HN user
arunkprasad.com
(2022 - 2025) Evergrow, Head of Engineering (2017 - 2022) Pinterest, user+pin modeling (2016 - 2016) Startup (2013 - 2016) Google, query understanding (2008 - 2013) Stanford, BS + MS
Maybe it's just the frequency illusion, but "X. Not Y." in particular is a pattern I strongly associate with LLM writing.
That’s confabulation. Not a metaphor. The same phenomenon.
Published. Replicated. Not fringe.
Not to validate it. Not to refute it. Not to engage with its content at all.
A similar idea from the Brihadaranyaka Upanishad, ~7th century BCE
'And here they say that a person consists of desires. And as is his desire, so is his will; and as is his will, so is his deed; and whatever deed he does, that he will reap.
What is the strategy, in your view? Maybe something like this? --
1. All government employees get access to ChatGPT
2. ChatGPT increasingly becomes a part of people's daily workflows and cognitive toolkit.
3. As the price increases, ChatGPT will be too embedded to roll back.
4. Over time, OpenAI becomes tightly integrated with government work and "too big to fail": since the government relies on OpenAI, OpenAI must succeed as a matter of national security.
5. The government pursues policy objectives that bolster OpenAI's market position.
I can't find an exact quote either, but AFAICT he wrote extensively on extensions and amputations, though perhaps less concisely.
As a side project, I'm creating resources for learning Tamil, my parents' native language:
https://akprasad.github.io/tamil/
It's been a lot of fun getting the basic tools going: transliterators, morphological generators and analyzers, and some other things on top. But the main goal is to improve fluency as quickly and efficiently as possible.
I use something similar: .. , .2 , .3 , etc.
The author agrees with you:
Not all requests for help are predatory. It’s part of your job to help out engineers on your team, and cross-org impact really does involve helping others sometimes, even if you get nothing in return. Predatory behavior is a consistent pattern of drawing on your time for nothing in return.
For context, Andreessen is talking here about the Biden administration, and his revulsion to this approach is why he endorsed Trump:
I, you know, look, and I would say like when we endorse Trump, we, we only did so on the basis of like tech policy. [...] Number two was ai, where I became very scared earlier this year that they were gonna do the same thing to AI that they did to crypto.
In Hindi and some other Indian languages, it is common to drop the last short “a” of a Sanskrit word, hence Ram for Rama, Vikram for Vikrama, etc.
You're thinking of koans [1], which are a specific kind of Zen speech used to shift the mind out of conceptual thinking. Outside of that context, Zen teachers often just talk in conventional language.
I cut a digression on forgery to keep my comment short, but my focus is less on provenance and the habits of art owners and more on what it feels like to look at a painting that you know is forged (or copied, etc.) The mental story around that painting, which is half of the fun and most of the meaning, is quite different.
For many people (most?), part of appreciating art is to engage with it beyond mere color and shape. I mean context about the artist themselves, and their life and work, and what they were trying to articulate, and why they made the choices that they did. All of this is embedded in a social context in which art expresses intention and meaning and some artifact of human experience.
It's because many/most people feel and understand this context that they react so differently to artwork that they know is forged. If a painting were just an issue of color and shape, forgery would be irrelevant to how we relate to art. But for most, a painting is embedded in the social context I described above, and forgery feels like a violation of that contract.
AI art seems to fall in a similar category, since AI systems are not social agents and not human beings.
Those were fun times! (& great to see you again after all these years). It's astonishing to me how far the tech has come given what we were working on at the time.
The title suddenly reminded me of this beautiful story [1] I read more than a decade ago, where spondylus plays a key role:
[I]t was a sense of morbid attraction that led me to Diane Ackerman's newest book One Hundred Names for Love, in which she documents the stroke and subsequent language deficit suffered by her husband, novelist Paul West.
[...]
Oddly, it was often the most obscure words that were easiest to recover. He struggled with words like blanket or bed, or his wife's name Diane, words that you would think over time should have seeped into his genes. Nevertheless, he could recruit words like postillion or tardigrades to get an idea across.
[...]
At times, his speech sent even his verbally-endowed spouse scrambling for a dictionary. When trying to ask her whether she'd received a check as payment for some work done, he resorted to the word spondulicks, which prompted the following exchange:
> "What's a spondulick?"
> "Money."
> "Really? Truly? Spondulicks?" In my mind's eye, I pictured a spastic duck.
> "Yes," he said emphatically.
> "Spondulicks?"
> "Spondulicks. It's British."
> Surely he was pulling my leg. I breezed into the library to look it up in an etymological dictionary, where I found this entry:
> 1856, Amer.Eng. slang, "money, cash," of unknown origin, said to be from Gk. spondylikos, from spondylos, a seashell used as currency (the Gk. word means lit. "vertebra"). Used by Mark Twain and O.Henry and adopted into British English, where it survives despite having died in Amer.Eng.
Over time, West recovers:
The intimate wordplay between West and Ackerman also eventually resumed, with West fashioning novel terms of endearment as gifts to his wife. The offerings were delightful. Deprived of the usual routes to language, and along with them, the common clichés that many of us struggle to shed, West bestowed on his wife exquisite pet names such as: My Little Bucket of Hair; Commendatore de le Pavane Mistletoe; Dark-Eyed Junco, My Little Bunko; Diligent Apostle of Classic Stanzas. And at one point, the man uttered what has to be the most searingly romantic sentence ever uttered in history, by anyone, in any language:
"You are the hapax legomenon of my life."
