lamename

I appreciate the sentiment, but I've found this resource [0] much more direct and comprehensive. It explains all of the nuance regarding dB and related terminology from audio engineering to perception (voltage, power, intensity, volume, loudness, etc.)

The format is a bit circular; just enjoy getting lost in it for half an hour.

https://sengpielaudio.com/TableOfSoundPressureLevels.htm

Also on HN today "I got fooled by AI-for-science hype—here's what it taught me" https://news.ycombinator.com/item?id=44037941

Thank you for saying this, I agree with your point exactly.

However, instead of using that known human bias to justify pervasive LLM use, which will scale and make everything worse, we either improve LLMs, improve humans, or some combo.

Your point is a good one, but the conclusion often taken from it is a shortcut selfish one biased toward just throwing up our hands and saying "haha humans suck too am I right?", instead of substantial discussion or effort toward actually improving the situation.

Nice writeup. F1, balanced accuracy, etc. In truth it depends on your problem and what a practical "best" solution is, especially in imbalanced scenarios, but Matthews Correlation Coefficient (MCC) is probably the best comprehensive and balanced blind go-to metric, because it guarantees that more portions of the confusion matrix are good [0,1].

I made a quick interactive, graphical exploration to demonstrate this in python [2].

[0]: https://biodatamining.biomedcentral.com/articles/10.1186/s13...

[1]: https://biodatamining.biomedcentral.com/articles/10.1186/s13...

[2]: https://www.glidergrid.xyz/post-archive/understanding-the-ro...

> this is aimed primarily at engineers who don't have ML expertise: someone who understands the business context, knows how to build data processing pipelines and web services, but might not know how to build the models.

I respect software engineers a lot, however ANYONE who "doesn't know how to build models" also doesn't know what data leakage is, how to evaluate a model more deeply than simple metrics/loss, and can easily trick themselves into building a "great" model that ends up falling on its face in prod. So apologies if I'm highly skeptical of the admittedly very very cool thing you have built. I'd love to hear your thoughts.

It's not just LLMs. Deep learning in general forms these multi-d latent spaces

If a Google engineer can get tricked by this, of course random people can. We're all human, including the flaws.

Education to the uneducated (or those who would prefer we remain uneducated in the face of power) can easily cast any education as "biased" against their purposes. Most people see through that for what it is, but an increasing population of Americans don't.

The saddest part is watching friends and colleagues who i previously considered healthily skeptical and rigorous thinkers now "blown away" by promotional videos, drinking the kool aid. Yes it is amazing. No I don't believe everything it comes up with. Why would we?

Probably because dB SPL doesn't match A-weighted human perceptual audiogram, so they're being specific? (I get that you could just translate it to dB SPL but still.)

Your story does a great job at highlighting the small but profound every-day experiences of working in a lab with living beings as test subjects. It's hard to put into words the strength of wonder you can acquire for life working as a scientist or animal handler -- you have immense depth of understanding of the microscopic scale biology while simultaneously seeing it at work in real time in the whole animal (or society of animals) on the macro scale. It's the biology equivalent to the "pale blue dot" sentiment.

The only similar deep, profound awe at life I've had outside of the lab was when my son was born. This might be the most common way people achieve this state of being. In all honesty this was one of the best parts of lab work. For me it happened every day; you're reminded of the insane complexity and the high degree of frailty of life. The terribly large power difference between yourself and a small animal in your hand. The deep similarities between humans and other animals, and at the same time, the worlds of difference. For me these experiences in the lab day after day put many other things in life on a lower rung, for better or for worse (like sustaining grad student pay and living conditions perhaps). But I wouldn't trade having that experience for anything.

Your story hints at this beautifully, and I hope others get the chance to feel that feeling.

The point of this, particularly for children (e.g. as a parent), is to build emotional bonds and stability, not to get them to do what you want. That can be a nice side effect in the moment, and is indeed more likely over the long term with this approach. But the chief goal is emotional safety, validation, etc.

> "At the moment that technology is limited because you need a dermoscopic lens which the public wouldn't necessarily have access to, but I'm sure with time the technology will advance and we will have effective apps that patients can access from the comfort of their own home."

+1 for comparative neuroanatomy indeed

Oddly, what tends to matter for competence in X domain in relation to the brain isn't the absolute size of the brain (or brain region) but size relative to 1. The other structures in the brain 2. Total brain size relative to body size

This is very cool and lends even more evidence to what has been more or less clear in the avian neuro community. This genetic/dev evidence in a similar brain region supports previous work by Karten (mentioned in the article) and others on bird auditory "cortex".

The mantra is "nuclear" or "regional" organization in birds rather than layers in mammals, but the anatomy [0, Fig 4] and electrophysiology [1] abstract this out in similar patterns.

[0]: https://pubmed.ncbi.nlm.nih.gov/20616034/

[1]: https://www.pnas.org/doi/full/10.1073/pnas.1408545112

I watched a discussion the other day on this "NNs don't overfit point". I realize yes certain aspects are surprising, and in many cases with the right size and diversity in a dataset scaling laws prevail, but my experience with real datasets training from scratch (not fine tuning pretrained models), and impression has always been that NNs definitely can overfit if you don't have large quantities of data. My gut assumption is that original theories were not demonstrated to be true in certain circumstances (i.e. certain dataset characteristics), but that's never mentioned in shorthand these days when data sets size is often assumed to be huge.

(Before anyone laughs this off, this is still an actual problem in the real world for non-FAANG companies who have niche problems or cannot use open-but-non-commercial datasets. Not everything can be solved with foundational/frontier models.)

Please point me to these papers because I'm still learning.

Neat, now I'm curious. Can you explain why?

I'd argue that your point is only true if the amount of money the government gives out per month per person vastly increases.

Related is the fact that other jobs existing, even IF desirable, doesn't mean they can be learned over night. Many jobs worth having take years of training. Switching careers overnight is not realistic for nearly anyone who gets laid off due to replacement.

It doesn't have to be. But often the executives or investors who stand to profit the most from innovation also have strong public facing influence over the narrative. Employees cost a ton, so it's self serving both to promote the product to like minded people, and to hype the product itself.

Last commit 1 yr. They welcome your contributions in CONTRIBUTING.md

You might find this helpful for prioritizing which knobs to turn first https://github.com/google-research/tuning_playbook

Nice. What laws are you referring to?

I don't see `tsai` in there either

This is flame war now, but there are > 1 person who can run a company successfully. No Eich != bad company. Blame current leadership not lack of specific leadership.

Whats the best casual (non Tor) browser these days? Brave? Opera?

This is the dream. But the author even mentions the reality

>Having tried this both ways, I can say that having a steady salary (and health insurance!) allows me to do much better work than when I was worried about my next source of funding.

So yes, there are, in theory, no gates of knowledge or funding. But overall this reads as a lucky person passionately encouraging those considering the same path, but from a very lucky position of survivorship bias.

Do you feel that your technical skills, people skills, or luck have helped you to avoid any ageist treatment you've encountered over the years? Especially in scenarios where "deciders" are younger than you.

This should be pinned as the top comment on 80% of these kind of blog posts as a reality check. Don't get me wrong, I'm (relatively) young, ambitious, and love growth and learning. So I eat these posts up.

But as I age and grow, I notice the challenge of the balancing act more and more.