Weekly Dose of Optimism #123
Brain Mapping, Walk Switch, Anduril x OpenAI, GenCast, Nuclear Meta, Isaacman
Today’s Weekly Dose is brought to you by … AlphaSense
AlphaSense recently acquired long-time Not Boring partner Tegus. Through their vast library of expert calls, the combined companies have more insight into how companies are actually using AI than anyone but the labs themselves. They’re where I turn when I’m trying to get the expert perspective on an industry or technology.
On December 17th, I’ll be talking to AlphaSense’s Director of Research, AI, Sarah Hoffman, about generative AI’s remarkable advances in 2024 and how enterprises will move AI tools out of proof-of-concept and into full integrations in 2025. Sarah’s experience and position give her a unique perspective on what to expect and how your organization can best take advantage of AI.
Join us to peek into the future, plan your 2025 strategy, and learn how to measure ROI. Register today to secure your free spot.
Hi friends 👋,
Happy Friday and welcome back to our 123rd Weekly Dose of Optimism. This was the most jam-packed week of optimistic stories we’ve had in a long time. There were probably 8 or 9 stories that could have reasonably made the cut here — stories that would have led the Dose five years ago sitting on the cutting room floor is a good sign for humanity.
Let’s get to it.
(1) A New Method to Map the Brain: Infect It With a Virus
Ashlee Vance for Bloomberg
E11’s approach, more or less, is to have a brain label itself. The lab does this by injecting a mouse brain with viruses that carry snippets of DNA to individual neurons and instruct the neurons to make special proteins. These proteins then become the biological equivalent of a barcode... The brain tissue, in essence, has built-in labels of its various structures and paths of wiring.
E11 Bio, a research organization backed by Eric Schmidt, is redefining brain mapping by injecting mouse brains with viruses that label neurons with unique protein barcodes. Basically the approach is like teaching mouse brains to barcode themselves so they can scan their neurons like groceries.
This brain mapping method drastically reduces time and cost, slashing the need for millions of microscope slices and tedious AI proofreading. E11 Bio injects viruses into the brain to create protein barcodes for neurons, enabling researchers to trace neural connections using colorful optical imaging (see image above!). This method skips the need for slicing the brain into thousands of pieces and avoids costly, time-consuming manual proofreading and electron microscopy.
If they’re successful, we could have detailed brain maps that unlock breakthrough neurological treatments, advance AI by mimicking brain function, and uncover the fundamental workings of the human mind. Long way to go — brain mapping has traditionally been costly, timely, and too static to provide real-time value — but if teams like E11 are successful, we might finally have a much deeper understanding of how these brains of ours work.
(2) Hypothalamic deep brain stimulation augments walking after spinal cord injury
Cho et al in Nature
A pilot clinical study showed that DBSLH immediately improved walking in two participants with incomplete SCI and, in conjunction with rehabilitation, mediated functional recovery that persisted when DBSLH was turned off. There were no serious adverse events related to DBSLH. These results highlight the potential of targeting specific brain regions to maximize the engagement of spinal cord-projecting neurons in the recovery of neurological functions after SCI.
OK, brace yourself for some acronyms. A new study revealed a breakthrough approach to restoring walking after spinal cord injuries (SCI) by targeting the brain's lateral hypothalamus (LH) with deep brain stimulation (DBS). In plainer language, the research uncovered a "walk switch" in the brain’s lateral hypothalamus, where targeted stimulation can restore walking after spinal cord injuries.
In animal trials, the research found that activating specific neurons in the LH dramatically improved walking by reorganizing the spinal cord’s remaining neural connections. Excitingly, limited human trials showed that DBS of the LH not only immediately enhanced walking but also led to long-term improvements when combined with rehab, even after the stimulation was turned off.
About 15M folks worldwide have some form of spinal chord injury and suffer varying degrees of paralysis as a result. Trials are early and the treatment won’t restore walking for everybody, but this is a big…step…toward using precise brain stimulation to repair movement after paralysis.
From Anduril
By bringing together OpenAI’s advanced models with Anduril’s high-performance defense systems and Lattice software platform, the partnership aims to improve the nation’s defense systems that protect U.S. and allied military personnel from attacks by unmanned drones and other aerial devices.
Technology and Defense. That’s what America does baby. 🇺🇸
Anduril and OpenAI announced they’re teaming up to advance America’s lead in both.
Anduril has established itself as the leading new-age defense prime. OpenAI is (and no shade to Anthropic, which we love here at Not Boring) the world’s leading AI lab. Together, they could produce some of the world’s most sophisticated defense technologies.
OpenAI is integrating its models into Anduril's counter-unmanned aircraft systems (CUAS), improving the drone system’s ability to detect, assess, and respond to aerial threats in real time. Notably, this is OpenAI’s first partnership with a defense contractor, reflecting a strategic shift to support national security initiatives. It’s no surprise OpenAI chose Anduril as their inaugural partner here — relative to its larger peers, Anduril is tech-first, much more immersed in Silicon Valley, and moves quickly.
We always get some pushback when we celebrate Anduril — something to the effect of, “How is the progress of this weapons company optimistic?” Our response is always the same: a more competent U.S. defense industry means a more powerful United States military which (if history is any indication) means a safer world which means more societal progress. And if we’re going to live in a world of drone-based defense systems (which at this point is inevitable) we might as well have them powered by the most sophisticated and competent AI technology.
