Readers it’s the 26th issue of the the White Pill, which means it’s this newsletter’s six-month anniversary. A LOT of good stuff in this issue: we get detailed about warp drives in our space section, open the LLM black box in our engineering and computing section, show one reason why Neuralink implants won’t need a tangle of wires sticking out of wearers’ heads in the medicine section, and have some good old fun stuff for you at the end.

Onward.

---

Space

The possibility of a warp drive. Broadly, the hypothetical warp drive works not by propelling the craft at faster than light (FTL) speeds, but by contracting the space in front of it and extending, or rarefying, the space behind it, which results in FTL travel. Imagine kneeling on a rug and reaching to pull the end of it to where your knees are, making the rug bunch up in a wave-like pattern. You’ve decreased the distance between you and the end of the rug without moving your body toward it. Imagine also that you are able to increase the space between each thread in the part of the rug that’s behind you, causing the distance between you and the end of the rug to increase, while your location still is the same. You are now closer to the front of the rug and farther from the end of the rug, while again, never having moved toward the front or away from the end.

FTL travel is, as far as we know, necessary for human travel across many interstellar distances that we would want to traverse. Without it, travel just takes too long. One of the closest systems with an exoplanet in the goldilocks zone — a planet we might want to colonize — is Proxima Centauri b. It’s 4.24 light years away from us, which sounds close, but isn’t. Even one of the fastest manmade objects, the space probe Voyager 2, would take 75,000 years to get there at its top speed.

Physicist Miguel Alcubierre worked out the physics behind the warp drive in 1994. But his solution required “something that anti-gravitated: something like ‘negative energy’ or ‘negative mass.’” That’s because “all of the particles and fields known to exist, even in theory, only have one type of mass/energy: the positive type, and hence they all curve spacetime the same way.” But the warp drive can only work if you can curve spacetime in two different ways, because again, in front of you, space should be contracted, behind you, space should be extended.

Reader, if you’re still with me, I’m sorry to inform you that the payoff for all this reading you’ve been doing — you might not like it. It’s long been hoped that antimatter could be a source for this “negative mass,” and without getting technical, a recent experiment at CERN showed that antimatter doesn’t anti-gravitate. “With that, humanity’s best hope for achieving warp drive has just died.” Bummer.

Regardless, you should read the entire explainer at Big Think.

Mission to Psyche. NASA’s Psyche asteroid mission is set to launch next Thursday, October 12th. It will go into orbit around Psyche in August 2029. Psyche is a large asteroid, 173 miles (278 km) by 144 miles (231 km), located in the main asteroid belt between Mars and Jupiter. What’s interesting about it is it’s thought to be made mostly of metal — part of the exposed core of a long smashed planetesimal, the building blocks of protoplanets. Apart from the scientific curiosity (we could learn a lot about how the cores of Earth and other rocky planets are formed) there is an obvious component of economic interest here, as Psyche could be a valuable target for future asteroid mining operations. (Space.com)

More:

• This image (above) of the Martian surface, captured by NASA’s Mars Reconnaissance Orbiter — which has been orbiting Mars since 2006 — shows dark streaks that looks like fungal mycelium. What are they? While astronomers aren’t totally sure, the current theory is “they’re avalanches involving bright Martian dust moving down slope and exposing the dark underlying rock.” Fascinating. (@ExploreCosmos_)

• A stream of gas about 600 million light years long (imaged above) has been detected using the Keck Observatory in Hawaii. This huge stream of gas feeds into galaxies, and provides fuel for star formation. Let’s see… remembering that it would take someone traveling the fastest speed we’ve ever gone 75,000 years to go just 4.24 light years to Proxima Centauri b, let’s do some back of the napkin math here… 600,000,000 divided by 4.24 is 141,509,434… multiply that by 75,000 and you get 10,613,200,000,000 (10.6 trillion) years to travel the length of the filament… let’s divide that now by 13.7 billion, the age of the known universe… so it would take us 774.7 times the amount of time the universe has existed to travel the entirety of this 600 million light-year structure! Distance and time in space are simply incomprehensible. (Keck Observatory)

• 🚨 BOOK RECOMMENDATION ALERT! 🚨 All this math reminds me of one of my favorite (though partly outdated I believe) physics / astronomy books, The Five Ages of the Universe, which almost exclusively deals in discussion of incredibly deep distance and deep time, but in a way that is actually understandable. Strong recommend.

• Gravitics, a company that “manufactures advanced space station modules,” posted a fun preview (screencaps above) of some of the concepts they’re working on. The video is part of a whole thread of pics and video of their stuff — pretty evocative.

