2023 年 4 月,一场没有画面的采访在 Twitter Spaces 上开播。上百万人在线收听,听到的是两个人越说越僵。一边是刚买下 Twitter 不到半年的马斯克,另一边是 BBC 的科技记者。
整场对话里最锋利的一刀,跟言论自由无关,跟假新闻无关,跟疫苗也无关。
马斯克只问了一句话:你能举一个例子吗?
记者举不出来。采访从这里开始崩塌。
那个被追问了五次的例子
记者抛出的判断很常见:自从马斯克接手 Twitter,平台上的仇恨内容变多了。他补了一句,这是他个人的感受。
马斯克的反应跟数据无关。他要一个样本。一条推文,一个账号,一张截图,任何具体的东西都行。
记者的回答是这样的:我现在不太看那个 For You 推荐流了,因为我不太喜欢它。
马斯克追问:那你既然不看,怎么知道内容变多了?
记者说:几周前我看到过,但我给不出确切的例子。
马斯克把话钉死在这个位置:你说你看到了更多仇恨内容,却一条都举不出来,那我认为你不知道自己在说什么。
接着记者搬出英国的一家研究机构(战略对话研究所,Institute for Strategic Dialogue),说那里有组织认为这类内容在上升。马斯克的回应很短:人们什么胡话都会说。
这段交锋看起来像抬杠,实质是两套取证方式撞在了一起。
一套用「我感觉」「有机构说」「大家都这么讲」。另一套要求可验证的最小单位:一个具体案例。当论据只能落到「我感觉」这一层,对话就失去了继续往前走的轨道。
马斯克自己把停顿说得很直白:我们时间不多,往下聊吧。
谁是错误信息的仲裁者,这句反问撑起了整场采访
记者问了一个在他看来很自然的问题:你是不是把言论自由放在了错误信息和仇恨言论之上?
马斯克把问题的前提拆了。
谁说某件事是错误信息?谁来当这个仲裁者?是 BBC 吗?(马斯克)
记者回了一句很关键的话:你就是 Twitter 上的仲裁者,因为你拥有 Twitter。
这句反驳站得住脚。平台的所有者当然有权力,删帖、降权、封号,每一项都是权力的行使。马斯克没有否认这一点,他换了个角度:一个人的错误信息,常常是另一个人的信息。这个判定权交给谁,才是真问题。
紧接着他主动送出一个让步:BBC 在有记录的一百年里,肯定也发过后来被证明不实的内容。哪怕你立志准确,也不可能次次都对。
这个让步很聪明。它把讨论从「谁更干净」拉到了「谁有资格裁判」。如果没有任何机构能做到零错误,那么把真伪判定权交给任何单一机构,风险都一样。
他的替代方案:法律划线,平台少管
马斯克给出的原则很朴素。
他认为,如果一个国家的民众反感某类言论,应该去找自己选出来的代表,通过法律把它禁掉。他举的例子是:你不能公开鼓吹谋杀某人,这在哪儿都违法。所以言论从来不是无限的。
这套逻辑的落点很清楚:边界由立法机构画,不由平台画。平台不该自己发明一套模糊标准,然后靠内部判断决定谁能说话。
他对言论自由的定义也很硬。
如果你不允许你不喜欢的人说你不喜欢的话,言论自由就是一句空话。而且在失去言论自由的那一刻,它不会自己回来。(马斯克)
这句话在现场没有被打断,也没有被追问。它其实是整场采访里马斯克唯一一次把话说到抽象层面。
四十万个机器人账号和一条没人说出口的因果链
关于「平台上错误信息是不是变多了」,马斯克给了一个非常具体的答案,这个答案指向钱。
他说 2023 年 3 月,Twitter 一个月清掉了 40 万个账号,绝大部分是机器人和诈骗号。
然后他讲了一段没有出现在任何官方声明里的内情。
前任管理层对机器人问题视而不见,因为他们的奖金跟用户增长挂钩。当你的报酬取决于用户数字,你就不会太仔细去看那些用户是谁。(马斯克)
这句话的解释力比任何内容审核政策都强。垃圾账号泛滥,人们习惯归因到技术能力,真正的开关在激励结构。指标定在哪个数字上,那个数字就会被制造出来。
所以他的结论听起来跟舆论相反:现在 Twitter 上的错误信息比过去更少,因为机器人问题被处理掉了。
现场没有人拿出数据反驳这个说法。记者手上是感受和第三方报告,马斯克手上是清理账号的数量和一条激励机制的因果链。
社区笔记:把判定权拆成几千份
马斯克提到的第二个机制是社区笔记(Community Notes)。
运作方式不复杂:普通用户可以在推文下面补充事实性说明,只有当观点不同的人都能认可这条说明时,它才会显示出来。跨立场的共识是显示的前提。
马斯克的说法是,这套机制运转得相当有效。
它绕开了「谁来当仲裁者」的难题。判定权没有被交给一个编辑部,而是被拆散成成千上万个互不认识的用户,还要跨过立场分歧才能通过。
这套机制当然有漏洞,冷启动慢、覆盖有限、容易受到组织化刷票影响。但在「平台自己判」和「没人判」之间,它开出了第三条路。
同一段对话里还有个细节值得留意。记者提到,他采访过被裁掉的内容审核员工,也采访过最近还在做审核的人,这些人的共同说法是人手不够,仇恨言论这块尤其吃紧。马斯克的回应没有否认人手变少,他换的是一整套治理思路:靠人盯内容,永远盯不过来。
采访为什么突然掉头:COVID 标签去哪了
对话进行到这里,记者换了个方向。他问马斯克:Twitter 原来有一套针对 COVID 错误信息的标注规则,后来消失了,为什么?
