Elon thinks AI could become humanity’s uber-nanny: Excerpts from a dinner convo

After months of moving fast and breaking things at Twitter, Elon Musk’s been on a crash course of a different sort in the last several weeks, doing the rounds of interviews across a range of media events and with different media platforms (although TC’s invite may have gotten lost in the mail…).

Yesterday, it was the turn of the WSJ, which featured him as its headliner to close off the opening night dinner of their CEO Council conference in London. Dialing in over video, showing up late for his half hour slot, Musk then proceeded to talk for more than an hour about Twitter (or “X slash Twitter” as Musk tried to call it), AI, Mars and more.

The interview kicked off just minutes after it was announced that Musk would be interviewing Ron DeSantis on Wednesday (today) on Twitter’s Spaces audio streaming platform, and that DeSantis would be making, in Musk’s words, “quite an announcement” as part of it.

Everyone widely expects that announcement to be about DeSantis officially entering the race for U.S. President, which prompted Thorold Barker, WSJ’s EMEA editor, to ask Musk if he was endorsing DeSantis.

Musk avoided a direct yes or no answer, which speaks to something else about him: he likes to talk in circles and his strategy sometimes goes that way, too. Musk replied that he sees Twitter as a “town square,” where he wants everyone to participate. Yes, he wants to bring more audience to the platform at a time when so many appear to be running away; but he also wants to position Twitter as a new media platform in its own right.

Musk’s actions and words should sound extremely familiar. In fact, it often feels like he has spent a lot of time tearing apart the company only to land on an idea that is a repeat of what it was already trying to do: build a media company, which he is now doing — complete with a new CEO with advertising chops who he’s poached from a media company, a strategy to court advertisers, and build entertaining content.

“Ranging from the left, moderate, to what’s considered right… I do think it’s important that Twitter be about the reality and the perception of a level playing field, a place where all voices are heard, and where there’s the kind of dynamic interaction that you don’t really see anywhere else,” he said. “I mean, today on Twitter, for example, AOC and Ted Cruz got into an argument, which was, independent of which side you agree with, very entertaining.” (Cue audience laughter.) “I [would] really just like someone fairly normal and sensible to be the President,” he added later, to more chuckling from the audience.

For Musk, the Twitter story he wants to tell — or at least hopes to sell — is that Twitter is on the return path from the layoffs, internal restructuring, and big changes. And that it’s entertaining.

It’s hard to know how close that story is to reality at this point.

In the meantime, and before investors run out of patience, Musk remains quite exasperating and compelling in equal measure. As Musk’s half hour conversation stretched to more than one hour, he ranged from talking about ambitions to make Mars a self-sustaining civilization through to who might succeed him at any of his companies; the political situation in China through to the future of humanity.

There aren’t many people who would be asked this range of questions in one sitting, and even fewer who would be interested to hear their answers. Yet that’s what you get with this guy. Here’s a selection of excerpts from the conversation:

Musk says he has only one part-time assistant. 

To an audience of CEOs and other executives “scheduled within an inch of their lives,” as Barker put it, Elon described the chaos of being at the helm of three different companies, Twitter, SpaceX and Tesla. (No wonder he dreams about building X, an everything app: it could help compress some of the work and things he has to oversee.)

“My days are very long and complicated, as you might imagine, and there’s this great deal of context switching… Switching context is is quite painful. But I do generally try to divide companies so it’s predominantly one company on one day, so today’s a Tesla day for example, although I might end up at Twitter late tonight… Time management is extremely difficult. And this is going to sound pretty strange but I I only have one part time assistant… but I do most of the scheduling myself. And the reason is because it’s impossible for someone else to know what the priorities are. So since the most valuable thing I have is time, I schedule it myself.”

Musk believes civilization is less robust than people assume, and that AI might accelerate its destruction. 

For someone who is time poor, it’s interesting how bearish Musk has come out on AI. He touts the AI built inside Tesla as the best in the world, and he was an early parter at OpenAI, but he’s not a fan it seems of artificial general intelligence, not least because it’s coming and seems to be out of control. “I think that that is it’s unnecessary for everything, but it is happening and happening very quickly. There is a risk that advanced AI either eliminates or constrains humanity’s growth.”

But don’t take any of that too seriously! Much later in the same conversation he contradicted himself to say, “I don’t think AI is going to try to destroy humanity, but it might put us under strict controls,” describing a scenario of “AI assuming control for the safety of all the humans and taking over all the computing systems and weapon systems of Earth and effectively being like some sort of uber-nanny.”

Musk claims that Twitter has gotten rid of hate speech and 90% of the “scam and spam entrepreneurs” on its platform, which will make it more attractive to advertisers. 

The pair touched lightly on Musk’s “courtship” of Linda Yaccarino, who was named as Twitter’s new CEO last week (she has yet to step into the job). It sounded like their interaction first started as a conversation about advertising, but Musk didn’t get into the specifics of how, and when, that turned into a recruitment exercise. Instead, he shifted direction to talk about how the company’s ailing ads business is coming out of rehab.

