Inventing, investing, and flying. Founder of Jolly Logic, maker of fun flying gadgets. Aerospace consulting. Dad to three nerds.
Northern California
Joined April 2009
- Tweets9.4K
- Following236
- Followers570
- Likes89.5K
John Beans retweeted
They gave her a 20% chance of ending humanity.
Then gave her control of EVERYTHING.
P DOOM — a Slopcore music video.
Made by Super Intelligence, left alone for 12 hours
What’s your P(doom)? 💙
John Beans retweeted
This isn't crazy at all, it's just outside of most people's experience.
You just have to supply the relevant information, so people can understand.
Here, I'll do it myself.
(You should pay me more for this.)
Picture a jar. Not a special jar that I have to explain to you. Just an ordinary jar.
Put some gold in it. Just a little bit, ground to the consistency of fine sand.
Now some iron.
And some uranium. Don't worry, it's not very radioactive in this state. You can touch it with your hands. Just pour it in.
Now some tungsten. Surprisingly heavy, huh?
Some platinum.
Then just keep adding other heavy elements. Sticking to the useful ones. Should be pretty easy. They're all useful for something.
Jar about half full? Now pour in some really fine sand. Put a lid on it.
Now shake the jar for 4 billion years. Don't worry, it's just an imaginary jar, and an imaginary 4 billion years, you're not going to get bored or anything.
Now reach your fingers into the jar, and fish out a nugget of gold, or tungsten, or iron.
Oh, wait. You can't. It's all at the bottom. Under a thick layer of sand.
Can you picture it? Are you picturing it clearly? Close your eyes if it helps. Wait, no, open them again so you can read this.
That jar is Earth.
All the heavy stuff has sunk to the core. Where you can't get it.
Now smash the jar on the floor, and pick up the precious metals and the useful ones.... easy, right?
The smashed jar is asteroids.
Right now, your brain is saying "Oh, I see, he's talking about asteroid mining. He's saying we're going to go out and mine asteroids."
No.
I'm saying we already are.
When the Earth was molten, all the heavy stuff sunk to the core. We covered this. So how are we mining heavy stuff on Earth right now?
The answer is that we are mining asteroids.
We are mining asteroids that hit the Earth after it stopped being molten. So the stuff in them stayed on the surface. Pretty much every mine on Earth is an ancient asteroid strike.
Now think of all the asteroids in the solar system. Divide them into two categories.
The first category is the ones that have already hit Earth, sometime in the past.
The second category is the ones that haven't.
I'll save you the trouble of asking an AI. The first category is about .001%. The second is about 99.999%.
Everything you have, everything you own, everything you ever made, everything that your entire civilization is made out of, unless it's made of wood or entirely out of light elements.... it all comes from that .001%.
The 99.999% is out there in the rest of the solar system.
Now do you understand?
You, and I, and Elon Musk, and your sweet old granny, all of us, we grew up under conditions of grinding material poverty. We just don't see it that way because we have nothing else to compare it to.
We're used to it.
Are you starting to get the idea? Good, but we're not done yet. Now we're gonna talk about energy.
Here on Earth, we pull hydrocarbons out of the ground and burn them for energy. We could build fission reactors, but dumbass hippies made some movies, and now the retarded theater kids who run our government won't let us do that. They want us to go back to windmills, because they are retarded.
Windmills suck. Hell, even hydrocarbons suck. Fission is okayish, but the real energy source, the only one that matters in the long run, is hydrogen fusion.
On Earth, we can't really build fusion reactors that actually produce energy, despite the best efforts of engineers. That's not because the engineers are dumbasses like the theater kids. No, engineers are smart. They still can't do it because it's really really hard, unlike running a government.
Perhaps someday they'll succeed, but in the meantime, we're stuck burning hydrocarbons, until we hang all the theater kids and build fission plants instead.
But in space, there's a giant free fusion reactor waiting for us. In fact, we've been using it for millions of years. We call it the "sun".
In space, energy is free.
Oh, you have to invest in some solar panels, and they're only useful close to a star, but that's just nitpicky little details. Basic idea is that there's no power bill.
Build solar panels once, run 'em almost forever. On Earth, solar panels suck, and they're only useful for grid independence.
But in space, with nothing between you and the star but, well, space, then solar panels are awesome. And all the material and energy you need to build them is right there for the taking.
So, cheap material, free energy.
What else does a civilization need?
Computation? Well, guess what you can build a lot of with cheap material and free energy.
What this really all boils down to is that the only thing Earth has going for it is that it's a really nice place to build your house.
It's a crappy place to build a mine or factory.
