Discover more from The White Pill
The White Pill, a New Weekly
the white pill #1 // inspiring, evocative, and excellent developments in tech, engineering, physics, astronomy, space, and medicine this week
Welcome to the first edition of the White Pill, Pirate Wires’ brand new really-cool-sci-fi-shit-that’s-actually-real-and-happening-right-now digest. The aim is to highlight fun, inspiring, and evocative developments in tech, engineering, physics, astronomy, space, and medicine with a regularly scheduled link blast and brief, lead story. Here, because we want this to be an email you look forward to, we’ll never include depressing doomerism, culture war bullshit, or clown links.
Honestly, there’s only so much “bomb the data centers” and “respect my frog/frogself pronouns” a man can take before he inadvertently pills himself into extreme, tribalistic critique. The red pill, I mean. The black pill, I’m talking about. Sure, some critique is necessary. The doomers do need to be given space, and faced. The clown links do need to be shared, laughed at, woven into art. But we also have to remember who we are, and what we actually want: connection, growth, human potential. In other words, white pills.
Brandon, lead editor on the pill, plans to test this digest for a few editions. If there’s enough interest to justify a regular feature, we’ll make it official.
Please enjoy, and remember to blow up the comments below. What did we miss?
First, some excellent news
In a significant medical breakthrough, a new grain-of-rice-sized device developed by the Houston Methodist Academic Institute can be used to deliver drugs to the pancreas to treat pancreatic cancer. The amount of medication required for these devices to be effective is four times smaller than traditional treatment for pancreatic tumors. Excellent news. (Interesting Engineering)
Shout it from the rooftops: Ghana approved a “hugely effective,” “world-changing” malaria vaccine, of which a course has been shown to be 80 percent effective for adults in Burkina Faso. Excellent news. (BBC)
Chemist David Liu has improved upon CRISPR for a second time with “quality editors” which “can theoretically correct 89% of the 75,000 genetic variants associated with diseases.” A previous version of his tool was recently used in London to save the life of Alyssa, a 13-year-old girl with an aggressive form of leukemia. “Her cancer is in complete remission,” Liu told El Pais. Excellent news. (El Pais)
The lead story: AI Truman Show
Are we living in a simulation? The possibility feels counterintuitive, to say the least, but researchers at Stanford and Google are beginning to poke at this question with their paper about a Sims-inspired virtual town (“Smallville”) that they created and populated with 25 ChatGPT-powered artificially intelligent “agents” (self-directed, autonomous AIs — more on this in the AI section at the end of this email). This alone is interesting, though nothing weird or out of the ordinary seemed to happen over the course of the simulation. But a few of the details about how they pulled it off are extremely evocative. Namely, the researchers ‘implanted’ ‘memories’ in the agents, developed a way for the agents to create, retrieve, and act upon memories generated in the simulation, and controlled the agents by using their “inner voices.” Let’s dig in.
The AIs were based on identity descriptions and seed memories: similar to jailbreaks like DAN, where the user tells GPT to assume the identity of an LLM with particular qualities, the researchers started each agent with an identity description and seed memories using prompts such as —
“John Lin is a pharmacy shopkeeper at the Willow Market and Pharmacy who loves to help people. He is always looking for ways to make the process of getting medication easier for his customers; John Lin is living with his wife, Mei Lin, who is a college professor, and son, Eddy Lin, who is a student studying music theory; John Lin loves his family very much; John Lin has known the old couple next-door, Sam Moore and Jennifer Moore, for a few years…”
The AI agents had memory architecture: in addition to seed memories, the researchers developed memory storage and retrieval systems that updated continuously, with which the agents engaged during the course of the sim. Agents in effect created new memories during the simulation, which they used to inform future behavior.
The researchers spoke to the AI via an “inner voice:” the researchers could tell the agents what to do by means of an “inner voice,” the mechanics of which are similar to those which are described in discussions about free will and consciousness —
“To directly command one of the agents, the user takes on the persona of the agent’s “inner voice” — this makes the agent more likely to treat the statement as a directive. For instance, when told “You are going to run against Sam in the upcoming election” by a user as John’s inner voice, John decides to run in the election and shares his candidacy with his wife and son.”
