能爬坡、SpaceX 载人飞行、Neuralink 动物实验,没有一件是他一个人干出来的。
搬去德州:五条线索摆在一起
颁奖礼之外,12 月还有一个传闻:马斯克要搬去德州。
主播没有直接下判断,而是把线索一条条列出来。
- 大多数星期里,他已经在德州 Boca Chica 发射场待一两天。
- 奥斯汀东侧在建一座巨型工厂,规模大到建成后可能被叫作 Terafactory,而不是 Gigafactory。
- Neuralink 官网挂着一个招聘岗位:head of construction,建设负责人,工作地点写着 Austin, Texas 和 Fremont, California。这条最关键,因为它意味着 Neuralink 要在奥斯汀建东西。
- 2020 年 5 月初,他发推说自己“几乎卖掉了所有实物财产”。
- 一周多之后,他公开表达对阿拉米达县和加州部分政策的不满。
第三条值得多说一句。一家 120 人的公司专门招一个建设负责人,说明它接下来的瓶颈在场地,不在人。脑机接口的实验需要动物房、手术室、洁净间、信号测试区,这些都不是租个写字楼能解决的。选址跟着马斯克走,是顺理成章的事。
第五条引出了那条著名的推文。
5 月 9 日那条推文
“坦率地说,这是最后一根稻草。特斯拉现在将把总部和未来项目立即搬到德州、内华达。如果我们还保留弗里蒙特的制造业务,那取决于特斯拉未来受到怎样的对待。特斯拉是加州剩下的最后一家汽车制造商。”(马斯克,2020 年 5 月 9 日)
主播的判断很直接:按后来的走向看,这明显是虚张声势。或者说,按马斯克一贯的方式,属于“半虚张声势”。他脑子里大概同时转着几十件事,说出口的那句只是其中一件。
但主播也承认,马斯克很可能真的相信德州比加州给出更多“自由”。这个“自由”加引号,是主播在保持距离。
然后是税。这部分是整期节目最干货的一段。
州所得税:加州最高,德州为零
加州的州个人所得税最高档全美第一。德州没有州个人所得税。
企业税这边,德州的标准营业税率大约 1%,加州 8%(实际税率 8.84%)。差 7 个百分点,对制造业是实打实的成本。
主播的判断比一般人想的更冷静:以马斯克的薪酬结构,个人所得税的档位对他几乎没有影响。他的财富主要来自股票期权,工资只占很小一部分,税率算法跟普通上班族不是一回事。真正有分量的是企业税那一块。
接着他给了一个更聪明的结论:马斯克大概率会把主要居住地报在德州,拿到税收好处,然后继续他的“游牧”生活。洛杉矶、弗里蒙特、内华达的 Sparks、柏林、上海、奥斯汀、佛罗里达的卡纳维拉尔角,一年到头来回切。也许明年这个时候,名单上还要加上中西部超级工厂,或者澳洲、非洲。
这个判断的价值在于,它把一个看起来是“忠诚度”的问题,还原成了“税表填在哪”的问题。企业家搬不搬家,很多时候不是感情决定,是会计决定。
AlphaFold:一个悬了五十年的问题
节目后半段换了个话题。2020 年 12 月,DeepMind 宣布在“蛋白质折叠问题”上取得突破。这是整期节目里唯一一条跟马斯克无关、但跟大脑科学直接相关的新闻。
蛋白质折叠问题是什么?所有生物的细胞都由三维结构的蛋白质构成。科学界早就知道,这些三维蛋白质可以展开成一串氨基酸。反过来走不通:给定一串氨基酸,怎么预测它会折叠成什么三维形状。
这个“反过来”卡了几十年。药物设计、疾病机理研究,很多环节都受制于它。你知道了基因序列,却不知道它造出来的蛋白质长什么样、在哪里出问题,后面的工作就很难往前推。
DeepMind 的系统叫 AlphaFold。它分析了数百万个蛋白质,在预测病毒蛋白三维结构上,达到了一个显著的准确度门槛。
DeepMind 官网首页的第一句话写着,他们要“研究和构建安全的人工智能系统”。公司的目标是破解智能,为所有人推进科学发现。这句话放在一家做游戏 AI 出身的公司身上,五年前听起来像口号,2020 年底听起来像路线图。
新闻发布视频里,一位团队成员这样描述影响:
“影响是巨大的,从疾病如何进展,到怎么发现新药,没有尽头。”(DeepMind 团队成员)
主播的解读是,这属于整个生物学与技术携手过程中的一个中间环节。AI 和机器学习让这个领域对大脑和身体的科学知道得更多;工程师一边帮着找数据,一边用上已有的数据。能更准确地预测疾病如何进展,就能开发出新的预防方案,针对阿尔茨海默、帕金森、痴呆这些脑部疾病。
消息发布后,马斯克在推特上向 DeepMind 道贺。
这条新闻跟 Neuralink 的关系,表面看不大,往深了看是同一件事的两头。Neuralink 要解决的是“怎么读和写神经信号”,AlphaFold 要解决的是“怎么从序列推结构”。两个问题都需要把生物系统变成可计算的数据。一边是工具,一边是算法,缺一头,另一头都走不远。
CNBC 泼的那盆冷水
