aalam

Abliteration. The word and its descriptions read like pure sci-fi

https://huggingface.co/blog/mlabonne/abliteration

"Here's the part that nobody talks about"

"Two gotchas before you click buy"

I really think there could be a score for entropy in playfulness that should differentiate LLM output

As always, I _wish_ I had a use case for an iPad. Seem like such powerful machines hindered by where they live in the serious-computing space. The iPadOS being much more restrictive doesn't help either.

I wish they could repurpose macOS to touch screens... Oh well.

I'll never understand the hype of buying a Mac Mini for this though. Sounds like the latest matcha-craze for tech bros

Naive take but I feel the only way to combat this problem is with AI. Any manual queuing, or separate threads could be overrun all the same. The only difference being there is a dam of sorts before we let some of the projects out.

Any manual efforts to combat AI won't scale as models get better and better. Show HN is a place to show-case cool projects, I don't see why a 100% AI generated project can't be shown. AI has allowed many to rob themselves of their retirement projects and the uptick reflects that. My hunch is once we settle into the new shift, we can perhaps tweak the parameters around decay of Show HN posts.

Or we could allow upstanding members to signal-boost Show HN posts. Something along the lines of "Hey guys I _really_ want you to see this post so here is it (again)".

I was lagging behind in AI until a colleague shared your guide about Context Management in LLMs. Thank you for that!

This sounds great, do you think something like this could be ported over to other countries' laws?

Where I'm from, citizens _need_ more awareness of their rights today and in the future.

TIL about shading, and am surprised how less I've seen this term in grading tutorials. While different, I feel like shading is something that should be learnt before grading.

PS You might have pasted two different answer drafts above. Paras 1,4 and 2,5 deliver similar information

They don't do that anymore, at least not for me. I tried contacting two different support agents and both mentioned that the functionality has been removed for them by the higher-ups.

It's not always social status, especially for people who volunteer but don't tell anyone. It can often be ideology and principles, from people who want to pay a favour forward (e.g. if they received help from the past).

I do agree there are other intangibles, though (sense of community with fellow volunteers, sense of freedom by doing something exceptional, work experience). These intangibles are good things (mutual benefit).

Some people also benefit without expectation of reward. There could be an argument of self-interest (e.g. to feel good or alleviate some past wrong), but in practice, it makes sense to just thank these people for their actions without worrying about their motivations (which they might not even definitively know).

That's fair game to record the conversation too as an interviewee (most reporters wouldn't object). Most reporters by default also record the conversation directly (better for fact-checking and makes the job easier). Misremembering isn't typically a problem in practice; there are legitimate consequences for journalists, at least at respected publications, who misreport quotes (issued correction notice in the article, and the publication can get bad press from other outlets).

The main downside with providing all the on the record statements in recorded audio, is that some statements by an interviewee can be false. If a scientist misstates a figure and corrects it later, listeners can latch on to the incorrect figure. Keeping it to print allows the reporting of the correct figure, or the printing of the quote with a note that it's a misstatement. You could also try to contextualize such statements in the transcript itself, but it then becomes more resource-intensive than the current system (publications are obligated to verify the entire transcript, versus selected parts).

I have personal experience with transcription software (e.g. Otter.ai), and it's still not good enough. Even if you get 95% accuracy, which isn't a given if the audio is recorded outside with background noise, that's still 5 in 100 words for correction, more if there are technical terms.

So you need someone to review it and clean it for accuracy. It's not trivially easy at all to provide all the source materials for every interview because it's time consuming to clean and verify. For especially sensitive investigative reporting that isn't breaking news, and where there's more time to report, news outlets (e.g. CBC News) already publish recordings and full email transcripts.

I would also argue that it's the norm for journalists from respectable outlets (usually the ones with paid subscriptions) to put quotes in proper context. It's an ethical principle of the field. When they don't, the current system works when the interviewee, readership, and press watchdogs call out the unethical reporting.

