The original version of this story appeared in Quanta Magazine.

Among the myriad abilities that humans possess, which ones are uniquely human? Language has been a top candidate at least since Aristotle, who wrote that humanity was “the animal that has language.” Even as large language models such as ChatGPT superficially replicate ordinary speech, researchers want to know if there are specific aspects of human language that simply have no parallels in the communication systems of other animals or artificially intelligent devices.

In particular, researchers have been exploring the extent to which language models can reason about language itself. For some in the linguistic community, language models not only don’t have reasoning abilities, they can’t. This view was summed up by Noam Chomsky, a prominent linguist, and two coauthors in 2023, when they wrote in The New York Times that “the correct explanations of language are complicated and cannot be learned just by marinating in big data.” AI models may be adept at using language, these researchers argued, but they’re not capable of analyzing language in a sophisticated way.

That view was challenged in a recent paper by Gašper Beguš, a linguist at the University of California, Berkeley; Maksymilian Dąbkowski, who recently received his doctorate in linguistics at Berkeley; and Ryan Rhodes of Rutgers University. The researchers put a number of large language models, or LLMs, through a gamut of linguistic tests—including, in one case, having the LLM generalize the rules of a made-up language. While most of the LLMs failed to parse linguistic rules in the way that humans are able to, one had impressive abilities that greatly exceeded expectations. It was able to analyze language in much the same way a graduate student in linguistics would—diagramming sentences, resolving multiple ambiguous meanings, and making use of complicated linguistic features such as recursion. This finding, Beguš said, “challenges our understanding of what AI can do.”

This new work is both timely and “very important,” said Tom McCoy, a computational linguist at Yale University who was not involved with the research. “As society becomes more dependent on this technology, it’s increasingly important to understand where it can succeed and where it can fail.” Linguistic analysis, he added, is the ideal test bed for evaluating the degree to which these language models can reason like humans.

Infinite Complexity

One challenge of giving language models a rigorous linguistic test is making sure they don’t already know the answers. These systems are typically trained on huge amounts of written information—not just the bulk of the internet, in dozens if not hundreds of languages, but also things like linguistics textbooks. The models could, in theory, simply memorize and regurgitate the information that they’ve been fed during training.

To avoid this, Beguš and his colleagues created a linguistic test in four parts. Three of the four parts involved asking the model to analyze specially crafted sentences using tree diagrams, which were first introduced in Chomsky’s landmark 1957 book, Syntactic Structures. These diagrams break sentences down into noun phrases and verb phrases and then further subdivide them into nouns, verbs, adjectives, adverbs, prepositions, conjunctions and so forth.

One part of the test focused on recursion—the ability to embed phrases within phrases. “The sky is blue” is a simple English sentence. “Jane said that the sky is blue” embeds the original sentence in a slightly more complex one. Importantly, this process of recursion can go on forever: “Maria wondered if Sam knew that Omar heard that Jane said that the sky is blue” is also a grammatically correct, if awkward, recursive sentence.

Michael Calore: Recording works.

Lauren Goode: Recording. Yeah.

Michael Calore: Yeah. It’s like when people say, let me film that. You’re not actually filming anything. You’re shooting a digital video.

Lauren Goode: So then if you have a video podcast, are you shooting the podcast? What do you say? Do you say taping, then?

Michael Calore: I think you say recording because it just—

Lauren Goode: Recording the pod.

Michael Calore: Yeah.

Lauren Goode: We’re recording the pod.

Michael Calore: It covers all the bases.

Lauren Goode: We’re capturing it.

Michael Calore: That’s what we’re doing.

Lauren Goode: We’re sublimating it. All right. Well, should we record this pod?

Michael Calore: I would like to, yes.

Lauren Goode: Let’s do it.

Michael Calore: Honestly, I’m still recovering from last week’s Big Interview event. My throat is still feeling a little bit raw, even though it’s been like four or five days.

Lauren Goode: You sound delightful to me.

Michael Calore: Thank you.

Lauren Goode: But that really was an epic event.

Michael Calore: It was.

Lauren Goode: Yeah.

Michael Calore: You were on stage.

Lauren Goode: I was. I was first up in the morning. Katie, our boss, gave the intro to the conference and then it was me and Lisa Su, the CEO of AMD. And not only was it a really interesting conversation, but then I was done for the day. I didn’t have to do any more interviews after that. And I just got to listen and absorb, and there were some other really great talks.

Michael Calore: There were, yes. And we’re going to talk through some of them. We’re also going to listen to your conversation with Lisa Su, and then we’ll talk about it, and we’ll take listeners behind the scenes of The Big Interview.

SpaceX is planning to raise tens of billions of dollars through an initial public offering next year, multiple outlets have reported, and Ars can confirm. This represents a major change in thinking from the world’s leading space company and its founder, Elon Musk.

The Wall Street Journal and The Information first reported about a possible IPO last Friday, and Bloomberg followed that up on Tuesday evening with a report suggesting the company would target a $1.5 trillion valuation. This would allow SpaceX to raise in excess of $30 billion.

This is an enormous amount of funding. The largest IPO in history occurred in 2019, when the state-owned Saudi Arabian oil company began public trading as Aramco and raised $29 billion. In terms of revenue, Aramco is a top-five company in the world.

Now SpaceX is poised to potentially match or exceed this value. That SpaceX would be attractive to public investors is not a surprise—it’s the world’s dominant space company in launch, space-based communications, and much more. For investors seeking unlimited growth, space is the final frontier.

But why would Musk take SpaceX public now, at a time when the company’s revenues are surging thanks to the growth of the Starlink Internet constellation? The decision is surprising because Musk has, for so long, resisted going public with SpaceX. He has not enjoyed the public scrutiny of Tesla, and feared that shareholder desires for financial return were not consistent with his ultimate goal of settling Mars.

Data Centers

Ars spoke with multiple people familiar with Musk and his thinking to understand why he would want to take SpaceX public.

A significant shift in recent years has been the rise of artificial intelligence, which Musk has been involved in since 2015, when he cofounded OpenAI. He later had a falling out with his cofounders and started his own company, xAI, in 2023. At Tesla, he has been pushing smart-driving technology forward and more recently focused on robotics. Musk sees a convergence of these technologies in the near future, which he believes will profoundly change civilization.

Raising large amounts of money in the next 18 months would allow Musk to have significant capital to deploy at SpaceX as he influences and partakes in this convergence of technology.

How can SpaceX play in this space? In the near term, the company plans to develop a modified version of the Starlink satellite to serve as a foundation for building data centers in space. Musk said as much on the social media network he owns, X, in late October: “SpaceX will be doing this.”