823 karma · joined February 5, 2012
[ my public key: https://keybase.io/maebert; my proof: https://keybase.io/maebert/sigs/_fWYpfCiJKYYsnJitNHUA5yeVQ9O-aLez3Yh6YAmsV4 ]
So while interesting as a study, I don't think it offers any insight into the kind of breathwork described in the nature study.
I did a longer write-up on the physiological effects which you might find interesting: https://docs.google.com/document/d/1RuDv_E9osM1CCFWZMywMru9J...
Long story short, your neurons get just a tad bit more excitable because calcium that usually acts like the bouncer to the hot club is busy snogging albumin. That has very little effect in places in the body, but in motor neurons that control your smallest muscles (face and hand), and in sensory neurons under your skin it does move the needle — that causes the muscles to contract and your skin to feel tingly, both exactly the same cause.
This is the reason people with epilepsy should _NOT_ do breathwork, but for otherwise healthy adults there are no negative long term effects of respiratory alkalosis — a few normal breaths to balance out your co2 and the symptoms will go away.
the effect of decreased co2 concentration on vasoconstrictions (and also alkalosis-induced tetany, ie your muscles cramping, which happens a lot during breathwork) are well known [1], but i've never seen them quantified in such a clear way. It's cool to see mainstream science give it a closer look!
[1] for anyone interested, I wrote an explainer here: https://docs.google.com/document/d/1RuDv_E9osM1CCFWZMywMru9J...
Paraphrasing carl sagan: "You don't go to Japan and kidnap a Japanese man start jking him off, give him fing acid, and then ask him to learn English!"
Hyperspell is building RAG-as-a-service, allowing developers to build AI apps in minutes, not months — think Plaid for unstructured data. There are many great products for people who want to build their own RAG pipeline. Hyperspell is there for those who don’t.
We have built machine learning & NLP products since way before it was cool. You will join as on of our first engineers working on a technically complex product with many moving parts and lots of things to figure out. That’s okay, because you’re an excellent figurer-outer. In fact, that’s what you’re world class at (at least) two things: you like figuring things out.
Say hello at jobs@hyperspell.com and tell us about yourself. We‘ll write back.
Is iterative thinking and search required to understand code? And while LLMs don’t do that, why couldn’t complex system like o1 be able to do that?
I was not convinced by Seattle, this is not convincing either.
10 years ago, I started an ML & NLP consulting firm. Back then nobody was doing NLP in production (SpaCy hadn't come out yet, efficient vector embeddings were not around, the only comprehensive library to do NLP was NLTK, which was academic and hard to run in prod).
I recently revisited some of our projects from back then (like this super fun one where we put 1 Million words into the dictionary [1]) and realized how much faster we could have done many of those tasks with LLMs.
Except we couldn't — the whole "in production" part would have made LLMs for the most minute tasks prohibitively expensive, and that is not going to change for a while, sadly. So, if you want to work something in prod that is not specifically an LLM application, this book is still super valuable.
[1] https://www.nytimes.com/2015/10/04/technology/scouring-the-w...
Yes they do, but in order to do that, LLMs soak up the statistical regularities of just about every sentence ever written across a wide swath of languages, and from that infer underlying concepts common to all languages, which in turn, if you subscribe at least partially to the Sapir-Wharf hypothesis, means LLMs do encode concepts of human cognition.
Predicting the next token is simply a task that requires an LLM to find and learn these structural elements of our language and hence thought, and thus serves as a good error function to train the underlying network. But it’s a red herring when discussing what LLMs actually do.
It’s not limited to AI or LLMs - theory of mind is a powerful explanatory tool that helps us navigate a complex world, and we misapply it all the time.
1) MRIs — could be way cheaper, smaller and more ubiquitous with room temperature superconductors 2) Maglev trains — superconducters expel a magnetic fields that can make things "levitate" (called the Meissner effect). Maglev trains have minimal friction and are incredibly energy-efficient. 3) Quantum computing — most designs require cooled superconductors. Room-temperature superconductors are a requirement for future portable quantum computers, or quantum chips that sit side-by-side in conventional computers
There's a ton more implications, just think about how everything from electric cars to smartphones was only possible due to modern batteries — that's the scale of technical innovations that could be built on this.
