HNHacker News
TopNewBestAskShowJobs

harperlee

5,439 karma · joined February 12, 2013

To contact me write to atienzar256 at the mail of Google dot com.
submissionscomments
harperlee··on Jonathan Blow has spent the past decade designing 1,400 puzzles
I was just speaking generally, I also don't have a side there. But for clarity, what I meant is that templates to the tune of "[subset of people with X characteristic] are more / less prone to [Y characteristic]" can construct blatantly false sentences, and also sentences that, irrespective of whether they are true (or that they are falsifiable at all, as you add), have a heavy political penalty.

I also don't think that's bad - you can say blatantly racist things with that template, and I'm ok with those things not allowed to be said in lots of contexts.

harperlee··on Jonathan Blow has spent the past decade designing 1,400 puzzles
Whether one is allowed to pose some particular questions is a political topic though!
harperlee··on If AI replaces workers, should it also pay taxes?
100% agree! I find quite concerning that this point is not immediate in any conversation about AI or robots impacting the number of jobs, and the subsequent conversation about innovating new taxes. AI and robots are capital as any other automation on a factory, and capital gains should be taxed appropriately. This is not a new thing completely separate from the untouchable status quo wrt existing taxes. If it tickles your political kneejerk, explore that, but playing tax sci-fi is distracting and thus dangerous.
harperlee··on Europe is scaling back GDPR and relaxing AI laws
It is a government who says that…
harperlee··on Europe is scaling back GDPR and relaxing AI laws
That is not surprising. Regulations are a way to ensure things that are not easily reached by market forces. Doesn’t mean that we should not care for that.
harperlee··on Europe is scaling back GDPR and relaxing AI laws
GDPR is not about the cookie banner, it has massive implications around the whole lifecycle of data. For example you need to be able to gather all data of a particular client for them to access, and they have the right for all their data to be erased.
harperlee··on When 1+1+1 Equals 1
XOR is a simple operation that shows that behavior.
harperlee··on Why do we need dithering?
It's related to nostalgia. If you have lived at the time where dithering was used, you will have an emotional response to it. For example, if you played games as a child with dithering, then playing The Return of the Obra Dinn, which was mentioned in other thread, will take you back to a happy place.

And even if you did not live at that time, exposure to that distinct visual style will also start having meaning to you. Like how an exposed brick interior wall has a distinct aesthetic, and carries connotations of an industrial space.

harperlee··on Attention lapses due to sleep deprivation due to flushing fluid from brain
That only works if the mean stay in the hospital (or at least the critical care period) is several hours but also way below 24h…
harperlee··on Why I'm teaching kids to hack computers
Game for kids, where you dedicate a third of the screen to a locked hint list and a very prominent "Buy Hint Tokens" button? Hard pass.

https://www.hacktivate.app/img/framed-ipad-3.png

The game industry needs to move away from milking vulnerable people with pay-to-win schemes.

harperlee··on US probes Waymo robotaxis over school bus safety
Not sure about a movie but that reminded me of the "Driver" short story in the "Valuable Humans In Transit and Other Stories" tome by QNTM (https://qntm.org/vhitaos).

I'd recommend to buy the book, but here's an early draft of that particular story:

https://qntm.org/frame

harperlee··on Karpathy on DeepSeek-OCR paper: Are pixels better inputs to LLMs than text?
But assuming that pixel input gets us to an AI capable of reading, they would presumably also be able to detect HWLLO as semantically close to HELLO (similarly to H3LL0, or badly handwritten text - although there would be some graphical structure in these latter examples to help). At the end of the day we are capable of identifying that... Might require some more training effort but the result would be more general.
harperlee··on Deloitte to refund the Australian government after using AI in $440k report
Less cynically: there is only one partner; the team, if it works, manages to leverage the partner to ~5x - 10x.

And no, you don’t pay partner rates. The client generally knows that to do commercial pressure you need to get info about team size, put it in the contract, and then negotiate on blended rate so it is close to what the market pays for the junior.

harperlee··on Don't avoid workplace politics
I'd add that in my experience, when you are close to the action, the cynic "golfing nepotism" take is usually missing a point of view that is far more rational; just far from the developer/architect that is dismissing the decision. Perhaps not technically optimal, or fair, or even legal - but even so, more akin to "I know this person delivered in the past, and the alternative is also good on paper" or "I really need to save my ass" (nobody got fired by choosing IBM) or "business-wise, this technical recommendation I don't really care for". Perhaps I'm optimistic but in general I don't really think (or want to believe) that people are quick to wage their careers on an acquaintance that is clearly selling something as part of their job, unless the stakes are really not so high from their point of view. Then again, open a newspaper :)
harperlee··on iRobot Founder: Don't Believe the AI and Robotics Hype
I read here genetics, because you're avoiding a sensitive topic.