I actually experienced a similar issue between the jack and the cable with exposed wires. It didn't affect audio quality for me, so it was just cosmetic; I wrapped electrical tape around the exposure as a fix.

This shouldn't happen considering the price, but it wasn't a dealbreaker for my personal preferences (as the fix wasn't too noticeable, and I enjoy the noise isolation and sound quality).

It's common because it would be too resource-intensive to provide transcripts for every quote.

Articles from high-circulation newspapers often have five or more interviewees per article. Quality newspapers are averse to publishing quotes without fact-checking them if possible (e.g. if a politician makes a false claim, you don't want to publish it without indicating it's false). To maintain the same principle, newspapers would need to fact-check the entire transcript, versus just the quoted part, for accuracy.

But if you want only a partial transcript, then you're essentially at the current state of affiars where you only quote the part you need, and paraphrase the rest. Journalistic ethical guidelines already require quotes to be in context for fairness. Reputable publications have an incentive to publish quotes in context (the interviewees, journalist watchdogs, and many readers would criticize that publication if they don't). So I don't see anything wrong with reading quotes and assuming they are published in good faith.

To mitigate ethical lapses, you can also read the same coverage from different sources (e.g. Wall Street Journal and New York Times) to get broader context about particularly important articles, and also subscribe to newsletters on reporting (e.g. the American Press Insitute newsletter, the Columbia Journalism Review).

My perception of Bose, as a former regular audio forum reader, is that they are great for sport earphones (though this was before the Airpods were released, so this might be outdated). They also last a long time and are excellent for noise cancellation, which I learned when looking for ways to block noise during work (ultimately I stuck to Etymotic's ER4XR to block out noise via isolation, versus noise cancellation).

For audio quality, however, there are definitely better alternatives at the same/lower price (but not necessarily as good for exercise or noise cancellation). I also heard that the AirPods Pro are comparable to Bose's QC35s for noise cancellation, but not as good (though the AirPods are far more portable), and their batteries don't last as long in the long-term.

I switched to Comply foam, then back to silicon for good. The dealbreaker for me was cleanliness. The downside you mentioned of squishing the foam was the biggest reason; I always aimed to sanitize my hands before rolling the foam (pre-Covid, to prevent colds), so it was an extra hassle to take them out often.

Also, it's easy to clean silicon tips with dish soap and water (alternatively isopropyl), but I found that Comply tips break down really easily with washing. They're also expensive to replace frequently, and I found that cheaper foam alternatives didn't measure up.

If regular cleaning isn't a concern (isn't strictly necessary, it's just a personal preference), Comply tips work, though I would only use them for IEMs that don't fit well for silicon.

If you haven't already tried it, a filter replacement with the tool and filters that come with the device could do the trick [0] (fixed a problem I experienced myself). Another common fix is to switch out the cable, though I've only experienced failure due to old filters (3 years usage).

[0] https://www.youtube.com/watch?v=qbDgTa8Htzo

I had a similar problem, and the solution was to change out the filter in the non-functioning earpiece [0] with the tool and replacement filters that came with the device.

I also bought a replacement cable, which didn't work, then contacted support for a replacement. Their suggested fix of changing the filter worked. I suspect the reason may have been moisture; I washed the silicon tip with soap and water, then put it back on the earpiece and filter without waiting for it to fully dry.

[0] https://www.youtube.com/watch?v=qbDgTa8Htzo

For a good historical example, see Linus Pauling’s notebooks (with page numbers and dates): http://scarc.library.oregonstate.edu/coll/pauling/rnb/

The blog by biologist Dr. Colin Purrington also has advice for maintaining a notebook (discussing why, backups, and what to note down): https://colinpurrington.com/tips/lab-notebooks/

Your advice on introspection is helpful, especially for people who find themselves repeating the same mistakes.

However, I would like to add a caveat on introspection. For some people, an absorption with introspection and self-understanding can be a distraction from making concrete progress in life goals. Sometimes, it helps to shut off or turn away from the stream of thought to plug away at important tasks. An excess of introspection can create an illusion of progress, though it’s still a crucial habit to practice.