马斯克的回答只有一句:COVID 已经不是问题了。
然后他立刻把枪口转了回去,连着两句追问。
第一句:BBC 是否为自己在口罩、疫苗副作用上的错误信息负责?是否对这些内容完全没有报道?
第二句:BBC 是否曾被英国政府施压修改编辑方针?你知道这件事吗?
记者的回应是:这不是一场关于 BBC 的采访。我不代表 BBC 的编辑方针。
马斯克接了一句带刺的话,大意是你原来不是这么想的。场面僵住,记者说我们聊点别的,马斯克也说聊点别的。
这段只有一分钟,但它解释了整场采访的火药味从哪儿来。马斯克没有把自己放在被采访的位置上,他在把采访变成一次反向质询。
那两个没有人回答的问题
英国《每日电讯报》(The Daily Telegraph)的报道标题写得很重:马斯克指控 BBC 掩盖疫苗副作用。
回看原话,「掩盖」这个词算不算过度解读,观众自己判断。马斯克问的是 BBC 是否为自己在口罩和疫苗副作用上的错误信息承担责任,是否对这些内容完全没有报道。这是提问,还够不上定罪。
第二个问题更敏感:BBC 是否被英国政府施压,要求修改编辑方针。
报道里提到,BBC 对马斯克的这些说法拒绝置评。
这里的沉默需要琢磨。如果指控毫无根据,一句「纯属虚构」就能解决。选择不回应,等于把解释空间留给了对方。
视频的讲述者、英国退休护理教育者约翰·坎贝尔(John Campbell)在镜头前说得很直接:如果 BBC 真的被政府施压改过编辑方针,那会严重损害英国的民主。他希望这没有发生。
英国媒体内部的两种反应
《每日电讯报》发了两篇东西。一篇是新闻,另一篇是评论,标题叫《BBC 的自由派气泡终于破了》。评论作者对马斯克大加赞赏,说他用犀利的机智揭穿了企业式的集体思维。
BBC 自己的回应是一篇盘点文:《我们从马斯克采访中学到的六件事》。
坎贝尔在视频里说,他读完的感觉是 BBC 在为自己辩护,语气偏批判。他也补了一句:当然,也可能是我这个人比较多疑。
同一场采访,两家媒体的框架完全不同。一家把它写成亿万富翁逼问公共媒体,另一家把它写成六件事,顺便看看这个人有多不靠谱。这场对话的主题是「谁来定义真伪」,而报道这场对话的媒体,正好给出了两种互不相容的定义。
坎贝尔的自述:那个「时间相关」的麻烦
视频里有一段内容是讲述者自己的经历,它不算马斯克的观点,但它是理解这场争议为什么在英国引发共鸣的钥匙。
坎贝尔说,他在接种第三剂新冠疫苗之后,开始需要服用降压药。他用了一个很谨慎的说法:时间相关性。症状出现的时间点,和接种的时间点,是吻合的。
然后他加了一句:奇怪的是,很多人也有同样的时间相关性。他正在收集这一块的证据。
他也提到自己因为发布这些内容,受到过外部机构的压力,收到过相当严重的威胁。他说自己能做的,只是根据看到的证据尽量准确地给出信息,不保证每次都完全正确,也愿意接受反驳。
马斯克那边的说法是:第二剂疫苗之后,他有几天难受得像要死掉一样。
时间相关不等于因果,这一点在方法学上没有争议。同样在方法学上,大量个体的时间相关报告累积到一定数量,就构成了值得正式调查的信号。把前者说清楚,同时不用后者一句话把人打发走,这才是公共卫生沟通该有的严谨度。这场采访之所以刺耳,是因为它暴露出这个中间地带长期是空的。
这场对话真正的信息量在哪里
把三十多分钟拆开看,马斯克其实只说了三件事。
第一,判定真伪的权力不该集中在少数机构手里。因为没有任何机构做到零错误,而一旦同一个机构同时拥有「犯错」和「定义错误」两种身份,纠错机制就自动失效了。它没有动机承认自己错,也没有外部力量能强迫它承认。
第二,平台的治理要靠机制,不靠人手。清机器人、社区笔记、把边界交给法律,这三招都落在机制层面,不需要招聘几千人的审核团队,也不需要有人每天早上决定哪条推文危险。这也解释了他为什么愿意裁掉那么多人还不慌。
第三,媒体自己也该被提问。这是他反复把话题拽回 BBC 的原因。一个以核查他人为己任的机构,如果拒绝回答关于自己的问题,它的核查权威就会被削掉一层,而且削掉的这一层很难补回来。
反过来看 BBC 记者的处境也不轻松。他在一场直播里,被采访对象连续质询自己雇主的编辑方针,而他的职务并不允许他代表雇主发言。他的应对方式是划清界限、转移话题,这在职业上说得通,在直播里显得被动。
至于「平台上仇恨内容变多了」这个说法,最后停在了「我几周前好像看到过」这个位置。这大概是整场采访里最尴尬的十秒钟,也是最诚实的一刻:我们对平台的很多判断,建立在自己推荐流的感觉之上,而推荐流恰恰是被算法挑出来喂给我们的。你看到的比例,和平台上的比例,中间隔着一整套排序模型。
还有一层容易被忽略。马斯克在采访里反复强调「谁来判断」,这个提问方式本身带着一个前提:判断真伪这件事是可以被清晰界定的,只要找到合适的判断者就行。可现实里大量争议内容卡在灰色地带,措辞半真半假,动机含混,利益相关方各有一套解释。这种内容既不该被简单删掉,也不该被简单放过。机制能解决的是流程问题,解决不了判断本身的模糊性。
所以他最后没有赢下什么结论。他只是把一个问题摆在了桌面上:当所有人都同意要打击错误信息的时候,没有人愿意回答由谁来判断。