Zooming out: what are some fun possibilities here? Could AI agency + simulations be the basis for:
Improvised sitcoms, movies, etc, fueled by seed memories, identity descriptions, and inner voicing?
Endlessly distinct scenarios differentiated by unique base conditions; e.g. what if you made one of the agents a serial killer? What if the town was surrounded by walls, outside of which were zombies trying to get in?
Simulating a million years, instead of letting it run in human time?
Read the paper here, but first, leave a comment describing what you would do with a sim like this.
Space and astronomy
Let there be light (in a dark universe). A new theory posits that the intersection of gravitational waves in the early universe — just after the big bang — created an effect at the quantum level that resulted in the release of the first photons (particles of light). The theory will be published in May in the journal Physics of the Dark Universe (hell yeah); read the paper online at ScienceDirect. (The Debrief)
The origin of water on earth may have been the process by which primordial planetismal (referring to the celestial bodies that came together to form our planet) magma interacted with molecular hydrogen. Researchers at the Carnegie Institute of Science say that this is a good model for explaining the presence of water on other exoplanets, too. (Futurism)
We’ll need to create many or most earth-like conditions for deep space exploration, starting with 1G of gravity and a social environment that humans are used to — as opposed to just providing the bare necessities like food, oxygen and shelter — a new paper says. (Gizmodo)
Watch the new SpaceX Starship Mars video that charts Starship’s course from Earth to a Mars colony. Starship may launch as early as next week. (YouTube)
Powered by nuclear energy and built with lunar mud, China plans to build a moonbase in five years. (Time)
Researchers at the Max Planck Institute for Extraterrestrial Physics have developed membrane-based, ultra-thin mirrors that can be rolled up and “precisely reshaped” in orbit for large space-based telescopes. (optica.org)
On Thursday, the European Space Agency launched a satellite that will be the first to orbit moons in the outer solar system. Dubbed JUICE (Jupiter Icy Moons Explorer), it’ll dedicate at least three years to Jupiter’s moons, some of which may hide massive subsurface oceans. (The Conversation)
In Texas, NASA let the press into the 3D printed habitat it created to prepare astronauts for life on Mars. (Guardian)
Is an advanced civilization protecting us from nuclear annihilation? In September of last year, a Russian fighter jet attempted to fire on a manned British reconnaissance unit over the Black Sea, but missed because of an equipment malfunction. Some have speculated that the aliens are protecting us from WWIII (yes flimsy but I Want To Believe). (New York Times)
More in AI
AI agents are here. Self-directed AI — you might have seen people referring to this as agentic AI, AutoGPT, or AI agents — works by continually prompting itself to complete the next task in the multi-step goal that a user gives it. Take a moment before reading the rest of this entry to imagine what self-directed AI might look like in 20 years.
Now, read this incredible piece of microfiction about AI agents. A taste:
Agent Based Corporations, swarms of autonomous intelligent agents form corporate entities in temporary alliances to compete against other swarms. Corporate mergers, hostile takeovers, and liquidations in hour timeframes. Generative Adversarial Corporations are created to simulate competition, maximizing the explored feature space of economic warfare.
The size of our shared world model shrinks further, as Hyper-Personalized Markets emerge around individual people. Hundreds of millions of agents are evolved to target each person, and compete to shape both their desires and the products that fill them.
Intelligent agent authored Predictive Economic Models begin to be shared between agents, and form fully functioning shadow economies between swarms. Vast networks of inter-agent communication networks emerge, with powerful influential agents acting as nodes.
“Access to public safety and demographic data should be democratized.” SanFranciscoGPT.com provides real-time public safety data for San Franciscans accessible via simple natural language searches. (@0interestrates)
There’s a new, seemingly powerful autonomous AI assistant available on a waitlist. “From booking flights and ordering food to researching complex topics and managing your email, it can handle almost any task you’d typically do on the web.” (@mattshumer_)
Rumor has it that Elon is developing plans for a startup that will rival OpenAI. “A bunch of people are investing in it . . . it’s real and they are excited about it,” a source allegedly close to the issue told Financial Times. (Financial Times)
Bonus: 70s space colony art
All courtesy of NASA Ames Research Center.
Touch grass (like that above) this weekend. And email me if there’s something I should include in the next edition of the White Pill: brandon[at]piratewires[dot]com.