• Elon Musk gave a live talk for the International Astronautical Conference this week, including lots of discussion about Starship. 

• Swedish digital artist Erik Wernquist released a wonderful six-minute space film called One Revolution Per Minute that you should watch with your headphones on and the sound up. In an era where terrestrial hotels are expanding into space, the film would serve as an excellent promotional tool for a space developer seeking business from space hoteliers. But today, it’s ‘just’ a very realistic (I think?) look at what it would be like to visit a torus-shaped space hotel.

---

Engineering, Computing

Opening the LLM black box. The AI company Anthropic, most recently in the news for inking an investment deal with Amazon worth $4 billion, says that it has made a significant step toward illuminating the black box that conceals our current understanding of how LLMs like GPT work. Their announcement is technical so read it if you dare, but — roughly — up until now, our understanding of individual neurons, LLMs’ fundamental computational units, has been quite limited. While some specific neurons have been identified with certain functions (like detecting negative sentiment), most remain a mystery. But now, Anthropic claims to have a method to group neurons together such that it can derive interpretable features. In other words, we were using trial-and-error to put together a puzzle whose pieces had no distinguishing features except for shape; today, we’ve discovered some puzzle piece features that allow us to rely less on trial-and-error for putting the puzzle together.

Why do you care? AI doomerism, a pessimistic position on AI that assumes AGI is, categorically, an existential threat to humanity, is the driving force behind many influential people’s and organizations’ attitudes toward AI. It compels them to call for bans of AI research, or for the government to strictly regulate it, in an FDA-like manner. It isn’t out of the question that their calls will be heard, and their efforts will ultimately be successful. For example, under pressure from Christian conservatives in 2001, President George W. Bush announced federal funding could not be used on embryonic stem cell research, and this restriction lasted until 2009, robbing us1 of nearly a decade of science that has, in the years since, given us advanced treatments for blood disorders, cancers, skin grafts, and corneal diseases, to name a tiny fraction.

Assuming what they’re claiming isn’t just good marketing, and they’ve actually achieved this technical breakthrough, Anthropic’s new understanding of how LLMs work represents a step in the direction of controlling AI, therefore potentially helping us prevent the hypothetical rogue AGI that turns us all into a gray protein paste or whatever (if that is something that will happen). By understanding how the AI thinks, we can trust its decisions more. There’s more potential to this development. Once we understand something, we can often improve it. If we get why certain parts of the AI make mistakes, we can potentially fix those parts. And as we deploy AI in more critical areas (like healthcare, transportation, or finance), understanding their decision-making process is vital to ensure they don't make harmful mistakes.

(Thank you to Founders Fund’s John Luttig for helping me parse this news.)

More:

• Laser rust removal is a process where laser beams remove rust and other contaminants from a surface. It’s based on the principle of laser ablation, where the laser energy is absorbed by the rust and other unwanted materials, causing them to be “blasted” away from the surface. Above is a screenshot of such a process in action, and here’s a video of it. Definitely need the sound on when watching 🤘. (@10x_er)

---

Medicine

• To avoid having to use a tangle of wires that stick out of wearers’ heads, one of the features Neuralink implants will need is wireless charging. Here’s a video of their charger doing just that.

• A new technique from researchers at the University of Oxford has demonstrated that 3D-printed neural cells can mimic the structure of the cerebral cortex (the outer layer of the brain) in mice. This technique, once sufficiently advanced, could be used in the treatment of strokes, trauma, and brain tumors, which damage this area of the brain in humans. (MedicalXpress)

---

Finally, the fun stuff

• Swales are a simple permaculture technique that can be used to ‘terraform’ desert (pic above unrelated, sorry had to include). By holding onto rainwater or runoff longer than a flat landscape would, they allow for better soil infiltration, which allows plant growth and recharges the groundwater table. This technique can transform dry, unproductive areas into thriving ecosystems. All this sounds kinda boring, but this video will convince you it’s not.

• A new paper determines that the preserved human footsteps in White Sands, New Mexico are around 23,000 years old, meaning humans have been in the Americas for at least that long. Traditionally, researchers believed humans hadn’t been here before 16,000 years ago, meaning, among other things, that “humans were present far south of the ice sheets during the Last Glacial Maximum.” 23,000 years ago, the White Sands ecosystem would have been pretty different than today’s — it was a cooler, maybe moister environment with a large lake and a mix of grasslands and woodlands; the fauna included both familiar animals and now-extinct megafauna, such as mammoths, ground sloths, and antique bison. Pretty cool. (Science)

Touch grass this weekend.

-Brandon Gorrell