同一个月,CNBC 发了一篇文章,标题是《马斯克的脑机接口公司 Neuralink 有钱有热度,但坎也不少》。
发布时间是周六。主播对这个时机的评论挺直接:找不到什么由头,也许只是想写点马斯克、又不写特斯拉和 SpaceX 的东西。
这个评论有点刻薄,但说中了科技媒体的一条潜规则:马斯克这个名字自带流量,只要沾上,标题就能起。
文章的核心对比是:特斯拉和 SpaceX 马斯克都熬过来了,Neuralink 面临的科学难题不一样。理由是,神经科学家对大脑的了解还有大片空白。
匹兹堡大学的一位教授在文章里说:
“火箭科学和电动车背后的科学,都已经写清楚了。大脑是这些东西开始失效的地方。脑科学大部分是未知的。如果你不懂大脑怎么工作,很难用工程超越大脑。”(匹兹堡大学教授,转引自 CNBC)
这段话的逻辑很清晰:前两家公司的难题在工程,Neuralink 的难题在科学。工程有明确的目标函数,科学没有。
主播对这段话的反应很强烈,说“我不能再不同意了”。
工程不必等科学读懂大脑
主播的论证只有两句,但很硬。
不需要完全理解科学,才能让工程有用。工程建立在已知的东西上,也能带来科学发现。下一代工程方案,再建立在那个时候已知的科学基础上。
换句话说,工程和科学是互相推着走的,不必等科学全部读完再开工。
这个观点在科技史上有很多例子。蒸汽机发明的时候,热力学还没建立;飞机上天的时候,空气动力学还在补课。先做出能用的东西,再从使用中总结规律,是工程的常态。
主播还从商业角度补了一句:恰恰因为这个,Neuralink 有可能比特斯拉和 SpaceX 更快地甩开整个行业。
这句话值得推敲。特斯拉的难题在制造和成本,工程可以硬推,一条产线一条产线调。SpaceX 的难题在可靠性和复用,也是工程问题。Neuralink 的难题在生物体,需要科学和工程交替推进。交替推进的好处是,每解决一个工程问题,就多一份数据;每多一份数据,就反过来推进科学。循环转起来之后,速度会比单向推进快。
这也是为什么那 120 个人里,工程师和神经科学家要坐在一起。只懂一边,循环转不起来。
节目最后,主播引用了 Neuralink 早期团队成员 Philip Sabes 博士的一段话,说这是他最认同的部分。
“能在细节上理解大脑的起源和神经元的组织方式,达到让神经科学家彻底满意的程度,那是另一个问题。我们不需要解决所有那些科学问题,才能取得进展。”(Philip Sabes,Neuralink 早期团队成员)
这段话是整期节目的落点。它把“科学未知”从一个障碍变成了一种前提:正因为未知,才需要更好的工具去测量;工具本身会带来新的已知。
反过来说,如果非要等到神经科学家对大脑“彻底满意”才动手,那这件事永远不会开始。
2020 年 12 月的 Neuralink,处在这样一个位置:公司很小,目标很大,科学有一半是黑的,工程已经跑到了科学前面。这种状态在生物技术领域不算罕见,罕见的是它背后站着一个愿意把 120 人放在 6 万人旁边、还说“这也很小”的人。
如果想继续跟进,这几件事可以做
- 把 Neuralink 官网的招聘页面加进收藏夹。招聘岗位比新闻稿更早暴露战略意图,奥斯汀那个建设负责人岗位就是例子,它比任何“搬去德州”的传闻都实在。
- 读完 CNBC 那篇原文,再读 Philip Sabes 的完整表述。二手转述容易把“科学未知”和“工程不可行”混成一件事,原文里这两件事是分开的。
- 去 DeepMind 官网看 AlphaFold 的技术说明,别只看新闻稿。重点看它在什么指标上达到了什么水平,以及他们自己承认的局限在哪。
- 判断“马斯克搬德州”这类传闻,别盯推特,盯三样东西:公司注册地变更、高管住址申报、以及新工厂的选址文件。推文是表态,文件才是决定。
- 关注 Neuralink 后续的人体临床试验申请进展。从动物到人这一步,审批时间比技术时间更能决定时间表。
- 如果身边有人把脑机接口当成科幻,把 2020 年 8 月那场发布会和这次访谈连着看一遍。三只猪的实时信号,加上“第一批给四肢瘫痪病人”的明确表述,已经足够把话题从科幻拉回工程。
0:00In short-medium term, Neuralink is really just going to help cure brain injuries. Hey everyone, welcome back to Neuralink. There's been quite a bit of news related to Elon, and he recently made some comments about Neuralink. So, I figured I'd put together another update style episode.