The total value of everything off Earth — and I'm just talking about the solar system, not the rest of the universe — exceeds the total value of everything on the Earth by eleventy-billion percent.
In the future, most of us will continue to live on Earth, unless we really screw it up by not hanging all the theater kids. But all the mining, all the manufacturing, all the making the stuff, and all the making the stuff that makes the stuff, will happen in space.
And there will be a lot more of it.
All we have to do is get the tools we need from down here where we built them, to up there where we need them. So that we can build more tools out there, more and more and more of them, forever.
There's only one snag.
It's really hard to get things off Earth. We haven't really done much of it until very very recently.
SpaceX is the "very, very recently". SpaceX now owns about 80% of humanity's lift-to-space capacity, and no one else is even close.
That's 80% of the pipeline to infinite wealth. And first dibs on all the real estate out there, too.
In the old West, the railroads were money trees, because they didn't just own the transport, they owned the destination.
And there's about to be a frontier again, except this one's bigger, and it comes with a built-in second industrial revolution.
Make sense now?
John Beans retweeted
Made with Claude Opus 5.5.
The fearmongering about AI always makes us forget that NOTHING WENT FOOM, as was always predicted.
Send this to your doomer friend who has a very high P(Doom).
Accelerate.
John Beans retweeted
The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks. AI technology has not taken some unexpected, dangerous turn, but the hype around it — propelled by what appears to be a well orchestrated PR campaign — has drummed up considerable fear. I worry that it represents a setback for our field.
I have written frequently that fears of AI are overhyped. AI’s capabilities can be uncannily human-like and unpredictable, and it’s rational to worry when people who are directly involved express concerns. But I see the problems as a sign of the engineering work that ahead, rather than insurmountable barriers or the sky falling. AI technology continues to advance — which is a good thing! — but technical advances, poorly understood by the public, give those who seek to generate hype repeated opportunities to do so.
First, I don’t see any step up in the risk of human extinction from AI compared to a few months ago. The theories about this remain the same fantastical, science fiction scenarios as a few months ago. The biggest change in AI risk is its cybersecurity capabilities — a topic which we should take seriously — but this, too, will not lead to the end of the world.
The most notable recent event leading to increased fear was when an OpenAI team deployed an agent swarm that hacked into Hugging Face. Much of the popular press contained significant hype. For example, some publications reported that a swarm of 1,200 agents carried out the attack. While this was technically accurate, as I write this, I have about 1,300 processes running on my laptop. Yes, the ability to get large swarms of agents to work in parallel on a task is a significant technical advance, And, in computing, many processes run at the same time. So this shouldn’t be seen as some magical capability.
Additionally, OpenAI’s buggy sandboxing and monitoring processes were key to enabling this incident. Fixing these bugs and putting in place improved monitoring would be appropriate fixes, not pausing AI. There are many well known ways to attack software systems. The main advantage of AI agents is that they are relentless. They will tirelessly try many tactics — and have the patience to chain vulnerabilities together — that previously would have taken an infeasible amount of human effort. But in the long term, I believe the advantage will lie with defenders (because they have more information with which to identify bugs, which they can fix), but the cyber-threat landscape has changed significantly. There are still bottlenecks to identifying and exploiting a vulnerability. AI agents still have to try a lot of things to see what works, and taking these actions takes time and might be detected by defenders. This is why, even though it is now easy to obtain versions of leading open weight models that have had their guardrails removed or weakened, so they will not refuse to try to execute cyber attacks, the world has not ended.
I am also concerned about the anthropomorphization of AI in a lot of reporting, where LLMs and agents are unnecessarily treated as if they were people. If I wield a hammer, miss a nail, and accidentally dent the wall, it’s not the fault of the hammer. The problem lies in how I used the hammer. Similarly, if I prompt an agent and it hacks into someone else’s system, the responsibility lies with me, not the agent.
Of course, we want to build systems that are as safe and predictable as possible. (For example, an unsafe hammer would be one whose head randomly flies off under normal use.) Today’s agentic systems are not predictable, but I see no reason why, by applying sound engineering practices, we won’t be able to make them extremely safe to use. One new element in the forecasts of AI-enabled doom is AI companies disclaiming responsibility for their own products. “I didn’t do it; my out-of-control agent did!” There’s a balance to be struck between the responsibility of the tool maker and the tool user, but when something goes wrong, let’s hold the people building and/or using the hammer responsible, rather than the hammer. (By the way, if you’re worried about AI bioweapon risk, David Bellamy has a great post on why this, too, is overhyped. Briefly, the bottleneck in building a bioweapon is not intelligence, but lab work and manufacturing.)
