HNHacker News
TopNewBestAskShowJobs

shermantanktop

5,227 karma · joined July 10, 2023

submissionscomments
shermantanktop··on What will our economic future look like?
Insurance and HMO CEOs salaries come from somewhere, and it’s not malpractice judgments.
shermantanktop··on Shopify acquires Tailwind
Funny, my opinion is the opposite. Excel is a much better spreadsheet than Word is a word processor or than PowerPoint is a presentation tool.

Of course it’s got quirks and unfortunate relics of its long history, but I reach for it willingly, and I can’t say that about the rest of the Office suite.

shermantanktop··on Large language models develop novel social biases through adaptive exploration
Are you sure? I mean, yes of course tiny sample windows have this effect. But it seems at least possible that LLM is more prone to this effect when making estimates of human performance or behavior than when doing it for other topics.

In any case I feel the paper is interesting but almost begging to be misinterpreted.

shermantanktop··on Large language models develop novel social biases through adaptive exploration
There is such a thing as lack of bias in statistical outcomes, right? E.g. fair dice? Measuring it may be probabilistic, but it exists.

What I'd like to see is if the LLM would exhibit the same behavior wrt other types of predictive selections. For example, rather than choosing people from four tribes, choosing flower seeds from four packets, or choosing lottery tickets from four machines.

shermantanktop··on I-have-ADHD: A skill to stop coding agents from burying the answer
“It is hardly my place, sir, to criticize the facial peculiarities of your friends.”
shermantanktop··on Large language models develop novel social biases through adaptive exploration
I knew before I opened this comment section that it would trigger a bunch of reactions, all because of the word “bias.”

Please just go read the abstract; your first reactions to the headline may not be relevant.

shermantanktop··on Large language models develop novel social biases through adaptive exploration
I think the paper is about bias formation, not reflecting existing bias.

If the formed bias was against HN usernames that started with “r,” would it still seem daft?

shermantanktop··on Show HN: I built an aesthetically pleasing puzzle
I was baffled by this because the rightmost square didn’t seem to have any pieces where the letter would fit vertically — the unplaced pieces had a border which made it impossible to see which might.
shermantanktop··on Splash-free urinals (2025)
The gents loo at a non-posh UK pub is about the lowest you can go in the western hemisphere.
shermantanktop··on Splash-free urinals (2025)
At home? 100%. Out in the world? Very much depends.

I was just in a stall where the inside of door had a pictogram forbidding users from standing on the edge of the bowl and squatting. I assume this might be for visitors from countries where squatting is common.

Point being, you just never know what random members of the public will do with a door to hide behind.

shermantanktop··on Ganon's Mysterious Origins (Revisited)
I read through those links (ha) and I agree, they knew the overall story was interesting to players, but intentionally elevated story over continuity. That was my original point.

The same debates happen about early games made with far less intention and artistry. For those games I’d be surprised if there was much care about canon at all.

Especially when the studio which made game v1 was fired and a new studio was asked to take some random game in development and update the sprite to look like v1’s box art. And 40 years later people are arguing about timelines and character arcs?

shermantanktop··on 'You Can See Everything' Review: Nathan Fielder's Doc About Elizabeth Holmes
Here’s how I think it makes sense: accountability for unethical behavior arrives via more or less direct routes depending on who you are.

The more you look like a high-status leader, the longer it will take, and the more likely it arrives as a fine or a suspended sentence followed by an eventual redemption story published in the New Yorker or equivalent.

But targeting other high-status people ejects you from the white glove treatment.

shermantanktop··on Smartphone makers don't bother to comply with EU repairability requirements
Being in favor of right-to-repair is generally popular on HN, for obvious reasons.
shermantanktop··on Ask HN: How do you manage skills files?
I don’t seem much point to intentionally curating a set of skills, and specifically invoking them by name, only to watch the ai skip them all and do better by just reading code and internal/external websites.
shermantanktop··on Ganon's Mysterious Origins (Revisited)
Sometimes? I’m no expert, but I’m pretty sure these games were generally not built with a hefty bible full of inviolable lore. Plot and continuity errors are common enough to show that.