这个问题到今天依然没有答案。它带来的后果很具体:你看到的每一条被贴上「不实」标签的内容,背后都有一个没有被公开审查过的判断者。这个判断者可能是平台员工,可能是某个机构的评审委员会,也可能是一套没人读过说明书的模型。
采访结束的时候,两个人都说聊点别的。这场对话没有结论,它留下的是一份清单:谁在判断,依据什么,能不能申诉,错了谁来负责。这四句话可以拿去问任何一个正在替你筛选信息的人。
- 下次听到「错误信息」「不实内容」这类词,先问一句:谁做的判定,判定标准写在哪里,有没有申诉通道。三个问题答不上来,这个标签就先别当真。
- 看到任何关于平台内容趋势的论断,先找具体样本。一个账号、一条推文、一张截图。举不出来,就把它归到「印象」这一类,别当证据用。
- 区分三件事:法律禁止什么、平台规则禁止什么、你不喜欢什么。这三条边界经常被混成一团,混在一起讨论必然吵不出结果。
- 看到「某研究机构说」的时候,去找原始报告。看样本量、看时间跨度、看它怎么定义「仇恨内容」。定义方式一改,结论往往会翻个面。
- 遇到个人讲述的不良反应,别急着用「时间相关不等于因果」一句话打发,也别急着当成铁证。记下时间线,等队列研究。两边都别站太早。
- 自己去用一次社区笔记,给一条你熟悉的推文补充事实。你会立刻明白跨立场共识有多难达成,也就明白为什么「平台自己判」这条路走不通。
- 如果你自己也在做内容,把这条当规矩:给出可核查的具体案例,比给出一个响亮的判断更能说服人。马斯克在这场采访里能压住对方,靠的就是这一条。
0:00Well, I want welcome to this video. It's Thursday the 13th of April now I've had an inordinate amount of requests to look at the information contained in this Interview so let's have a quick look at that now Because I'm saying that's what I saw a few weeks ago. I can't give you an exact example
0:18Let's move on we have we only have a certain amount of time well, COVID misinformation You change the COVID misinformation has well BBC changes COVID-19 information? The BBC does not set the rules on Twitter, so I'm asking you... No, I'm talking about the BBC's misinformation about COVID.
0:41I'm just asking you about... You change the labels, the COVID-19 information labels. There used to be a policy, then it disappeared. Why do that? A COVID is no longer an issue. Does the BBC hold itself at all responsible for misinformation regarding masking and side effects of vaccinations and not reporting
1:12on that at all. And what about the fact that the BBC was put into pressure by the British government to change the editorial policy? Are you aware of that? This is not an interview about the BBC. Oh, you thought it wasn't. And I see now why you've done Twitter spaces. I'm not a representative of the BBC's editorial
1:31policy I want to make that clear. Let's talk about something else. We want to talk about the EU too. All right, let's talk about something else. You weren't expecting that. Let's talk about something else. Has the BBC changed its Covid misinformation? I'm talking about the BBC's misinformation about Covid. Questions there from Elon Musk.
