Imagining the wild things that might happen when we mix and remix new technologies
Welcome to the 1,800 newly Not Boring people who have joined us since last Monday! If you haven’t subscribed, join 139,616 smart, curious folks by subscribing here:
🎧 The audio version is back! To listen to me read, head over to Spotify or Apple Podcasts.
Hi friends 👋,
Happy Monday! As sometimes happens, over the weekend, I got walloped with a case of Writer’s Block. I couldn’t figure out a topic. Normally, I’ll head into the weekend with at least a rough idea of what I want to write about, but this weekend: blankness. I stared at the computer screen. I turned to Twitter. I looked at the whiteboard on which I write potential future ideas, and got bored by all of them. I started writing a piece on The Bear and the Hard Work War, and then I stopped, 1,266 words in.
What’s in the news right now that would make a compelling piece? Twitter? I’ve written about it enough, and Matt Levine is doing Hall of Fame work there. The Fed? That’s been covered by much smarter people, no one knows what’s going to happen, and Arthur Hayes wrote this banger.
Sometimes, you just don’t have it. Realizing that this was one of those times was freeing. Instead of trying to do a deep dive or a bunch of analysis, I decided to let my imagination run wild (and keep it brief). It’s summer, things are slow… let’s have a little fun.
Let’s get to it.
Today’s Not Boring is brought to you by… Arrived
The bulk of the time we spend managing our money is boring. Paying bills - boring. Filing your taxes - boring. Budgeting - boring.
You know what isn’t boring? Building the foundation of your financial future using one of the most tried and true ways - investing in residential rental properties for the steady price appreciation and passive income the asset class offers. However, buying a rental property comes with high upfront costs and operational headaches like sourcing tenants and maintaining the property, which tend to push investors back to “easier” asset classes.
That’s the problem Arrived set out to solve in 2019.
With Arrived, you can buy shares of individual rental homes that were handpicked by their in-house experts based on their long-term appreciation and income potential. They also take care of the operations and property management so you can kick back, relax, and watch your wealth grow - definitely not boring.
Sign up is simple, minimums are flexible, and the portfolio diversification is *chef’s kiss*.
Get started with Arrived today.
One of the things that makes the future so hard to predict is that new technologies don’t operate in silos; they’re intertwining threads that weave together in unpredictable ways.
Uber is a canonical weave: for anyone to be able to dial up a ride from their pocket, GPS, smartphones, payments, and online ratings systems needed to combine in ways that their respective inventors couldn’t have predicted. This conversation between Patrick O’Shaughnessy and Kenneth Stanley is rich with examples of discrete innovations combining in ways that couldn’t be planned, like the invention of the vacuum tube unintentionally leading to the computer.
As the individual technological threads each get wilder and more powerful than they were in previous generations – replace “vacuum tubes” with “AI” as an input, for example – I suspect that the remixed outputs will be even harder to predict.
Even in sci-fi, where the job is to imagine fictional futures by extrapolating technological progress, most of the technology is fairly linear extrapolation: fusion-powered spaceships, AI robots, terraforming. I think we’ll have all of that at some point, but I also think we’ll have all sorts of crazy combinations that are harder to predict, and then we’ll have crazier combinations of those crazy combinations.
To show you what I mean, we’ll walk through some simple examples (with the caveat that these are just some of the ones I can dream up; the crowd will create things that make these look quaint).
Over the weekend, I came across this thread:
When I first saw OpenAI’s DALL•E 2 in the wild a few months ago, the first thing I thought was, “Woah.” The second thing I thought was, “It’s going to be really wild when this gets combined with other technology.” I started jotting down notes for an essay on combining technologies, then I abandoned it for a couple months until I saw Nicholas Ptacek’s tweet.
Nicholas’ experiment was one I’ve wanted to run: ask GPT-3 to write prompts for DALL•E 2. Ask the AI to ask the other AI to create something.
Using outputs from one OpenAI system as an input for another is an on-the-nose example of composability, combining one lego block with another, even manually, to create something new. It’s not hard to imagine that someone will automate this, so that I can type in a random idea, GPT-3 creates a bunch of DALL•E 2 prompts from it, and DALL•E 2 spins up images based on those prompts.
But that’s just the first and most obvious step here. Presumably, the models will continue to get better and be able to do more things. Ben Thompson wrote an excellent piece on this idea: DALL-E, the Metaverse, and Zero Marginal Content.
DALL•E 13 might be able to create short films from scripts generated by GPT-11.
I’m sure that there are all sorts of things that make this many orders of magnitude harder than creating static images (which itself is very hard)…
… and I’m equally sure that, at some point, it will be a reality. I’ll probably create and watch some of those short films myself; Dev certainly will; his kids will find them terribly old-fashioned.
Zoom forward a little further. Imagine combining DALL•E 22, GPT-19, and Snap Spectacles 15. Then remember that Snap recently acquired brain-computer interface startup NextMind.
Dev’s kids will be thinking things like “space battle jungle” (they won’t know what a “video game” is) and will pop into a full-blown, freshly-created AR/VR game of their imagination.
Dev will be lucky to have something to keep his kids entertained. The flight to his family’s Lunar vacation will be shorter than the current three days it takes to get to the Moon, but parents will always do whatever it takes to keep their kids from screaming on flights.
Pull on just a few of the many open tech threads, and you land in a world in which anyone can create their own entertainment instead of passively consuming it, at the speed of thought. It will make TikTok look slow and archaic, for better and for worse.
But that’s just one pair of threads. The world won’t, I hope, look like the DALL•E version of that scene in Wall-E.