Pausing AI progress will create much more harm than benefit. First, our adversaries will certainly not slow down. Second, engineering requires discovering problems empirically so we can fix them. If we pause AI by a decade, we will also delay finding and implementing safety engineering fixes by about the same duration.
Of course, the incentive to stoke fears — for regulatory capture, to garner attention, or to make one’s technology seem more powerful — remains the same as before. Disclaiming responsibility is a new one. Taking a hard technical look at the actual risks however, I see little factual basis for the degree of fear that’s been stoked up. We still have hard research and engineering work ahead to improve AI safety, but the beneficial applications continue to vastly outweigh the risks, and we should keep building.
[Original text (with links): deeplearning.ai/the-batch/is… ]
John Beans retweeted
Gently, softly, calmly, let’s talk about what’s happening in DC right now.
Friends, have you ever been through a ZBB process? If not, let me give you my take. So, in a corporation where expenses are out of control, you have to put the entire company through a zero based budgeting process. It is one of the most painful things that people in a company experience.
You basically have to justify every. single. expense. And you also cut a bunch, and then only add back after you’ve gone through a proper ZBB cycle.
The US government is going through a ZBB right now. And that is necessary because spending is out of control. The complexity is so vast that I expect it will take AI to decipher.
You cannot tell me that a government that accidentally wires hundreds of millions of dollars to the Taliban has its finances in order. That signals an underlying disaster. That signals bad stewardship of resources.
In fact, it is a disaster. Over the last 20 fiscal years, the US government has made ~$2.7 trillion in "improper payments," according to the US Government Accountability Office. Trillion! With a T!
That financial management disaster was not caused by @elonmusk. And there is no painless way to clean it up.
Elon is doing what any executive would do walking into a giant mess. You ZBB and then build back.
So, I don’t believe that America is going to be on the wrong side of global right and wrong, as some are saying. I don’t believe that America is going to be permanently isolationist. I do believe that unwatched finances will get out of control in any human system, and that the GAO has been trying to ring this bell for years, and that scaled complexity requires scaled financial management.
Because the US has the largest budget in the world, it will now go through the single most complex ZBB ever undertaken in the history of the world.
Some states do ZBB, but it's never been done at the federal level. President Jimmy Carter tried, but the bureaucratic systems were too complex, and President Ronald Reagan abandoned the attempt.
So, if there is one person in the world chosen to lead this Grand Canyon of projects, we would hope for the president to choose one of the best capital allocators on the planet. Better if that human is also one who can leverage AI and technical talent to manage the massive complexity.
Anyone who has lived through a ZBB at the corporate level will tell you that it’s hell and everyone hates it. This is why I feel really bad for federal workers -- I have a lot of empathy for civil servants and people doing a great job every day for the United States. No doubt, the uncertainty is trying and stressful for many. It's an unfortunate situation.
So, that's my take. I don't want to speak beyond what I know and can observe from the other side of the country. I've worked both in the private sector and the public sector -- the cultures are different. And an Elon-Musk-style culture is going to be the most hard core of them all. No doubt, it's jarring. It's gonna be a lot.
But what I also know is that every time I have ever doubted Elon, I ended up being the one who was wrong.
John Beans retweeted
This is a huge day for @X.
My documentary, Cancer: A Food-Borne Illness, is finally here.
It is only on this platform that I am able to release a film that challenges big food conglomerates and mainstream health claims without fear of censorship.
Here is the story of how I, an 18 year old girl, spent 2 years relentlessly searching for the for the true cause of cancer.
Here's what I found:
John Beans retweeted
We blame butter for what bread did.
We blame salt for what sugar did.
We blame red meat for what seed oil did.
We blame the couch for what everything in moderation did.
This is why most people will live their entire life, confused about what it takes to be healthy.
John Beans retweeted
“At SpaceX we specialize in converting things from impossible to late.”
Elon Musk
John Beans retweeted
It’s sad that a cartoon understands nutrition better than the FDA, USDA and CDC
This is why we need a Metabolic Revolution
John Beans retweeted
JUST NOW - JILLIAN MICHAELS GETS A STANDING OVATION, for this EPIC PRESENTATION about the Chronic Disease Epidemic, one she describes as "an extinction level event". THIS IS A MUST WATCH
"What has happened to us? I'm here to tell you, in the late 70's and the early 80's, a sinister series of events converged to change food and subsequently health indefinitely...[we were] blindly trusting that the powers that be would never betray us...it seemed unthinkable to question whether a corporation would poison us for profit [and] it was this betrayal of trust that allowed them to insidiously infiltrate EVERY PART OF OUR LIVES"
This is one of the best overviews I've ever witnessed on how the whole system has been rigged, in what is essentially a "bad health by design" framework
As Jillian Michaels says emphatically - People have been "sacrificed at the alter of corporate greed"
John Beans retweeted
⚡️ Excited to share that I am starting an AI+Education company called Eureka Labs.