1:52Now, let's look at some of the commentary that was been going on about this. This is from one of the leading most respectable papers in the United Kingdom, arguably the most authoritative paper in the United Kingdom. Elon Musk accuses BBC of covering up vaccine side effects. What is that? Putting it a bit strongly? I don't know.
2:12I mean, you've seen the clip, you judge for yourself. Was he actually accusing the BBC of covering up vaccine side effects? Full interviews there, check it out. Surprisingly, surprisingly it seems to be still available. The BBC are still making this interview available. Mr Musk again, does the BBC hold itself at
2:37all responsible for misinformation regarding masking and side effects of vaccinations? not reporting on that at all. That's what Mr Musker said. Now, I must say watching the larger part of this interview, I'm very, very impressed about how well Mr Musker is informed about a whole range of issues. Really quite impressive and I think we can assume he's fairly
3:06well informed. I don't really know who that reporter was from the BBC, but I expect his bosses had a word with him for being perhaps not as prepared as he could be, more and more on that interview later on. And what about the fact again direct quotes that the BBC was put under pressure by the British government? Now Mr Musk seems to have information that
3:29the BBC was put under pressure by the British government, if so that is rather ominous and would totally undermine in my view British democracy. So let's let's hope that's not true. What were the BBC put under pressure to do according to Mr Musk to change the editorial policy? Are you aware of that? Now obviously the BBC
3:55being presumably lily white and squeaky clean on this would come back and say no this is complete rubbish but as far as the Telegraph article was concerned and the BBC declined to comment on Mr Musk's claims. Why? Why wouldn't the comment on Mr Musk's claims? Why on earth not? Mr Musk previously spoken about having major side effects after the second COVID vaccination.