(4) GenCast predicts weather and the risks of extreme conditions with state-of-the-art accuracy
From Ilan Price and Matthew Willson at Google DeepMind
GenCast provides better forecasts of both day-to-day weather and extreme events than the top operational system, the European Centre for Medium-Range Weather Forecasts’ (ECMWF) ENS, up to 15 days in advance. We’ll be releasing our model’s code, weights, and forecasts, to support the wider weather forecasting community.
Google DeepMind saw the praise we just lavished on OpenAI and they took that personally.
Google’s relationship with AI has been funny. On the one hand, Google Researchers discovered the Transformer architecture. On the other, the Google mothership has been slow and clunky in its rollout of generative AI into its flagship products. Remember the Black Nazis and Founding Fathers?
Here at Not Boring, though, we’re DeepMind defenders. While they might be losing the chat and image generation race, DeepMind continues to drop unique and valuable models — like AlphaGo, AlphaFold, GNoME, and now GenCast, which is more accurate than the most advanced operational weather forecasting system, up to 15 days in advance.
Results published in Nature show that its probabilistic weather model “has greater skill than ENS on 97.2% of 1,320 targets we [Google DeepMind] evaluated and better predicts extreme weather, tropical cyclone tracks and wind power production.” ENS is the ensemble forecast of the European Centre for Medium-Range Weather Forecasts, the current gold standard.
If there’s been a bigger loser in the AI Boom than Google, it’s been Europe. But the real winner is everyone in the world who gets access to better weather forecasts and to all the people who will be able to build with GenCast, since the company plans to release the model’s code, weights, and forecasts to the wider weather forecasting community.
Bad day to be extreme weather, good day to be a human, GREAT day to be Google DeepMind. The company also released Genie-2, “which can generate a diverse array of consistent worlds, playable for up to a minute.”
DON’T SLEEP ON THE DEEP.
(5) Accelerating the Next Wave of Nuclear to Power AI Innovation
From Meta
Today, Meta announced it will release a request for proposals (RFP) to identify nuclear energy developers to help us meet our AI innovation and sustainability objectives — targeting 1-4 gigawatts (GW) of new nuclear generation capacity in the U.S.
Zuck is going nuclear.
This week, Meta announced it was joining its hyperscaler peers in making its nuclear ambitions known. Google, Microsoft, and Amazon have each partnered with, invested in, or made commitments to companies that can provide nuclear energy capacity for their energy-intensive computing needs. For their part, Meta announced an RFP to identify nuclear energy developers to deliver 1-4 GW of new nuclear capacity by the early 2030s.
According to Meta, it’s at least 5 years away from having its data centers powered by nuclear, but this is yet another big development in nuclear energy’s comback story.. in the ELECTRONAISSANCE. In committing to nuclear energy, Meta is providing yet another point of financial and industry validation for nuclear, which could further accelerate innovation, deployment, and regulatory change. In a not so far off world, when all of the most important companies are powering their operations with nuclear energy, it’s hard to imagine that other companies and institutions won’t follow suit.
Bonus: Billionaire who performed the first private spacewalk is Trump’s pick to lead NASA
Marcia Dunn for The Associated Press
A tech billionaire who bought a series of spaceflights from Elon Musk’s SpaceX and conducted the first private spacewalk was nominated by President-elect Donald Trump on Wednesday to lead NASA.
He went to Jared.
If Elon Musk isn’t Iron Man, than Jared Isaacman most likely is. We’ve covered Isaacman a few times in the Weekly Dose, but if you’re not familiar with him, here’s the quick background: he made billions founding payment processor Shift4 Payments, parlayed that into founding aerospace defense contractor Draken International, and then famously commanded the first all-civilian spaceflight, Inspiration4…and most recently led the Polaris Dawn mission, becoming the first private citizen to perform a spacewalk. And now, in what you can only imagine is a boyhood dream come true, Isaacman has been nominated by President Trump to lead NASA.
Without knowing all of the politics around and nuances of NASA Administration, I feel confident in saying the Isaacman pick was a good one, for a couple of reasons. First, Isaacman is the perfect combination of successful businessman and passionate explorer — which hopefully means he’ll be able to not only run NASA more efficiently and effectively, but will also do so with the zeal of an astronaut missionary. Second, Isaacman has a very, very close relationship with Elon Musk. Musk has been, and likely will continue to be given his new role in the Trump Administration, NASA’s most important partner. If Isaacman can avoid obvious conflicts of interest traps and leverage this relationship with Musk properly, it should mean a much deeper relationship with the world’s most important private space company. And finally, Isaacman just seems like a legitimately great guy — the kind of guy that you look at and think “man, they don’t really make them like that any more.” Family-oriented, seemingly very kind (watch the Netflix doc and then tell me otherwise), financially successful, patriotic, and…a groundbreaking astronaut? He really is like Iron Man, but if Iron Man was from a middle-class family in Jersey and way less douchey.
After basically being stagnant for a couple of decades, NASA has started to get its mojo back with more ambitious and inspiring programs. Isaacman could be the man to cement the comeback and restore NASA’s glory.
Have a great weekend y’all.
Thanks to AlphaSense for sponsoring! We’ll be back in your inbox on Tuesday.
Thanks for reading,
Dan + Packy
we need 400000 (4000 terrawatts, US energy consumption) gigawatts of nuclear, we can make energy free if we want to. 1-4GW is a rounding error.
Anymore, the proper response to questions about celebrating Anduril is, "You're obviously not old enough to the answer. We'll try again when you're mature enough to be having these conversations."