Some people are very invested in making sense of things that just don’t have a ton of intention and meaning behind them.

shermantanktop··on Invisible Companies
Alas, mine is far too visible. But I will spare you the associated downvote.
shermantanktop··on Three sites made 215,128 “best software” pages for AI. Perplexity cites them
If you don’t know how to do that now, the fastest way to learn (for most people, surely) is to have an LLM teach you and critique your efforts. And then you’ll realize that it’s unlikely you’ll get practice at doing it from scratch, since the AI is faster and probably better.

Human learning is slow, AI production is fast. Very hard to steer a middle path.

shermantanktop··on The Snow/Leavis ‘two cultures’ clash
The only important question in this space is how to balance the speed of development of this technology with safety. And globally, as a human race, our answer is: speed. Yes, lots of AI scientists spend lots of time and energy on safety. But all the incentives (economic, reputational, professional, geopolitical) are on the side of speed.

So sure, there are introspective AI scientists. I fear their role in the history of this period will be a footnote about cautionary voices.

shermantanktop··on Humanity has built the records of FATE by accident
> Recently, I was at a baseball game with my partner. One of the reasons I enjoy going to baseball games in person is to watch the crowd. As I looked around, however, I noticed something different: many people had tuned out of the game and were instead on their phones, talking to various chatbots: ChatGPT, Claude and Grok.

I’d like the author to describe how they knew what specific applications these many people were using on their phones, all around them at a baseball game.

Many of them were actually on social media, which itself is a form of algorithmically driven mass influence. But because the content was typed in by other people, we’re ok with it, while the robotic token producer we call AI is an evil force?

shermantanktop··on The Snow/Leavis ‘two cultures’ clash
He points out the value of introspection, which the literary culture has in spades, but the scientific culture utterly lacks.

If we get the worst outcomes from AGI, the scientists who worked so hard to enable it are unlikely to reflect deeply about it. If they had that capacity, they might do it ahead of time. And some do…but not enough to stop the inevitable tide.

Ironic, given the role that chance and statistical prediction plays in the mechanisms of AI. Free will is in fact a thing, and moral choices have moral consequences.

shermantanktop··on Show HN: The load-bearing vocabulary of Claude
“Orthogonal” is similar. It’s been commonplace for decades in my world but seems Claude-ish.
shermantanktop··on Autism mutations drive neurodevelopmental pathology
You can invert that. Why do people with similar or the same genetic code (identical twins) have such variation in symptoms?

Ultimately, the symptoms are what people experience, and treatments often center on mitigating symptoms. Having an insight into the underlying genetics may or may not help anyone live with a disease, especially not in the near term.

shermantanktop··on Six Months of Writing Code Exclusively with Agents
Building understanding through use, as opposed to reading, is not a controversial idea.
shermantanktop··on Nitter and XCancel receive cease and desist notices
Literally the next thing that happened after I posted this yesterday: my kid texted me that Dolly Parton had died. That’s how some news gets to me—- it’s important to someone I know. It turns out that that is both a good filter and fast enough propagation.
shermantanktop··on CEO fired developers to make room for AI. Developers create open source AI CEO
The speech model for the voiceover should be named “Max Vocalfry.”
shermantanktop··on Mechanical Turk shutting down September 30
The tagline for MTurk used to be “artificial artificial intelligence.”

It appears now we can get along with just a single “artificial.”

shermantanktop··on The Hugging Face incident and the road ahead
If we acknowledge that humans are fallible, is human judgment unnecessary? and what replaces it? Pre-codified behavior rules are just delayed human judgment, and have holes. Machine judgment is very the thing you are trying to control. What's left?
shermantanktop··on AWS Acquires DuckLabs
Does LinkedIn activity paint an accurate picture of wherever you work?

I can’t say that’s true for me and my company. It’s not all lies or anything, it’s a fragmented view that emphasizes conflict and the most visible 10% of what the company actually does.

shermantanktop··on AWS Acquires DuckLabs
Anyone trying to use a kv store for relational workloads is The same kind of person who uses a kv cache with durability features instead of a kv store. You can’t blame the tech for their mistakes.
shermantanktop··on Dolly Parton has died
becomes? i think that happened like 15 years ago...
← PreviousPage 5 of 34Next →