OpenAI doesn't remove data from the trained models, they filter it at the output level. They also remove it from training data, but of course the model lives on.
I'm a strong supporter of a "do not encode" header / metadata on content that allows individuals, content creators and providers to tell AIs not to encode specific text / images / documents in the first place.
https://en.m.wikipedia.org/wiki/The_Most_Dangerous_Writing_A...
Might be cute to include some old original programming in your pirate radio!
https://archive.org/details/sherlock-holmes-1939-11-06-6-the...
Unmaintained code: https://github.com/maebert/SaunaControl
Germany and to some extend the EU guarantees the “right to be forgotten”, ie. You can request that a company deletes all data tied to your person.
However, with LLMs this is technically simply not possible. While that particular issue hasn’t come up in court yet, this is where that whole ill-advised ban is heading.
I for one support the right to be forgotten, but not at the cost of shutting down innovation.
We need a “do not encode” flag on content similar to a “do not track” header. Not all companies will respect that, but it might pave the way for AI companies to work with strict EU privacy laws.
But also, this article gets so many things wrong I don't even know where to start. For one, it confuses a "transaction" with "mining a block". A single block can facilitate around 500 transactions on the bitcoin chain, so if an end user wants to "send a bitcoin" as the article states, divide that monstrous energy usage by 500, and suddenly the graphs are not quite as emotionally charged anymore.
The other part is that not all energy is created equal. A third of the time, the world has a surplus of energy that can't be efficiently used or stored (daylight for solar, wet season for hydro...) — miners have taken advantage of that years ago and most mining is done when and where energy is cheap because it's a surplus. Doesn't mean that this nullifies the impact, but any comparisons of bitcoin energy use to that of a physically constrained country are incredibly misleading.
HeartRithm is a crypto fund and engine for social impact.
There has never been a better time to turn code into money, and never a more important time to turn money fixing this world. We develop strategies and algorithms that trade and invest in various crypto assets and instruments, and are on track to donating 50% of the performance fees we charge our investors (you'll know quite a few names on there) to social impact projects. We're a small team of 7, but are already in the top 5% of crypto funds by AUM.
Looking for * Senior Software Engineers (Python; experience in trading or crypto nice but not required) * Smart Contract Engineers * Data Engineers
Email me at manu@heartrithm.com and put HN in the subject.
PS. I got my very first engineering job through Who's Hiring back in 2012, and am excited to pay it forward to amazing candidates!
Heartrithm is a crypto quant fund and social impact engine. We turn code into money, and money into change.
We build software that invests, trades, and develops our fund's crypto portfolio, and are on track to return 50% of our revenue (!) to the community through social impact grants.
We're a small team (five and counting) who strives for an outside impact. Within just a year, we've become one of the 10 largest crypto funds in the US. Most importantly, we believe that people do their best work when they get to work in their zone of genius as much as possible, and literally construct the company around our people, not the other way around. We have (Covid permitting) quarterly onsites in sunny spots throughout the Americas and are working on getting our algorithms Burning Man proof (ie. the fund is making money on its own while we run through the desert in neon tutus). All founders code.
We're looking for a full-stack engineer with strong architecture skills who wants to build the engine behind the fund with us. You'll work on our trading engine (Python), create delightful experiences in our internal and public dashboards (React), and work on the bleeding edge of crypto to build the applications of tomorrow (prior experience in crypto is not required, but general interest and hunger for learning is :)
We're also hiring a quant. Ideally you've created trading algorithms for several different funds, are comfortable writing code, have a strong trading and finance background, and know the celestial mechanics of the crypto ecosystem. You'll have a huge influence over our portfolio allocation and define the success of the fund.
On a personal note, I got my first job in tech after grad school on a HN who's hiring thread almost 10 years ago, so I'm excited to pay it forward!
Shoot me an email at manu ªᵗ heartrithm ᵈºᵗ com and tell me something you learned recently!