0:19This one's going to include info about Elon's recent statements about Neuralink while accepting an award in Germany. And a rumor he's going to move to Texas. I'll talk about a new breakthrough made by DeepMind. And lastly, I want to share some thoughts on an article from CNBC saying that Neuralink has hurdles to overcome.
0:38Elon Musk accepted the Axel Springer Award about a week ago in Germany. According to the Axel Springer website, the award is given to outstanding personalities who are particularly innovative and who generate and change markets, influence culture, and at the same time face up to their responsibility to society.
1:00Elon fits this description pretty darn well. And a couple notable award recipients include Mark Zuckerberg and Jeff Bezos. Fortunately, in 2020, Elon's gotten a little more recognition for his and his team's accomplishments. He was named Fortune Magazine's 2020 Business Person of the Year.
1:18And he's also on every single one of Forbes 30 under 30 list. Okay, that was a bad joke. But, I really am happy to see he's getting more of the respect he should. Throughout the day, the hosts asked Elon a series of questions relating to all of his ventures. I particularly enjoyed them asking Elon simply how he manages all these companies.
1:40His response was a quietly confident and overly humble statement saying that Neuralink's a small company, very small. And when they talked about Boring Company, he said that's also small. Although his responses are true, it's also true that 99% of people would never dare try to create a company the size of Neuralink. Which, if you're a consistent 120-ish employees versus
2:08Tesla's 60,000-plus employee count. Later, he continued his explanation of Neuralink's ambitions with the following. Helped your brain injuries, brain and spine injuries. So, it's like if you if somebody's a I think the first implanted devices in humans will be for quadriplegics, tetraplegics, allowing them to control a computer or a phone just using their mind.
