🌀🗞 The FLUX Review, Ep. 137
Episode 137 — March 14th, 2024 — Available at read.fluxcollective.org/p/137
Contributors to this issue: Neel Mehta, Boris Smus, Dimitri Glazkov, Ben Mathes, Erika Rice Scherpelz, MK, Justin Quimby
Additional insights from: Ade Oshineye, Alex Komoroske, Robinson Eaton, Spencer Pitman, Julka Almquist, Scott Schaffter, Lisie Lillianfeld, Samuel Arbesman, Dart Lindsley, Jon Lebensold, Melanie Kahl
We’re a ragtag band of systems thinkers who have been dedicating our early mornings to finding new lenses to help you make sense of the complex world we live in. This newsletter is a collection of patterns we’ve noticed in recent weeks.
“Better to trust the man who is frequently in error than the one who is never in doubt.”
— Eric Sevareid
🤝🏠 The profound network value of trust
Imagine a world where engaging a contractor to fix a leaky gutter is straightforward. The contractor arrives, assesses the work, provides an estimate, performs the repair, and charges you exactly as discussed. Simple, efficient, and satisfying for both parties. You can rely on that single contractor to complete the task without follow-ups, fixes, damage, or overcharging.
Compare this to the reality of most home repair work. You expect any individual contractor may be suggesting unnecessary work or overcharging you, so you get multiple quotes with written estimates. You ask a bunch of questions to try to assess if the estimate is honest. Since the contractor needs to schedule when to come out, it will be a couple of weeks before the work gets done. When it finally does get done, there’s a good chance you’ll be overcharged or the work will not be done to the desired level of quality.
Trust minimizes the friction arising from doubt and skepticism. Trust is more than a feeling or a social nicety. It is also the bedrock of efficient economic transactions and the glue holding together the intricate mosaic of societal interactions. This echoes Coase's theory of the firm, which posits that organizations exist primarily to lower the costs associated with doing business. Organizational boundaries are one way of creating trust (within a limited context), but trust lowers transaction costs more broadly.
As our home repair example shows, the costs of low trust are not just theoretical. When there’s a legitimate basis for mistrust, the economic implications are profound—inflated transaction costs, diminished efficiency, and a tangible drag on transaction fluidity. Through this lens, mistrust isn't merely an emotional burden; it’s an economic impediment.
Consider the increased compute necessary to store data on a blockchain. Unlike a centralized and trusted database, blockchains are designed for when there isn’t trust. It’s not an accident many applications of blockchain technology exist in economic niches outside a high-trust society, such as avoiding international laws through money laundering. Blockchains are a technical solution to shared information in low-to-zero-trust environments. Naturally, they’d be used in such low-trust (i.e. illegal) use cases in addition to any positive use cases they unlock from extending trust to places it hasn’t been before, such as key rotation.
The implications of trust extend far beyond the economic sphere. They permeate the fabric of our workplaces, where the absence of trust can lead managers to micromanage, stifle innovation and autonomy, and corrode the spirit of collaboration. In teams, mistrust can derail projects and erode the sense of shared purpose necessary for breakthrough achievements. At a societal level, a lack of trust can lead to political stalemates and the politicization of even the most banal topics. Trust — or the lack thereof — shapes the very architecture of our interactions.
Furthermore, in dense professional networks, high or low amounts of trust change how individuals act and how well the entire network can act. If everyone in your industry is known to lie/cheat/steal at times, everyone has to verify, threaten violence, etc. to make little gains. In an industry known to be trustworthy, you can email someone, ask, and rely on their answer.
You can play this wonderful little game by Nicky Case to see how trust drives emergent cultures: vindictive, cheating, trustworthy, naive, etc. based on who trusts who. In short, shading the truth is essentially the same thing as defecting in an iterated game of prisoner’s dilemma. Any amount of deceit hampers the game, and so people have to start getting defensive and seek retribution lest they get cheated.
We generally avoid normative judgements, but this is one of the clearest “shoulds” we’ve encountered: If you increase distrust in a system by deceit, however subtle, you are forcing huge costs on those around you. You probably shouldn’t.
As we navigate the world, we see cultivating trust is not merely about fostering goodwill; it's about building a more efficient, effective, and humane society. Being trustworthy and reliable isn’t some abstract good, it’s about raising the value of every network you are in. When society seems to take mistrust as a default position, the challenge is to bring trust — justified trust — into our daily interactions. In doing so, we not only enhance our economic transactions but also enrich our collective human experience, proving that trust is, indeed, the currency of a thriving society.
🛣️🚩 Signposts
Clues that point to where our changing world might lead us.
🚏🎨 Researchers jailbroke an LLM using ASCII art
Large language models will usually refuse to answer queries relating to touchy or dangerous subjects like bomb-making or counterfeiting. However, one team of researchers realized they could trick the LLMs into answering questions about such topics by writing out a word with ASCII art, directing the LLM to parse the art into a word, and then talking about the word without explicitly saying it.
🚏🥊 One AI image startup banned employees of a rival company after alleged data scraping
Midjourney, the company behind the popular AI image generator bot on Discord, accused employees at rival firm Stability AI (which makes the Stable Diffusion image model) of trying to scrape Midjourney’s data using “botnet-like activity”; the staffers were allegedly “trying to grab all the prompt and image pairs in the middle of the night.” Midjourney further claimed the scraping had caused a “24-hour outage.” In response, Midjourney said it’d be indefinitely banning all Stability AI employees from the service, along with anyone else they think is using “aggressive automation.”
