steinvakt2

4 ms·

31.
by steinvakt2 4mo ago
Yes please! I have been frustrated with the state of object detection models especially. Everyone claims SOTA. So you end up having to test manually to find out which one actually is. And unlike LLM's, it should be pretty easily quanti
32.
by steinvakt2 5mo ago
Codex with GPT 5.5 is so good that I don't understand why people have this position. And I've been using cursor for a year, switching models (I did like both Opus 4.6 and 4.7)
33.
by steinvakt2 5mo ago
My art quiz is also tough: https://artle.eu
34.
by steinvakt2 5mo ago
https://news.err.ee/1609925000/estonia-s-arvo-part-remains-w... I guess it should have been worded "performed" instead of "played".
35.
by steinvakt2 5mo ago
Its the Norwegian word/pronunciation of "quiz".
36.
by steinvakt2 5mo ago
Thanks a lot. I'm afraid five questions is too much for most people and that they'll avoid coming back. Also, creating almost 100 high quality questions per month (30x3) is actually a bit of work. But thanks for the feedback. We w
37.
by steinvakt2 5mo ago
Better accuracy than whisper large? For English? What about multilingual?
38.
by steinvakt2 5mo ago
What do you consider to be the model with highest accuracy?
39.
by steinvakt2 5mo ago
Definitely something we will consider!
40.
by steinvakt2 5mo ago
Its been running for some months. New questions every day of course. Sometimes related to the season. Such as Easter questions during Easter. Glad you enjoyed
41.
Show HN: My retired dad and I made a daily, somewhat difficult, quiz(kviss.eu)
39 points by steinvakt2 5mo ago|20 comments
42.
by steinvakt2 5mo ago
Maybe https://kviss.eu/ is enjoyable for the people in this thread.
43.
by steinvakt2 5mo ago
What do you consider to be SOTA?
44.
by steinvakt2 6mo ago
This is not a new model. Also, it hallucinates a lot. Also, it's very heavy and slow in inference. It's also bad in multilingual. Edit: I'm talking purely about speech to text (STT). Not sure about the other things this can d
45.
by steinvakt2 6mo ago
Can't be just that. There was competition in the GPT-4 era. But we didn't get model drops every month.
46.
by steinvakt2 6mo ago
I’ve had great experience using opus 4.7 in cursor. Works for everything including iOS frontend
47.
by steinvakt2 6mo ago
This, as well as the NYT game, is heavily inspired by https://www.timdle.com ?
48.
by steinvakt2 6mo ago
And because she does less things in her day? Letting the brain process the instrument playing during day and night, not a bunch of other stuff. I’m sure brain age matters a lot but I think it’s a bit overrated. Probably a lot to do with env
49.
by steinvakt2 6mo ago
Many years ago. Not "a few years ago". Also you could make the sentence that 180 people lost their lives because of an evil war, of which USA and Israel are the aggressors. And we definitely don't talk enough about that part
50.
by steinvakt2 7mo ago
Not necessarily missed. Maybe just late.
51.
by steinvakt2 7mo ago
Seems weird. China is definitely falling behind. India is not.
52.
by steinvakt2 7mo ago
Still the weird case??
53.
by steinvakt2 7mo ago
This isn't enshitification. Most people have plenty of chargers. Its a good thing you can opt in/out.
54.
by steinvakt2 7mo ago
Don't you find transformer based models perform better? Are they too heavy?
55.
by steinvakt2 7mo ago
Let me know if you are hiring. I'm based in Norway and did my masters in AI computer vision (object detection, segmentation). Have also published a paper in medical imaging, as well as sold my AI webapp/SaaS for six figures $ some
56.
by steinvakt2 8mo ago
The recipient will get the forever chemicals, but they survive due to the blood transfer, so in a few weeks they can donate and get rid of the chemicals.
57.
by steinvakt2 8mo ago
EU chat control is also better than American government spying on American tech companies (which is effectively a kind of EU Chat Control, except its USA who gets to spy). Both are bad, but one is less bad.
58.
by steinvakt2 8mo ago
Interesting. Can we get in touch? I just sold my webapp/saas where I used NB-Whisper to transcribe Norwegian media (podcast, radio, TV) and offer alerts and search by indexing it using elasticsearch. Edit: It was https://mun
59.
by steinvakt2 8mo ago
I thought it was the other way. That 8k is problematic because the chance of dead pixel is so much higher, driving up the cost because of the higher ratio of ruined batches?
60.
by steinvakt2 11mo ago
Isn't "Proof of Humanity" kind of interesting here: https://proofofhumanity.id

More ›