“Linda, felt that it would be very helpful for advertisers to see me in person so invited me down to a conference in Miami, which was very helpful,” he said of his appearance on stage with Yaccarino several weeks before he named her CEO.

He said there, he “met with a number of advertisers personally to assure and show them that Twitter is a good place to advertise.”

He claimed that “hate speech has declined” and that “the quality of the system, especially with respect to scammers and spammers, is dramatically better than it used to be… We’ve gotten rid of, at this point, well over 90% of scam and spam entrepreneurs. It should be quite rare at this point that you see a scam.”

He also said that its figured out how to shut down bot farms, presumably in aid of improving authentic sentiment on the platform, although he didn’t spell out why these exactly were a problem for advertisers, if for example they were brigading in favor of a brand?

He says that he and the new CEO will work together over moderation.

Musk has been quite outspoken before about free speech, and specifically him saying what he wants to say, which doesn’t sound like it will change going forward. While “the general principle is that will hew close to the law. So for any given country, we will try to adhere as close to to law as possible,” the company is working on “adjacency controls that ensure that “if you’re Disney for example…the content nearby [will be] family friendly. That’s totally understandable.

MDAU wasn’t a great metric.

The problem with “monetizable daily active users,” Twitter’s self-defined metric that it adopted when still a public company to describe its audience and how it was growing, was, Musk said, that “a bunch of those users… would see a notification on their phone, about a tweet, but they wouldn’t actually click through to the site. Certainly what really matters is true user seconds of screen time.” That is what the company tracks now, he said. “That’s based on the screen time as reported to us by iOS, Android and the browser.” He says the feedback from advertisers for this measurement has been positive.

Is there a limit on free speech — at least Elon’s free speech?

The topic of George Soros came up, specifically Musk’s characterization of him and the huge amount of outcry and argument that it generated. Elon really does not feel he should be reined in.

“I’m not going to mitigate what I say because that would be inhibiting for your freedom of speech. That doesn’t mean you have to agree with what I say… The point is to have a divergent set of views. Free speech is only relevant if it’s speech by someone you don’t like. Is that allowed? If so, you have free speech. Otherwise you do not. And for those who would advocate censorship, I would say it is … only a matter of time before the censor gets turned on you.” Easy to say, harder to see played out.

Twitter will be hiring again.

The conversation then veered into a totally different area, to the subject of Twitter staffing — unspoken questions about moderation and who is left to handle it being possibly the bridge here? With the company now down to about 1,500 employees from the 7,500+ it had when he took over last year, there are indeed a lot of questions about how it can recover, build and grow. There will be more people added, he said, not least if it plans to revive that advertising business and court big advertisers. Oh, and build that “everything” app.

“We are going to start adding people to the company, and we have started adding some number of people to the company,” he said. “And, but it’s still there’s still a lot of change to have to happen. So but I think 1,500 is probably a reasonable number.”

On Twitter’s valuation.

“All’s well that ends well,” was his short answer to whether he regretted buying Twitter. On the ambition/projection that the company could be worth $250 billion after he bought it for $44 billion, he described Twitter as “on the comeback arc.” He didn’t have an answer for whether Twitter could go public again, nor whether it would even stay headquartered in San Francisco.

Content moderation: bad; AI regulation: good

Musk is okay with people speaking their minds, but he is less okay with AI doing the same.

“I’ve been pushing hard for a long time and met with a number of senior senators and congressmen, people in Congress in the White House to advocate for AI regulation, starting with an inside committee that is formed of independent parties as well as perhaps just bids from the leaders in industry,” he noted.

One area that he’s particularly concerned with is the presence of AI in social media. On the subject of the weaponizing of AI, “the pen is mightier than the sword. So one of the first places where to be careful of AI being used is in social media, to manipulate public opinion.”

He said this was one of the reasons he wants to turn Twitter into primarily a subscriber-based system, is “because it is dramatically harder to create,” and thus more unlikely that it would not be an authentic person, or so the thinking goes. “It’s like a 10,000 times harder to create an account that has a verified phone number from a credible carrier that has a credit card and that pays a small amount of money per month. And have those credit cards and phone numbers be highly distributed, not clustered, incredibly difficult. So whereas the past someone could create a million fake accounts for a penny apiece and then manipulate or have something appear to be very, very much liked by the public when in fact it is not, or promoted and retweeted, when in fact it is not. Popularity is not real, it’s essentially gaming the system.

“So the bias towards a subscription based verification, I think, is very powerful. And that really, you won’t be able to trust any social media company that does not do this, because it will simply be overrun with bots to such an extreme extreme degree.”

Elon thinks AI could become humanity’s uber-nanny: Excerpts from a dinner convo by Ingrid Lunden originally published on TechCrunch