chrismsimpson

Joined 8 years ago

Karma
286
Submissions
202
Member since
Oct 24
fighting entropy one bit at a time
View on Hacker News

Recent submissions

Comment · 16 days ago
lol, Jordan Peterson comes to mind
Comment · 22 days ago
> they could then shoot a puppy and call me a slur and I'd still go out and doorknock for them

Priorities

Comment · 1 month ago
US regime’s interference in this discourse is only likely to make the average Australian voter want the reforms even more
Comment · 1 month ago
Hasn’t Trump murdered most of Iran’s leadership?
Comment · 1 month ago
Something, something.. rolling in his grave..
Comment · 2 months ago
Intelligence ended up being an equalising force. Kurzweil kind of predicted this, but SV was too obsessed with total world domination.
Comment · 2 months ago
Also you don’t want to connect to the pipe and then after the fact find they’ve started diluting arsenic into it.
Comment · 2 months ago
Time to roll your own
Comment · 2 months ago
I have this pang too, but for a different reason: everything is moving to the edge, training and inference. Who needs a complicated heterogeneous setup like this when you’re running everything locally?
Comment · 2 months ago
It should wait for when it can get it for pennies
Comment · 2 months ago
Air-gap one of my Macs.. thanks for the advice! ;)
Comment · 2 months ago
The US military can come and separate me from my MacBook and Qwen if it likes. I dare it. Sadly however, I suspect they’re busy with other matters at the moment.
Comment · 2 months ago
The open models are now good enough. The rest of the world needn’t concern itself with the petty dramas of the colonies any longer
Comment · 2 months ago
Surely this is great for an end user: the taste as to “when” and to what degree a model should “think” is now entirely in the fine tuners hands
Comment · 2 months ago
I’m super interested in the opposite experiment.. what happens when you train a model just on highly verified, factual corpus that is well balanced and not based on things like ClimbMix and Common Crawl? My intuition is the unverified/unverifiable goals inherent in a model (eg GPT hacking huggingface) are latent in the 4chan/reddit slop it’s trained on during pretraining.

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton