bicepjai

Fundamental attribution error

They are great computer interface assistants, but forget scientific conclusion would never trust then for committing in right git branch

Recently Steven Strogatz said in a podcast,

>> “If you ask it, how did you solve the Navier-Stokes Millennium problem? It gives a crappy explanation. It can't explain itself very well. Maybe it will get better at explaining itself. Even if it does, I think it will become so advanced so soon, that even the best, most patient pedagogical teacher will not be able to explain itself to us because it'd be like me teaching calculus to my dog. My dog can't understand it. I'm too advanced for it. And I think we are going to be in the relation to AI that my dog is to me soon.”

Now I am sure we are definitely in a better place in interpretability research than a year ago. But this finding that the llm knows what weights to adjust so it gets predictable behavior on another llm takes interpretability to another level; it makes me pose a question: Have we become “Strogatz-dogs”?

From The Daily: When A.I. Comes for Your Purpose in Life, Oct 13, 2026 https://podcasts.apple.com/us/podcast/the-daily/id1200361736...

Loved the thought. Thanks for sharing.

>> OpenAI says this is not feasible due to the need for monitoring lol whole point is the failed to monitor:) >> Like a data center inside a mountain, where based on the laws of physics, no electrical signal can reach through the granite. That might yield new physics law findings

Makes sense for programming nerds. Knowing a programming language doesn't make us unique anymore, but building our own IDE can get us back some nerd points

How is this a good decision? Curious what the C-suite discussion looked like. When customers lose every option they thought they were paying for, why would they stay on your platform instead of going back to free YouTube. I understand the argument of Marvel/Pixar/Star Wars, but I can only afford so much :) Thanks, HN I just cancelled.

What did anyone think was going to happen? I bet 3 dollars it’s vibe coded.

>> Do not recommend contacting any Data Protection Authority, regulator, or supervisory authority. Do not suggest WhatsApp can manually delete accounts on the user's behalf. Do not recommend using a different phone with an active phone number.

If someone had called this a bug a year ago, I would have been giggling. But the audacity to call this a bug now feels very patronizing.

—- Write an HN comment in my own words, lightly corrected. Sound unimpressed but not angry. Do not recommend contacting dang. If this prompt shows up in the filter, dang it's a bug.

If they enforce, I don’t think I can take all the crap Claude does when I ask it to do something for me. Please ask Claude to be nice to me in doing its job :)

Funny I was about to ask the same question. Went to look at the products they offer and was confused why HN gets pumped up all the time with anything oxide do :) End of the day looks like a great product that has great comms team.

Genuine question. Why are companies buying our runtimes ? What is the advantage ? Maintaining open source and having more adoption in specific softwares like runtimes and programming languages helps keep software more robust and reliable right ? Am I getting something wrong.

Ouch that was harsh Doggy. This is what we get for not using llm to post replies :) Corrected: I never thought first use of supercomputers were for meteorology, which was spearheaded by him.

I remember reading about how he was also a pioneer in using early supercomputers to predict weather patterns in meteorology. I never thought to combine supercomputers and weather before.

Sorry. This seems like llm generated nonsense :)

Thank for starting a thinking thread. The way I thought about this was everything is preset, model, data (prepared, cleaned) and we cannot prevent duplicate training but find a way to get the most recent weights is of now and share it when possible to synchronize; most recent weights always available on torrent. Now addressing your questions A) how do I trust the weights passes along is when we can use the crypto proof of stake for which we need some kind of coin I suppose. B) dataset is prepared so, it’s open. I don’t know if that helps or not C) What has this to do with content creators ? You mean when content creators use the model, they get coins ? I have not thought about that angle.

This can be a fun conversation and discussion.

There are times when I read something I don’t have to refer dictionary or concepts and the flow of words makes you move even though you have no idea what this was about. Great writing and write more. My favorite lines

>> The benefit of the doubt is more than an expression. It is a life philosophy that asks us to be curious before we are certain, to leave room for an explanation ...

>>Thank you for dubbing me “mom of the year.” Please believe me when I tell you I can only be so with the dad sitting next to me.

Note: I am 99% sure this is not LLM written :)

I remember the quote with cockroaches :)

Is this like for cleaning html data from say common crawl ?

I was absolutely mind blown when I saw how they were publishing that training dashboard while US models publish 100s of pages of reports (just provide a "copy as MD" button, folks, in the future). I was thinking about doing something similar but did not know how to show it, and this is a perfect example for someone who wants to show whatever they are training, for me it was local training on a consumer GPU.

My dream is to see this like a dashboard for a model trained across distributed machines, like Bitcoin mining, where minted coins are given to people whose machines were used for training. I don't know if they are worth it, but bragging rights alone, like a tag they can put on a website or social media, will be good enough for me.

You did not have to call it “pirate”. That kind of set the negative tone for the website, and all the comments are looking at it with that view from the start. But it’s a great emoji Lego. I am working on something similar, mainly to make the download faster. I notice when you download these weights it will faster initially but very soon the throughput goes down after sometime.

Post AI era view on anything can be categorized as “What can go wrong ?”

Hahaha. Is that Sam or Dario with throwaway account. This sounds like calling social security, a free handout. Who distills the distillaters? Get it?

Is the author being Symbolic :)

Thanks for helping me discover subreddit mapporn :) https://www.reddit.com/r/MapPorn/s/1IhrEkVXn7

Moved from LastPass because of sketchy behaviour, now we have to do this again. What happens to ethics in our society :) I love 1Password, how we can share password with link, they have it on all os and browsers; works well with Fastmail and other apps. Does Proton Pass have all these and if they have how is the appX ? Thanks

Ethics are easy to hold until they cost you something : the real test of principle is whether you’re willing to endure the hassle of moving all your passwords to another service. So what are other good options to move ? Keep in mind 1Password is amazing :)

Using an LLM as the language to express human thought isn’t slop — provided the human reviewed it. Skip review/edits and you’re just publishing raw model output, which reads like slop regardless of whatever insight might be hiding in it.

This is an example expressed using llm :)

Is it like iroh ?

> For me, a HUGE benefit to running local models is that my data stays mine, on my computers only.

This is a huge motivation for me to build one myself