4:26I became aware of my assumed adverse reaction to the vaccine after my third dose. a musk was aware of his after his second dose. And he says this left him feeling as if he was dying for several days. So I really hope he's made a full recovery. If what I consider to be my vaccine injury, I haven't made a full recovery. I need to take blood pressure medications now,
4:56and that just started a few months after my third dose. Temporal correlation, of course, of course, temporal correlation. Many other people seem to have the same temporal correlation strangely enough. More on that on another video when we look at that in here, I'm collecting evidence on that now. So Mr Musker had unfortunate effects there. The BBC's Liberal Bubblers
5:18finally burst, this opinion piece of course. And we can't, yeah, it's an opinion piece by Telegraph writer. Huge thanks to Mr Mus for exposing the corporations group think with such forensic dancing wit. That's just one view. Now BBC have come back of course with a what you might call a counter piece. Six things we've learned from Elon Musk
5:50interview they are talking about. Now I'm not going to go through these things because it felt to me like the, what do I know, I only read it. But it looked to me like the BBC was trying to rather go at Mr. Must To Justify itself. Might be completely wrong, of course. It did seem to be rather critical to me, but there again, I'm just maybe a bit of a suspicious person.
6:15But six things we learned from the interview is what it's called. Now, this is what Mr. Musk said. In March, Twitter said it removed 400,000 accounts in one month alone to help make Twitter safer. Now we believe these are bot accounts because Elon Musk of course is very technically savvy
6:31and you would probably be able to give the instructions to remove these automatic bot accounts and he says this, in my experience there is less misinformation rather than more on twitter at the moment and as I say if you want to see more from the BBC's perspective on that it's in that
6:48article, we always give full links on this channel. Let's listen to a bit more of that interview now just that you might find it interesting. But nonetheless, free speech is meaningless unless you're a lot of people you don't like to say things you don't like, otherwise it's irrelevant. And if at the point of which you lose free speech, it doesn't
7:16come back. I think the issue some people have is that a lot of people were brought back. I mean some people who brought back who were previously banned for spreading things like queuing on conspiracies. You have people like Andrew Tate who were brought back who were previously banned for things like hate speech. Do you think
7:33you prioritize freedom of speech over misinformation and hate speech? Well you You know, who is to say that something is misinformation? Who is the arbiter of that? Is it the BBC? Yeah, you're literally asking me. Yes. Well, no, you're the arbiter. Who is the arbiter? You are the arbiter on Twitter because you own Twitter.
8:00Yes, I'm saying who is to say that one person's misinformation is another person's information. The point of which you say that this is most information, like who is going to this sign? This information can be dangerous, that it can cause real world harms, that it can potentially cause them.
8:16Yes, so the point I'm trying to make is that the BBC itself has at times published things that are false. Do you agree that that has occurred? I'm quite sure the BBC have said things that turn out to not be true. It's whatever it is, 100 year history, I'm quite sure. Yes. Even if you aspire to the accurate, your times won't rule, you will not be.
8:40I think in the grants that you think the VC does aspire to be accurate. You accept there has to be a line in terms of hate speech. I mean, you're not looking at total 100% unrestricted speech. I mean, I generally am of the opinion that if the people of a given country are against
9:01a certain type of speech, they should talk to their elected representatives and pass a law to prevent it. So for example, you cannot advocate murdering someone. That's illegal in the United States, everywhere really, I suspect. So there are limits to speech. I mean, I guess taking your argument to a logical conclusion then, do you accept that there's more misinformation on the platform if it's not being policed in the same way?
9:31I actually think there's less these days because we've eliminated so many of the bots which pushing scams and spam. And previously previous management turned a blind eye to the bots because their bonuses were tied to user growth. And if your compensation is tied to user growth, well, you're not going to look too closely at some of the users. That's part of the problem.
9:55So I think we've got less misinformation because we don't have the bot problem that we used to do. And we also have given a lot of attention to community notes, which corrects with community itself, correct misinformation. It's been very effective. I mean, I would only just add that we have spoken to people who have been sacked that used to be in content moderation.