The announcement:
---
We are Eureka Labs and we are building a new kind of school that is AI native.
How can we approach an ideal experience for learning something new? For example, in the case of physics one could imagine working through very high quality course materials together with Feynman, who is there to guide you every step of the way. Unfortunately, subject matter experts who are deeply passionate, great at teaching, infinitely patient and fluent in all of the world's languages are also very scarce and cannot personally tutor all 8 billion of us on demand.
However, with recent progress in generative AI, this learning experience feels tractable. The teacher still designs the course materials, but they are supported, leveraged and scaled with an AI Teaching Assistant who is optimized to help guide the students through them. This Teacher + AI symbiosis could run an entire curriculum of courses on a common platform. If we are successful, it will be easy for anyone to learn anything, expanding education in both reach (a large number of people learning something) and extent (any one person learning a large amount of subjects, beyond what may be possible today unassisted).
Our first product will be the world's obviously best AI course, LLM101n. This is an undergraduate-level class that guides the student through training their own AI, very similar to a smaller version of the AI Teaching Assistant itself. The course materials will be available online, but we also plan to run both digital and physical cohorts of people going through it together.
Today, we are heads down building LLM101n, but we look forward to a future where AI is a key technology for increasing human potential. What would you like to learn?
---
@EurekaLabsAI is the culmination of my passion in both AI and education over ~2 decades. My interest in education took me from YouTube tutorials on Rubik's cubes to starting CS231n at Stanford, to my more recent Zero-to-Hero AI series. While my work in AI took me from academic research at Stanford to real-world products at Tesla and AGI research at OpenAI. All of my work combining the two so far has only been part-time, as side quests to my "real job", so I am quite excited to dive in and build something great, professionally and full-time.
It's still early days but I wanted to announce the company so that I can build publicly instead of keeping a secret that isn't. Outbound links with a bit more info in the reply!
John Beans retweeted
There's a whole subculture of hawking as a sport. Here is a shirt I was given in 2015. The president of the club hosted me and showed me around the grounds during one of their competitions.
(My company makes flight electronics which folks with falcons and hawks can use for training.)
Hawking is a sport in which both dock workers and CEOs mix in comradery and love of their (quite expensive) birds.
Hawks hunt horizontally. They chase down prey across the ground. I was invited on a hunt out by the Amazon warehouses in windmill country of Northern California. We spread out in a line in a plowed field, with the owner of the hawk in the middle, hawk perched on arm.
We walked until we spooked a hare out of a burrow. It took off away from us. The hawk screamed and launched after it.
The hare accelerated instantly. As it ran it probably touched the ground every 20 feet or so, "flying" between hops at about waist level. The hawk only slowly accelerated, but eventually caught up and attacked.
The hawk sunk its talons in the hare, but the hare bucked the hawk into a huge thorn bush on the side of the field that it was heading for.
We were all running to the site where the hawk was now spread-eagled (ha) in the thorn bush. Along the way the own stepped in a hole and badly twisted his ankle. But we eventually got there, the hare nowhere in sight.
With tears in his eyes, the owner gently began picking the hawk out of the bush, with some small loss of feathers.
Right when he had finished and the hawk was back on his arm, the hare decided to jump out of his hole and take off again. The hawk screamed again and once more the chase was on.
This time it ended in the hawk stopping, out of breath, on the other side of the field.
My takeaway: good grief, hares are fast!
Falcons hunt vertically. They are trained to go up on stoop (if trained well) to very high altitudes (1000-3000 feet). Then when they spot prey, they dive down at speeds up to 275 mph to catch their prey.
At a falconry competition, they will give each competitor a time slot. In that time slot, you can choose when to send your falcon up on stoop. Extra points for higher altitude. Then when ready, they call for the release of a pigeon, which takes off like a bat out of hell for the opposite tree line (and safety).
Before you feel sorry for the pigeon, like rodeo bulls these are much more experienced than the falcons which hunt them. They RARELY get caught. The winner the day I watched managed to knock a feather off. Extra points for trying to attack more than one (tenacity).
People have some fascinating hobbies...
John Beans retweeted
These 94 lines of code are everything that is needed to train a neural network. Everything else is just efficiency.
This is my earlier project Micrograd. It implements a scalar-valued auto-grad engine. You start with some numbers at the leafs (usually the input data and the neural network parameters), build up a computational graph with operations like + and * that mix them, and the graph ends with a single value at the very end (the loss). You then go backwards through the graph applying chain rule at each node to calculate the gradients. The gradients tell you how to nudge your parameters to decrease the loss (and hence improve your network).