2:30Elon often talks about how we should have things to wake up for and look forward to. For me, this future where Neuralink can send and receive enough neural information to and from the brain to help paralyzed patients move prosthetic limbs, that future is incredibly exciting. Not only is Neuralink working towards this functionality, but they should also be able to help eliminate the primary negative effects of Alzheimer's or Parkinson's disease.
2:58Think about the frustration of a child when their grandparent or parent can no longer recognize them or remember prior experiences. That situation just destroys me. A potential future where this frustration no longer exists, that's inspiring. I hope if you feel the same way, you'll help educate those around you about the potential of the team and the tech at Neuralink.
3:22The conversation primarily revolved around Tesla at the award ceremony, given the Berlin Gigafactory is starting to make more progress. If you're curious about the times Elon discussed Tesla, I recommend listening to Rob Maurer with his channel called Tesla Daily. Of course, Elon's endeavors don't stop there. The Starlink portion of SpaceX is starting to generate more headlines as it's becoming increasingly likely that they'll be able to connect people around the globe with incredibly high-speed and reliable internet within the next several years.
3:55Likewise, the actual mission of SpaceX to establish a multi-planetary species on Mars is also making good progress. The launch of SpaceX's most powerful vehicle, Starship, should occur plus or minus a few days of me recording this episode. Lastly, it wouldn't surprise me if The Boring Company's tunnels and bricks are laying the foundation for other larger construction projects down the road.
4:21At the end of the interview portion of the award ceremony, in typical Elon fashion, he took the time to thank all of the great teams at each of his companies. But I I don't want to just say to acknowledge the great people at at Tesla and SpaceX, Neuralink, Boring Company. really I accept this award on on your behalf for all the great things that that you've done. So, super super appreciated. but I really want to acknowledge very strongly the people at the companies who made it happen.
4:47Okay, so moving on from the award, there have recently been some rumors that Elon's going to move to Texas. This move wouldn't be too surprising given that he's already spending a day or two most weeks in Texas with the Boca Chica launch site for SpaceX. Also, the gigantic Texas Gigafactory and potentially Terafactory by the time it's all finished being constructed is being built on the east side of Austin.
5:13And of course, last but not least, it's highly likely Neuralink is going to be constructing some facility or factory close to Austin as there's a job opening on the Neuralink website for the head of construction. The posting says, "Seeking an experienced, dedicated, and well-organized head of construction to join our facilities team working in Austin, Texas and Fremont, California." Other things that align with the rumor are that earlier this year, at the beginning of May, Elon tweeted about, quote, selling almost all physical possessions.
5:50Just over a week later, he tweeted about his unhappiness with some policies and structure in Alameda County and California. On May 9th, he wrote {quote} Frankly, this is the final straw. Tesla will now move its HQ and future programs to Texas, Nevada immediately. If we even retain Fremont manufacturing activity at all, it will be dependent on how Tesla is treated in the future. Tesla is the last car maker left in California.
6:20Based on how things have shaped up since then, I think it's pretty clear that this was a bluff. Or like all of Elon's bluffs, they're actually semi-bluffs. There are probably dozens of things he's considering, but it wouldn't surprise me if he does really believe Texas allows for more {quote} freedoms than California.
6:39The tax consequences of living and operating in California, where state personal income tax is the highest, are also considerably less favorable than in Texas, where state income tax doesn't exist. I believe, based on how Elon is compensated for most of his work, these tax brackets wouldn't matter much at all. But, the corporate taxes are also significantly lower in Texas at a standard business tax rate of 1% versus the 8% in California.
7:08The question becomes, do these factors, plus traffic and the other factors that I'm not considering, outweigh the good things that California does have to offer? Overall, it probably won't matter too much because he'll probably indicate his primary place of residence, get the tax advantage, and continue his nomadic lifestyle, where he's constantly splitting time between Los Angeles, Fremont, Sparks, Nevada, Berlin, Shanghai, Austin, and Cape Canaveral in Florida.