This quote is from Larry Page, one of Google’s co-founders.
Of course it is an improved experience, but I don't then get to say that I'm a good cook.
I've updated my comment to explain my view, which I think closely aligns with yours.
Yes, much in the same way that hiring someone to cater a dinner party makes me a great chef.
(edit to give some body to my comment above:
Hosting a great dinner party is hard work and requires coordination between food, decor, seasonality, people attending, etc. It is akin to a director coordinating the parts of a film. So I do think hosting a good dinner party can count as artistic expression.
I don't know the parent comment's intended reading, but I was reacting to the idea that typing a Sora prompt makes someone a good artist. If the parent means instead that AI allows people to coordinate multiple media in a broader expression that was not possible otherwise, then I fully agree.
)
I'm trying to point to the set of all jobs a human being could do, which includes future jobs enabled by future technology.
This is not as nebulous of a set as it sounds because it has real human boundaries: there are limits to how fast we can learn, think, communicate, move, etc. and there are limits to how consistently we can perform because of fatigue, boredom, distraction, biological needs like food or sleep, etc. The future is uncertain, but I don't see why an AI system couldn't push past these boundaries.
do we really have any data to back that up
By definition, we don't have data for events we haven't seen before. So instead I reason as well as I can:
Consider the set of all jobs a human being could do. Consider the set of all jobs an AI system could perform as well as a human being but more cheaply. Is the AI set growing, and if so, how quickly?
Prior technology is generally narrow and dumb: I cannot tell my cotton gin to go plant cotton for me, nor can I ask it to fix itself when it breaks. Therefore I take on a strategic role in using and managing my cotton gin. The promise of AI systems is that they can be general and intelligent. If they can run themselves, then why do I need a job telling them what to do?
Sentence 1 seems historically illiterate, and I think pg knows how ridiculous it sounds because he walks it back almost immediately. "Historically people made a sacrifice, but not intentionally, for the most part" is incoherent.
No one weaves now, and that's fine.
Did horses find new jobs when we moved to steam power? Leave aside the odd horse show and fairground ride. By the numbers, what do you think happened?
I'm reminded of Maciej Cegłowski's "The Website Obesity Crisis" (2015):
The article agrees:
Of course, businesses have long segmented customers
The article's argument is that:
- This kind of segmentation is more widespread and automated than it was previously.
- It's appearing in areas it wasn't in before (e.g. Disney World).
- It's extending to public arenas like vaccine queues, access to public lands, and TSA queues.
You might also be interested in the Charvaka school of ancient India [1], which is a close counterpart to Epicureanism. The Charvaka school was likewise influential and widespread, and it likewise become obscure over time, for reasons I don't know.
for the “smart” person, failing at a task takes a stab at the heart of their identity. Self cushioning in a cloud of rationales “well, I can’t do X but at least I’m smart” means that if “being smart” is ever questioned, the possibility of them failing terrifies them.
Yes, I see it much the same way. If you're curious, I wrote about a related theme a few years ago here [1].
When the work is uncertain, she withers: uncertainty means the unknown, which means risk, which means failure, which means she's worthless. The thought of fumbling in the void terrifies her.
The piece defines its use of "soul" quite clearly:
But there is a fourth sense in which when we talk about [a soul], we all mean more or less the same thing: what it means for someone to bare it, for music to have it, for eyes to be the window to it, for it to be uplifted or depraved. Even if, religiously, we know by revelation that other people possess them for eternity, we only engage with or know anything about them at a quotidian level by way of the same cues and interactions that a more this-worldly view would take as their sum total: bright eyes, a dejected slump, a sudden manic inspiration or a confession of regret.
Also Embraer [1].
I'm not so sure. Where a hypothesis is vulnerable to evidence, a butterfly idea seems vulnerable to emotion:
I’ve been on the butterfly crushing end of this myself- I’m thinking of a particular case last year where my friend brought up an idea that, if true, would require costly action on my part. I started arguing with the idea, they snapped at me to stop ruining their dreams. I chilled out, we had a long discussion about their goals, how they interpreted some evidence, and why they thought a particular action might further said goals, etc.
The author mentions "dropping [his] religion" here [1], which undercuts the idea that he is lying to himself about his religious doubts.
I don't think much good comes from analyzing a private stranger in this way, especially given such limited evidence.
I'm glad that the author enjoys Ruby! As someone with limited Ruby exposure, these quotes stood out to me:
The language is meant be joyful to use. [...] Everything else that Ruby is stems from this value.
This is important and underrated. I think many programmers have a bias that working on a difficult problem entails using a "real" programming language with sharp edges. I had some version of this bias for a long time until I started exploring the most recent generation of systems languages.
Well written ruby code can often read like natural language.
I see where the author is coming from, but I find a healthy dollop of symbols to be very helpful for reading and understanding code at a glance.
Feeling recognition in the language you’re programming is so powerful.
This is the feeling I had the first time I used Python, and later Rust. It's a wonderful feeling!
[As] Kent Beck said at RailsConf in 2020, “Software design is an exercise in human relationships.”
Especially true given all of the components involved in supporting a language: compiler, docs, standard library, third-party libraries, package managers, frameworks, formatters, profilers, ...
It depends on what you mean by "system language." For me, this category mainly includes languages that provide fine-grained control over memory management (C, C++, Zig, Rust, ...), so I personally wouldn't include OCaml.