10:21And we've spoken to people very recently who were involved in moderation. And they just say there's not enough people to police this stuff, particularly around hate speech. in the company. Is that something you would trust? What is your future? I mean, you use Twitter. Right. Do you see a rise in hate speech?
10:38I mean, just a personal anecdote. Like, what do you do? I don't. Personally, my four year, I would see I get more of that kind of content. Yeah, personally. But I'm not going to talk to the rest of Twitter. You see more hate speech personally. I would say I would see more hateful content in that.
10:56Content you don't like or hate for. What do you mean to describe a hateful thing? Yeah, I mean, you know, this content that will solicit a reaction to something that may include something that is slightly racist or slightly sexist, those kinds of things. So you think if something is slightly sexist, it should be banned?
11:15No, I'm not saying anything. I'm saying. Well, I'm just curious. I'm trying to say what you mean by hateful content. And I'm asking for specific examples. And you just said that if something is slightly sexist, that's hateful content. Does that mean that it should be bad? Well, you've asked me whether my feed, whether it's got less or more.
11:37I'd say it's got slightly more. That's what I'm asking for examples. Can you name one example? I honestly don't you. I honestly don't. You can't name a single example. I'll tell you why, because I don't actually use that for you feed anymore, because I just don't particularly like it.
11:50A lot of people are quite similar. I only have one second. second you said you've seen more hateful content you can't name a single example not even one i'm not sure i've used that for the last three or four weeks and i i don't know how did you see the people content content because i've been using a lot of using twice a since you've taken over for the last six months is that you must have at some point seen that you for your hateful content i'm asking for one example right i know i can't give a single one and i'm saying i thought that i'd say so that you don't know what you're talking about really yes because you can't be a single example of hateful content or even one tweet
12:24and yet you claimed that the people content was high that's a false no what i can't be applied what no no what i can was .. there are many.. organizations that say that that kind of information is on the rise now whether the weather has a lot of people have lots i mean i'd write a little bit like this to the.. strategic dialogue .. institute in the u in the uk they will say that So, look, people will say all sorts of nonsense.
12:50I'm literally asking for a single example and you can't name one. Right. And as I've already said, I don't use that feed. But how would you know that? I don't think this is getting anywhere. You literally said you experienced more hateful content and then couldn't name a single example.
13:02Right. And as I said, I have absurd. I haven't, I haven't actually looked at that feed. Then how would you know if there's a single content? Because I'm saying that's what I saw a few weeks ago. I can't give you an exact example. Let's move on. We have, we only have a certain amount of time.
13:15Well, COVID misinformation. You changed the COVID misinformation rule. BBC changes COVID misinformation? The BBC does not set the rules on Twitter, so I'm asking you. No, I'm talking about the BBC's misinformation about COVID. I'm just asking you about, you changed the labels, the COVID misinformation labels.
13:38There used to be a policy, and then it disappeared. Why do that? COVID is no longer an issue. Does the BBC hold itself at all responsible for misinformation regarding masking and side effects of vaccinations? And not reporting on that at all? And what about the fact that the BBC was put into pressure by the British government to change the editorial policy?
14:12Are you aware of that? This is not an institute about the BBC. you thought it wasn't. And I see now why you've done Twitter spaces. I am not a representative of the BBC's editorial policy. I want to make that clear. Let's talk about something else. I've heard from you too. All right, let's talk about something else. Sorry about the
14:31duplication of part of that content that I found myself rather enthralled with the discussion, as I'm sure you did. The reason I put that bit in is there's been so much problem with misinformation and disinformation. And do try and catch that video I've just released yesterday. Or in fact, was it today? With Dr. Seneal Dan, where we've both had pressure from
14:53outside agencies and indeed, well, personally, I've had quite extreme threats. So this is a problem. This is a problem. But all I can say is I try and give the information as accurately as I can based on the evidence that I see and don't always get it completely right. Same as Dr. Dand, he's open to counter-arguments as well as I am.
15:24But it is a problem and let's hope there's been no national collusion in the United Kingdom or the United States or anywhere really. We'll leave that there as a point of reflection. Thank you for watching.
资源与出处
- 视频在 YouTube 收看
- 视频原始视频文件
- 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