7:39And maybe by this time next year, we'll add Giga Midwest or Australia or Africa to the list. Now that I spent plenty of time on a rumor, it's time to move on to recent events that have actually already occurred. The first story of note is one that broke out of the company named DeepMind.
7:58They've been recognized for a breakthrough in solving the {quote} "protein folding problem". I'm not entirely sure what the ins and outs of the breakthrough are, but based on my initial research, it looks like this is actually a pretty well-known problem in the scientific community.
8:13And this is the most significant publicly acknowledged breakthrough. The protein folding problem is that cells in all living things are comprised of 3D proteins. And although the scientific community understands that these 3D proteins can be unraveled into a string of amino acids, they haven't known how to go backwards.
8:35So, they don't know how to take the information from a string of amino acids and predict the 3D shape of a protein. This is no longer the case after DeepMind's development of an artificial intelligence system called AlphaFold. In line with the first statement on their company website, DeepMind has been able to {quote} "research and build safe artificial intelligence systems".
9:00Their goal is to solve intelligence and advance scientific discovery for all. The AlphaFold system analyzed millions of proteins and was able to reach a substantial threshold for predicting the 3D structure of a viral protein. In part of the YouTube video press release, a team member referenced the implications of this breakthrough.
9:22He says, "The implications are immense, from how disease progresses, how you can discover new drugs, it's endless." I believe this is one intermediate step of many for how the entire field of biology is working hand in hand with technology. By using artificial intelligence, machine learning, and other techniques, the field is going to become more knowledgeable about the science related to the brain and our bodies.
9:50Simultaneously, engineers can help scientists find the data while also using the data that is available. By being able to better predict how diseases progress, new solutions can be developed for preventing brain disorders like Alzheimer's, Parkinson's disease, and dementia. After their news release, Elon congratulated them on their progress on Twitter.
10:15Next is an article that broke from CNBC. I'm curious about their timing here. This past Saturday, they released an article with a headline, "Elon Musk's brain-computer interface company Neuralink has money and buzz, but hurdles, too." It feels like there is no reason or spark to release an article like this, other than maybe they just wanted to cover something about Elon that wasn't Tesla or SpaceX related.
10:40The article outlines the differences between Elon's prior successes with Tesla and SpaceX versus the upcoming hurdles with Neuralink. My takeaway was that many people think Neuralink will have a tougher time making it through the science-related challenges than Tesla or SpaceX did because, {quote} "There's still so much that even neuroscientists do not know about the brain." A professor from the University of Pittsburgh says, {quote} "The differences rocket science and the science behind electric cars have been written.
11:13However, the brain is where that begins to fall apart. Brain science is largely unknown. It's hard to out-engineer the brain if you don't understand how the brain works." To say that this is the reason the team's going to have more trouble with Neuralink doesn't feel right to me.
11:29I could not disagree more with the idea that you need to fully understand the science to make engineering useful. Engineering is based on the things that are known and can lead to scientific discoveries. Then, the next engineering solutions will be built on the foundation of whatever science is known then.
11:48From a business perspective, I believe that this is the reason they'll be able to leapfrog everyone in the industry even faster than Tesla or SpaceX has. At the end of the article, the author includes a note from one of Neuralink's original team members, Dr. Philip Sabes. I agree with his comments. He says, "Being able to understand the brain's origin and the organization of neurons in fine detail in a way that would satisfy a neuroscientist to the core, that's a separate problem. And we don't need to solve all of those scientific problems in order to make progress." That'll wrap it up for this episode.
12:24Thank you all for listening. Please like, share, and follow @neurapod on Twitter. And if you're feeling extra generous, please also comment below the video any thoughts you have about Neuralink. Hope to have you again next Wednesday. Thanks.
资源与出处
- 视频在 YouTube 收看
- 档案silicon-valley.tech 档案页本页视频与文稿数据来自 Elon Musk Archive
- 精读本篇精读 Markdown 版