🚏🥡 Certain “forever chemicals” will no longer be used in food packaging in the US
The US’s Food and Drug Administration has announced “per- and polyfluoroalkyl substances,” or PFAS, will no longer be allowed in food containers like takeout boxes and burger wrappers. PFAS resist water, grease, oil, and heat, which has made them popular for food packaging and cookware, but they also accumulate in the bloodstream and have been linked to weight gain, decreased fertility, immune system harm, and possibly cancer. Manufacturers have been voluntarily phasing out PFAS over the last few years — they phased out the toxic chemicals faster than expected — but it may take a year or two before the remaining product stocks are used up.
🚏🌤️ Weather forecasts have gotten a lot more accurate, but developing countries need better access to them
Thanks to improvements in data collection and computing power, weather forecasters have gotten substantially more accurate since the middle of the 20th century: “a four-day forecast today is as accurate as a one-day forecast 30 years ago.” The improvements have been especially noticeable in longer-term forecasts, such as those looking 5 to 10 days out. Still, experts say that we need to work harder to make these advancements available to people in the developing world, who need the data for their jobs (since more people there work in agriculture) and are at higher risk of climate change-related catastrophes.
📖⏳ Worth your time
Some especially insightful pieces we’ve read, watched, and listened to recently.
When Robustly Tolerable Beats Precariously Optimal (Amanda Askell) — Argues that, especially when the stakes or cost of failure are high, it’s desirable to find a solution that’s “robustly tolerable”: something that works pretty well in a wide range of situations. While such systems will never be optimal on any one axis, they also have far lower downside risk, a fact that’s often under-appreciated.
The Munger Operating System: How to Live a Life That Really Works (Farnam Street) — Eclectic advice from the legendary investor: be reliable, don’t wallow in self-pity but be tolerant of weakness in others, avoid being part of a system with perverse incentives, and work with and under people you admire and trust.
What Should We Have Learned From the Collapse of the New Economy (1998–2000)? (Dave Karpf) — Draws striking comparisons between the dot-com era (praised as the “New Economy,” a term nobody could clearly define) and the recent crypto/Web3 boom and bust. Concludes that “the excessive spending, overheated rhetoric, and bad financial advice” of the late 1990s continue to play out with each subsequent tech hype cycle.
A Few Laws of Getting Rich (Collab Fund) — Shares some lesser-known disadvantages of being rich: wealthy people have fewer people around them to tell them they’re wrong; wealth gained quickly can also be lost quickly; success breeds envy in others, who may ultimately try to undermine you; and (of course) most of the things that make us happy in life have little to do with money.
🔍📆 Lens of the week
Introducing new ways to see the world and new tools to add to your mental arsenal.
This week’s lens: Socratic humility.
Socratic humility refers to a particular mindset toward approaching particularly gnarly — especially seemingly hopeless — situations.
A famous quote attributed to Socrates paints this mindset in broad strokes: “I know that I know nothing.” When walking into the unknown, it seems prudent to humbly expect anything to transpire. We are better off leaning toward sensing (as opposed to judging) and being prepared to learn eagerly.
However, humble acceptance of not knowing is just one component of this mindset. The other component is the actual “Socratic” bit. Socrates (at least, according to Plato) was famously concerned with identifying inconsistencies in thinking and incoherence of concepts. Instead of making claims of judgment, the Socratic method focuses on questioning claims in the broader context of the claimant’s beliefs and looking for tension between them.
The famous Laches discussion of bravery is a great example of a Socratic process at work. Every attempt to define bravery confounds other beliefs. As the discussion progresses, it appears the participants know less about the term than when they started! And yet, all are enriched by the knowledge gained—even if it is a revelatory insight into one’s inconsistencies.
When we apply Socratic humility, we recognize we know less than we may think, but we are eager to learn from our observations and inconsistencies. We focus on puzzling over these details rather than trying to gain a holistic understanding of the situation. We don’t immediately try to solve the problem. We try to find weird (often smaller) parts of the problem that seem incoherent so we can learn more about them.
By focusing on the details, we progress in understanding the larger problem. It is still possible, and perhaps (even likely), we will never fully solve the problem. But when armed with Socratic humility, we are at least guaranteed to learn more about it — and about ourselves.
🔮📬 Postcard from the future
A ‘what if’ piece of speculative fiction about a possible future that could result from the systemic forces changing our world.
// The rate of improvement in Generative AI in the past year is incredible. How might the video game development industry change as a result of ever increasing AI capabilities?
"Welcome to the 2039 Game Developers Conference! Today kicks off 5 days of presentations, schmoozing, parties, and more than a few morning hangovers... Today is Monday, and it is the start of two days of focused tracks on various topics, from game prototyping to business development to the latest in platform development. Since you are here, you've decided the gAI track is for you.
"In the past decade, the average size of game dev teams drop precipitously. Contracting firms assisting with art, animation, testing, and programming have dropped in size as well. Grand Theft Auto 13 was built by a team of 25 over two years. A team of 21 built and shipped Hell Divers 6: Return to Super Earth.
"So welcome to two days of talking about how generative AI systems can help you accelerate the development of your game. Quick show of hands: who is building a game by themselves?
<vast majority of the room raises their hands>
"That's about what I thought. 92% of the games published last year on Steam list fewer than 5 team members. Which is part of the reason we insist on having these sessions in person. While building a game can be done alone, humans are social creatures. So, in addition to showing off tools and game development techniques, we will also talk about how to build relationships and support networks with other developers crucial for keeping your creative juices flowing. Yes, there are plenty of chatbots and companion-bots out there who will listen to you, encourage you, and even a few will push you to be better. Still, we believe relationships with other humans are a critical part of the creative process.
"Plus it means we can keep selling these expensive tickets for an in-person event! <a soft chuckle rolls through the crowd> "Alright, let's get started. Turn your ocular implants to shared workspace Red 5...
© 2024 The FLUX Collective. All rights reserved. Questions? Contact flux-collective@googlegroups.com.