Hacker News: elhospitaler

New comment by elhospitaler in "The Dawn of a New Era for Supernova 1987a (2017)"

elhospitaler — Wed, 23 Oct 2024 04:12:05 +0000

That time-lapse video is insane!

New comment by elhospitaler in "Self-Rewarding Language Models"

elhospitaler — Sun, 21 Jan 2024 00:54:31 +0000

It's common for ML papers to be posted on arxiv in advance of official publication (hence, preprint server). I haven't heard of the other authors but Kyunghyun Cho is legit and well known in the field.

New comment by elhospitaler in "Most AI startups are doomed"

elhospitaler — Tue, 28 Nov 2023 20:07:08 +0000

I think this is mostly true, except I imagine OpenAI has a more defensible head start due to all the user data they've collected from people's queries. If they can maintain their lead in model quality (which is definitely possible - the engineering task of training the best and largest LLMs is not for the faint of heart) and remain the "best" general use chatbot, I could definitely see them build an insurmountable lead with all of their user data.

New comment by elhospitaler in "Show HN: Trains.fyi – a live map of passenger trains in the US and Canada"

elhospitaler — Mon, 27 Nov 2023 18:18:24 +0000

Meanwhile, Europe: https://mobility.portal.geops.io/world.geops.transit?layers=...