HNHacker News
TopNewBestAskShowJobs

dado3212

1,150 karma · joined August 17, 2016

Hi, I'm Alex Beals, I'm reachable at <firstname>c<lastname> at gmail dot com.

https://github.com/dado3212

https://blog.alexbeals.com

submissionscomments
dado3212··on The impossible predicament of the death newts
I also handled these all the time as a kid in northern Oregon, literally hundreds in a day because they would hang out in a series of lakes nearby. Is there a non-poisonous rough-skinned newt variant? I find it impossible to believe that as a young kid I wouldn’t have licked my fingers after years of doing this.
dado3212··on How a single line of code could brick your iPhone
Neat, $17,500 is pretty good, I’m so used to these blog posts being for peanuts, or where companies fix the vulnerability but don’t pay out at all. Apple’s gotten better about this since 2019.
dado3212··on Colossal squid filmed in ocean for the first time
A juvenile, just under a foot. Direct footage on YouTube here[0] if you don’t want to click through ads.

[0] https://youtu.be/lzPoG9H8Hlo?si=udDxsRPeLOcR_smk

dado3212··on Spain to impose fines for not labelling AI-generated content
Wrong title?
dado3212··on Part two of Grant Sanderson's video with Terry Tao on the cosmic distance ladder
Any specific parts?
dado3212··on I helped fix sleep-wake hangs on Linux with AMD GPUs
> So I did the natural thing: I saved and extracted the amdgpu.ko kernel module, decompiled it in Ghidra, and mapped the location of the crash in dm_resume to the corresponding lines in the kernel source.

This is always my favorite part of debugging.

dado3212··on Ending our third party fact-checking program and moving to Community Notes model
> So, we’re going to continue to focus these systems on tackling illegal and high-severity violations, like terrorism, child sexual exploitation, drugs, fraud and scams.

I don't think this is exhaustive, and I think SSI (suicide/self-injury) + ED/etc. stuff is considered high-severity.

dado3212··on Kelly Can't Fail
Very cool writeup, would’ve benefited from some LaTeX formatting.
dado3212··on Lawsuit blames Character.AI in death of 14-year-old boy
https://www.nytimes.com/2024/10/23/technology/characterai-la...

as well

dado3212··on Why is it so hard for the U.S. to build quality transit?
Why is this unique to America? Why doesn’t this apply to the UK?
dado3212··on Why is it so hard for the U.S. to build quality transit?
This article does an okay job posing the question, and then stops. Is there an actual answer or work on an answer that people would point to?
dado3212··on Llama 3.1
> We use synthetic data generation to produce the vast majority of our SFT examples, iterating multiple times to produce higher and higher quality synthetic data across all capabilities. Additionally, we invest in multiple data processing techniques to filter this synthetic data to the highest quality. This enables us to scale the amount of fine-tuning data across capabilities. [0]

Have other major models explicitly communicated that they're trained on synthetic data?

[0]. https://ai.meta.com/blog/meta-llama-3-1/

dado3212··on From GPT-4 to AGI: Counting the OOMs
Are there any good OCR packages that are state of the art for general-purpose transcription? (i.e. give it a business card and get it to format it for you, give it a comic and have it transcribe it, give it nutritional info and have it table it)? When I looked recently I pretty much just got GPT-4o as the best API.
dado3212··on Show HN: Simulating 20M Particles in JavaScript
The videos look awesome but the "try it out here" codesandbox links don't work for me on MacOS Chrome desktop. I get 'Uncaught ReferenceError: SharedArrayBuffer is not defined' and some CORS errors: 'ERR_BLOCKED_BY_RESPONSE.NotSameOriginAfterDefaultedToSameOriginByCoep'.
dado3212··on Singapore doubles down on lab-grown meat as Silicon Valley backs off
CAR-T treatments are expensive because they’re individualized. The process has to be repeated per-individual. That’s completely different from lab-meat.
dado3212··on House votes to ban TikTok in the U.S. if it's not sold
Admittedly TikTok is China-owned, and China already bans multiple U.S. based social media companies.
dado3212··on French artists in 1899 envisioned what life would look like in the year 2000
This might be a better source, with all of the images.

https://publicdomainreview.org/collection/a-19th-century-vis...

dado3212··on Money bubble
Archive while this is down - https://web.archive.org/web/20240229191137/https://www.tbray...
dado3212··on Behind a Secretive Global Network of Non-Consensual Deepfake Pornography
Cool analysis, though it doesn’t make for a very linear read. This is the kind of issue where laws really need to be updated, it’s crazy that non-consensual pornography is legislated against across the US but that deepfakes (for the most part) aren’t included.

Feels like a high importance area to move fast in with current real world harm, yet we move slower than our premature attempts to litigate AI compute guardrails.

dado3212··on Do we think of Git commits as diffs, snapshots, and/or histories?
That’s interesting! I think of them as diffs, though I know they’re snapshots. More familiar with Mercurial though, which may influence my thinking.
dado3212··on California seal pups were turning up headless. Experts confirmed the culprit
> Otherwise please use the original title, unless it is misleading or linkbait

I think this probably qualifies, a better title would be something like “Experts confirmed the culprit: coastal coyotes”

dado3212··on YouTube doesn't want to take down scam ads
YouTube does not have the scale to manually review every time someone flags an ad. HN community believes that “oh, but MY report is high signal” which may be true, but user reports as a whole are not high signal. So instead it’s going through some ML classifier first.

An automatic answer saying “this doesn’t go against policy” does not mean that it escalated to YT staff who carefully considered the ad and said “actually, this is okay because we really want the revenue.” It means that whatever ML classifier took a look at it thought it was okay (which is probably bugged, given their typical speed at taking down fake livestreams in this vein).

dado3212··on Developer account removed by Apple
What makes you say that? The article is light on details about what the apps actually do, and why they might be violating.
dado3212··on Emu Video and Emu Edit, our latest generative AI research milestones
I don’t think either of these (or the base Emu model) are open source.
dado3212··on [dead]
This feels like it’s a ChatGPT synthesis of a better article. Samsung’s proposed headset is NOT called the Vision Pro, nor is it really inspiring a reboot (perhaps the AI missed that Apple’s inspired a Samsung reboot?).

Submitter has also only submitted articles from this website, somewhat dubious.

dado3212··on How one patient found errors in the algorithm making transplant decisions
Or an alternative framing: using data can help to remove human biases and increase overall lifespans (which the studies have shown). How is it sociopathic to work on software that demonstrably saves lives?
dado3212··on The false positive rate of AI detectors and its effect on freelance writers
Alternative reading: company determines that paid contractor is producing content that they can’t distinguish from a borderline free LLM, and decides to use it instead. Isn’t it likely that AI (not the detector) is the real threat to freelance writers’ livelihoods?
dado3212··on Meta to charge for ad-free versions of Facebook and Instagram in Europe
Out of curiosity, did you opt out of ad tracking?
dado3212··on OpenAI in Talks for Deal That Would Value Company at $80B
> peak interest in LLMs is already over

Where are you seeing this? It feels like it's just beginning. Also really not sure that the software development and writing applications have not materialized, I know multiple people who actively use it for both.

dado3212··on Meta's new AI dating coach will kink shame
Bing does the same thing, I think just optimizing for latency. Admittedly it probably shaves off 10-15s in the usual response, I’d probably make the same decision.
← PreviousPage 2 of 7Next →