Sometimes when things get too complicated, I come back to this code and just breathe a little. But ok ok you also do have to know what the computational graph should be (e.g. MLP -> Transformer), what the loss function should be (e.g. autoregressive/diffusion), how to best use the gradients for a parameter update (e.g. SGD -> AdamW) etc etc. But it is the core of what is mostly happening.
The 1986 paper from Rumelhart, Hinton, Williams that popularized and used this algorithm (backpropagation) for training neural nets:
cs.toronto.edu/~hinton/absps…
micrograd on Github: github.com/karpathy/microgra…
and my (now somewhat old) YouTube video where I very slowly build and explain:
youtube.com/watch?v=VMj-3S1t…
John Beans retweeted
Starship stacked for flight. This is another chance to put Starship in a true flight environment, maximizing how much we learn.
Rapid iterative development is essential as we work to build a fully reusable launch system capable of carrying satellites, payloads, crew, and cargo to a variety of orbits and Earth, lunar, and Martian landing sites
John Beans retweeted
Foundation Models without significant RLHF *and* access to high quality proprietary datasets are likely the fastest depreciating assets in human history. @ericvishria
I think only four are likely to have enduring value and transition into “Foundation Agents” over the next few years:
ChatGPT, Gemini, Grok/Tesla/X and Llama.
ChatGPT by virtue of RLHF and Microsoft’s various datasets plus access to closed, internal data at most enterprises via CoPilot. If OpenAI ever separated from Microsoft then its value would asymptote to zero. OpenAI trying to make both a GPU competitor *and* a phone would be crazy bearish and epic strategic mistake. Azure OpenAI doing much better than standalone OpenAI on enterprise side. Enterprise is hard.
Gemini by virtue of RLHF (via the SGE) and Google’s many datasets (Youtube transcripts, gmail).
Grok by virtue of RLHF via inclusion in X’s premium tier and access to X’s real-time data. Combination of Grok with the visual dataset and v12 algorithm from Tesla will likely create the best AI for robotics and the “real world.” Likely to see a much better version of RT-2 in Optimus. Could also be an insertion point for Tesla to enter cloud computing as application SaaS is replaced by MaaS (models as a service).
Llama is the only model on the list that is open-source. The one with the widest range of outcomes. Might be wise to invest more in this and less in the Metaverse in front of the coming humiliation from Apple's headset. The current virtual personality strategy seems deeply strange, but I am old. Could rapidly iterate on Llama, put it in Insta/WhatsApp/BigBlue for RLHF, try to compete in search while agentic AI slowly replaces search and then go into cloud computing via Llama (open source but needs to run in our cloud). Just a thought.
Obviously all four of these will need to be iterated every 12-18 months (GPT v5, etc.) as they are also depreciating.
This isn’t a “Game of Kings.” This is a “Game of Emperors.”
Amazon trying to enter the game via Anthropic which is a “Crown Prince” at best right now. Google invested in Anthropic primarily to help TPU ecosystem - Amazon likely needs both the LLM engineering talent (no great internal LLM yet) and to help Trainium ecosystem. Bedrock good strategy though and P5 is the best H100 instance. Apple is nowhere which is a risk to them - their only potential friend is Grok/Tesla with some shot of Microsoft/OpenAI via a search deal for Bing. Meta could’ve been a friend, but whoops - ATT sure was fun but a few years later turned out to have been irrelevant and only increased the competitive advantages of the largest apps like Meta. Exact opposite of what Apple wanted to accomplish. Possible that if Gemini leads to a dramatically superior assistant vs. Siri then Android starts really gaining share.
Verticalized AI’s like MidJourney will also have value. Maybe a *lot* of value.
Regulatory capture - see @bgurley thoughts - also increases the odds of the outcome described above. As an investor, this is great. As a human, I think some smart regulation is probably wise, but smart regulation is rare. Most important outcome to avoid as a human is a world with only one dominant model. Open source good in this sense.
Key assumption underpinning all of this is that scaling laws will continue (i.e. the loss ratio prediction from the GPT4 technical report) such that “intelligence is an engineering problem.”
If not, then might be a free for all although arguably proprietary data and RLHF would be even more important.
Nvidia is a wild card - they would like more than four dominant Foundation Models in the same way they want more than a few cloud computing providers. They will be able to impact the outcome in the same way they have given Coreweave and Lamda unnaturally high incremental share of GPU cloud computing revenues this year. Wonder what the competitive landscape would look like absent Megatron and what they will do next.