A major technique mentioned by the author is omitted from the article’s headline: co-working. She writes about benefitting greatly from logging on to a muted Zoom room with hundreds of other people to work at a set time. It’s interesting how this is effective for some people; for another anecdote, blogger Alexey Guzey wrote about a similar system [0] with a smaller group.

Psychiatrist Dr. David Burns writes at length about more techniques on addressing procrastination (called “do-nothingism”). He has a chapter on the subject in his book “Feeling Good,” and he wrote a similar passage online in an article for his podcast [1]. The methods he writes about that work best for me have been breaking down tasks into sequence of simple steps. This helps greatly for straightforward tasks, though it is less effective for difficult tasks that involve uncertainty and learning. These harder tasks require persistence and preferably time.

Separately, a personal technique I use is to really sell myself on the value of completing a task. The phrase, “discipline matters more than getting motivated” is thrown around a lot on the web, and though it has merit, I find motivation crucial to sustaining projects in the long-term (and also figuring out when it’s worthwhile to exit from a method or approach).

[0] https://guzey.com/co-working/

[1] https://feelinggood.com/2017/09/18/053-ask-david-i-dont-feel...

There can be positive effects of matching a story, but it depends on the journalists and the publication. Let's say The Wall Street Journal publishes a major investigative scoop, but you only pay for a Washington Post subscription. The Washington Post then matches the story to give readers access to the information, maybe with additional information or a unique angle, and sends a push notification to readers with the app. The WaPo article cites, and likely links, the WSJ if readers want to read the original report, and its readers become aware of the news.

That is the high-quality, professional version of matching. The low-quality, unprofessional version is echoing a story for clicks, but you're less likely to run into this if you are a newspaper subscriber who reads reports from publication apps (or alternatively, a reader who curates high-quality sources of reporting without paywalls).

There's a term for that in journalism: "matching" a story. This traditionally involves legwork, such as contacting the same sources quoted in the first article to verify that facts are true.

This legwork is usually done by traditionally respected publications (i.e. not just copying and pasting). Other outlets and blogs with different goals do blatantly copy and paste, which produces articles less valuable than those traditionally matched.

Matching—versus copying and pasting—is important because it avoids the spread of false reports. From "The Elements of Journalism" by Bill Kovach and Tom Rosenstiel: "Originality is a bulwark of better journalism, deeper understanding, and more accurate reporting. […] In the era before curation and aggregation, the tradition of “matching” stories was rooted in the same idea. Rather than publishing another news outlet’s scoop, journalists tended to require one of their reporters to call a source to confirm it first. This tradition of matching was a way for news organizations to avoid having to credit their rivals, which in this earlier era was considered an embarrassing admission of being scooped. Yet the tradition of matching had another more important and salutary effect. Stories that couldn’t be independently confirmed would not be repeated."

A norm in the journalism industry is to credit the first publication who reported an investigative scoop, naming the outlet (e.g. this New York Times article [0] cites information reported by less-read Punchbowl News).

[0] https://www.nytimes.com/live/2021/01/21/us/joe-biden

“Defective” could have been more precise, and would match the reporting (though I kept the title close to the source material’s to meet HN guidelines, truncating the start to meet the character limit).

The initial rollout is limited to the US, with no concrete plans reported yet on expansion.

“The scheme will initially roll out only in the US. […] Apple’s neuralMatch algorithm will continuously scan photos that are stored on a US user’s iPhone and have also been uploaded to its iCloud back-up system.”

Researchers interviewed for the article would agree with your analysis. “Security researchers [note: appears to be the named security professors quoted later in the article], while supportive of efforts to combat child abuse, are concerned that Apple risks enabling governments around the world to seek access to their citizens’ personal data, potentially far beyond its original intent.”

Article link for ease of access: https://www.ft.com/content/14440f81-d405-452f-97e2-a81458f54...