I actually think that it is culture. Often people attribute to melanin what is very obviously an issue with values and cosmovision, as part of the culture that someone breathes since childbirth. Which is solvable with education but only on generational timescales, unfortunately. Similar to Planck's principle, "Science progresses one funeral at a time".

harperlee··on Delete FROM users WHERE location = 'Iran';
451 coming from here by the way: https://en.wikipedia.org/wiki/Fahrenheit_451
harperlee··on How can I influence others without manipulating them?
The way I read manipulation is that someone is handling (manus: latin for 'hands') someone merely as a tool, bending and pushing and whatnot, just as a mere means to an end, without big regard for the tool, in an egotistical way. Objectifying the other, to some extent.
harperlee··on The health benefits of sunlight may outweigh the risk of skin cancer
Soap also kills parasites. And antibiotics!
harperlee··on The health benefits of sunlight may outweigh the risk of skin cancer
Might it be that when outside you tend to look farther than when inside? So distance vision gets used more and body adapts. Similar to how kids that spend time outside are less shortsighted.
harperlee··on Many hard LeetCode problems are easy constraint problems
I believe NoahZuniga is perfectly aware of the intent and denouncing an abuse of (unneeded) notation.
harperlee··on Take something you don’t like and try to like it
What about the converse: showing your significant other e.g. the best film ever, only to discover with horror how bad your taste was a number of years ago?
harperlee··on Steve Ballmer Interview
Owning stock is the polar opposite of having money in a vault - it is giving money to the economy. Specifically to a part of the economy that is a money-generating machine: a company. Put in circulation precisely to be invested on the economy.
harperlee··on The Math Is Haunted
Perhaps this thread is a good place to ask, could anyone contribute their own opinion about the relative future of lean vs. idris/coq/agda? I want to dedicate some time to this from a knowledge representation point of view, but I'm unsure about which of them will have less risk of ending up as so many esoteric languages... I sank a lot of time on clojure core.logic for a related project and got burnt with the low interest / small community issue already, so I've been hesitant to start with any of them for some time.
harperlee··on A simple way to improve spaced repetition scheduling
It does - it is easy to see when answering a new card, where the default times for the different options are <1m, <6m, <10m, 25d.

On the other hand, people do have a circadian rhythm and schedule, and learning is dependent on sleep - so it makes sense that learning plans of a given spaced repetition system do consider a day step (even if there is more resolution underlying the higher level planning).

harperlee··on Solar power has begun to transform the world’s energy system
Diesel, iron or aluminum, from your parent post, are difficult to explode… (personally, no clue about magnesium); and the point of the latter two is that you can “store” energy by upstreaming its consumption when power is available, you don’t necessarily need to produce an actual reversible energy store.
harperlee··on X-Clacks-Overhead
GDPR goes way beyond cookies so that dichotomy is nonsensical.
harperlee··on Denmark to tackle deepfakes by giving people copyright to their own features
I stand corrected, what is illegal is to operate them continuously, not to just have them as I simplified above.

"If, for example, you use a continuous recording of the road in which other vehicles' license plates are visible to defend yourself against a traffic ticket, you could be violating data protection, a serious offense that could be punishable by a fine of up to 300,000 euros."

https://www.race.es/camara-para-coche

harperlee··on Denmark to tackle deepfakes by giving people copyright to their own features
As an example, in Spain it is illegal to have dashcams, and its content cannot be used in a trial - but you can share the content with insurances and policeman, and recording is generally not prosecuted. It is nonetheless an opening if an officer is searching for a way to fine you...
harperlee··on LLMs are cheap
Nitpick: Au contraire
harperlee··on Type-constrained code generation with language models
From my experience around less-used languages (with clojure on one hand and code aster's python on the other), LLMs may be able to generalize syntax but availability of APIs, functions, etc. is something that you can't solve by generalizing. Or more precisely, you can generalize but that means hallucinating non existing tools.
← PreviousPage 3 of 34Next →