As Matt Ball pointed out on Derek Thompson’s Plain English, the time Dev’s kids spend in the Metaverse will hopefully replace the 3-5 hours per day of “lean-back” TV consumption with something more engaging and creative.
But I want my grandkids to spend more time outside, in nature, too. A different, and eventually intersecting, track is one proposed by Not Boring’s own Elliot Hershberg yesterday in Viriditas.
Laying out his personal mission – “I want life to flourish in the universe. I view biotechnology as the most logical means towards this end.” – Elliot makes the case for more biological growth:
Growing trees, the living systems that pull CO2 out of the atmosphere while turning sunlight and water into beautiful three dimensional structures that carpet large portions of our planet. Growing food using the awesome power of microbes in a way that could sustainably feed the entire planet. Growing our medicines—advanced nucleic acid, cell-based, and microbial therapies that enable us to live healthier and longer lives. Growing materials to build advanced cities with abundant housing.
Instead of escaping into virtual worlds, at some point, we’ll be able to grow new ones. Imagine Dev’s kids dreaming up the ideal treehouse … and then growing it in their backyard.
Now, it could take a century or more to get there. “We wanted flying cars. Instead we got 140 characters.” Obviously, a lot needs to happen between now and then.
We need to figure out how to code cells to manufacture human-scale biological structures, then how to make those capabilities so widespread that anyone can use them (without making the ability to grow biological weapons equally widespread), then how to use AI/ML to create grow-ready plans for biological buildings (in a way that adapts to each specific backyard’s terrain and environment), then how to hook that up to some brain-machine interface that lets kids think “treehouse” and grow a treehouse, all while ensuring that no one gets hurt in the growing process and they don’t break zoning laws. I skipped a lot of steps, too.
But the threads are there.
Elliot wrote about Ginkgo’s vision of making biology programmable. CRISPR is a programmable biological system for editing DNA. These technologies are already in the wild. As biology becomes programmable, it stands to reason that at some point, it will interact with the new suite of AI programming tools.
GitHub Copilot, which uses the OpenAI Codex, can assist developers in writing “code and entire functions in real-time.” In April, Replit created an AI/ML team, which is building tools to help people learn how to code, and then assist by finding errors in code as they write it. CEO Amjad Masad tweeted, “The crazy big idea here is that we think this could be a start of a general-purpose learning environment, a human-machine symbiosis that transcends mere coding.” AI-assisted cell programming to grow treehouses is something that sounds impossibly wild but could very well become a reality as threads weave together.
Even if it takes 500 years to achieve full-blown Artificial General Intelligence, even if it’s never possible, narrow AI is already becoming an “interactive intellectual amplifier.” As more things can be expressed via code, and AI gets better at writing code, the amplifier’s reach will expand. The limiting factor will be our imagination.
New technologies will create new challenges. There’s the challenge of who owns the technology: Google owns leading AI lab DeepMind, Microsoft invested $1 billion in OpenAI and owns GitHub. When GitHub announced that it would charge $10/month for Copilot, it stirred backlash among developers, who were upset that the company trained the model on their open source code and then turned around and charged them to use it.
While this won’t be a satisfactory use case to web3 skeptics, I think that this is an interesting place to weave in the web3 thread. As transaction costs come down to the sub-penny level, I’d imagine that people will be able to own their data, opt-in to contributing it to models, and then receive rewards when those models are used. Already, GitHub gives free Copilot access to maintainers of popular open source projects, and it’s not hard to imagine a future in which those developers get paid more than $10/month for their work.
More broadly, as more of our data is used to train AI models of all sorts, owning, permissioning, and getting paid for the use of that data will become more important. And as our grandkids create more content, I expect they’ll be backed by some future version of NFTs that prove creatorship and even pay out the people whose data went into feeding the models that helped generate the content. As they spend more time in realistic virtual worlds, tools that give digital items physical properties will become more useful.
All of this – AI, space travel, biotech, web3, immersive virtual worlds, and all of the other threads that we haven’t touched on – are going to require a lot of energy. Getting to the point where any of this is possible, let alone possible in a world that’s not a dystopia with radical inequality, is going to require cheap, abundant energy.
Eli Dourado and Austin Vernon’s Energy Superabundance: How Cheap, Abundant Energy Will Shape Our Future pulls at more threads, like Suborbital Point-to-Point Travel, Hyperloop, Vertical Farming, Desalination, and New Cities, that cheap, abundant energy make possible. Each of them intertwines with other threads, like growing buildings for those new cities, or living and working anywhere thanks to Hyperloop and Vertical-Takeoff-and-Landing Taxis.
And, of course, each of them will come with their own problems, red tape to cut through, and unintended consequences. For some, the risks will outweigh the rewards. The future won’t be perfect; it will be wild.
We could play this game for hours – mix one cutting-edge or futuristic technology with another and see what pops out – but I told you I’d keep it short, and I’m more curious to hear what you come up with: drop your examples in the comments or on Twitter.
Ultimately, all of this is why I try to keep an open mind – and even a sense of wonder – towards any new technology solving hard problems. Each individual thread has the potential to do something predictably powerful on its own – ranging from fun to lifesaving – but their combinations will be indistinguishable from magic.
Thanks to Dan for editing!
Have a great week, and see you back here on Thursday!
Thanks for reading,
Really liked this post, you should this type of post more often
I'll believe the promise those intertwining threads hold only when new technologies like AI and ML are combined with other existing ones like gene editing and 3-D printing are used to solve hard problems such as health equity by class, gender, and race among Americans. Until then, these new technologies will be nothing more than platforms for only the rich and powerful to use and benefit from. I enjoyed reading your article though and wish you well in your writing and business endeavors.