sashank_1509

58 ms·

541.
by sashank_1509 4y ago
This sub and blog post are all talking about ChatGPT which is no doubt amazing and far ahead of the curve. However I would also point to Metas new vision model: SAM (Segment Anything): it is so far beyond any other vision model, I actually
542.
by sashank_1509 4y ago
Extremely impressive system. Blows everything else (including CLIP from OpenAI) out of the water. We are inching closer to solving Computer Vision!
543.
by sashank_1509 4y ago
in the real world, when you write war criminal, people think someone convicted of a war crime, it’s really not meant to be used a figure of speech. John Yoo is still a UC Berkeley Law Professor. You’re making war criminal as meaningless a t
544.
by sashank_1509 4y ago
What are you on about, who’s a war criminal, genuinely curious?
545.
by sashank_1509 4y ago
Law means anything if its applied equally to all sides. Biased applications of law, does not mean anything. Nuremberg Trials were political: We humiliated and destroyed the Nazi's for loosing and killing 6 million jews. Nothing wrong w
546.
by sashank_1509 4y ago
Alright let me make the argument why indicting Trump is a horrible idea. It's really not hard to see why, if you have any insight into what the other side thinks: 1. First there are two alternative realities being played in real time
547.
by sashank_1509 4y ago
This is how our society becomes the dystopia in Atlas Shrugged. Every single fast moving technology that we do not understand, needs to be stopped on its track and regulated to check for “safety”, “inclusion”etc because really the biggest p
548.
by sashank_1509 4y ago
It’s per 100km. I tried a few prompts, wolfram didn’t get it right!
549.
by sashank_1509 4y ago
Its all to do with how the API is designed. Now the prompts need to be sent in the form of chat, and we need to worry about text being output from the model, and not just code. Basically, users need to comb through their codebase, change th
550.
by sashank_1509 4y ago
First I asked it, what is something you rub one your body. It answered sunscreen. Then I asked can you make a 6 letter word from given letters. It answered lotion. Then I asked can you make a 7 letter word and it just repeated its previous
551.
by sashank_1509 4y ago
This is better answered in stack overflow but x is an array. I think a better written string is e ^ (x_i - max(x)) / sum_i(e^(x_i - max(x))" x_i is a specific element in the array x. Basically an array of real values is converted
552.
Simplest thing GPT4 can’t do?(lesswrong.com)
6 points by sashank_1509 4y ago|9 comments
553.
by sashank_1509 4y ago
I’ve found it quite bad at unscrambling letters. Asking a seven letter word from “solution” I got , outlins (not a word), loutins( a long explanation about how this relates to hawks which was also made up, this is not a word) and after lots
554.
Show HN: Lightweight YAML Config CLI for Deep Learning Projects(github.com)
1 points by sashank_1509 4y ago|0 comments
555.
by sashank_1509 4y ago
Polgar sisters clearly proves a genetic point if you notice their ancestry? But I dont want to go individual examples, because individual geniuses possibly exist in every genetic pool. Let’s talk about averages where genetics do matter. On
556.
Ask HN: A good project for getting into Language Modelling?
3 points by sashank_1509 4y ago|2 comments
557.
by sashank_1509 4y ago
I think they’re planning to release fine tuning for gpt3.5 later but the only way to avoid a long prompt at start is fine tuning
558.
by sashank_1509 4y ago
I think it’s fine you’re focused on retrieval but you should add that as a caveat to your results, 100 times better at retrieval. As an ML researcher in grad school here’s what >80% use case of clip I’ve seen: 1. Take a random image, tak
559.
by sashank_1509 4y ago
You seem to be grossly mistaken: https://openai.com/research/clip Image retrieval is mentioned zero times, original CLIP was built for robust classification. You provide a thousand words and contrastively it shows you
560.
by sashank_1509 4y ago
They seem to be only testing for the image retrieval task, but I don’t think CLIP is actually used for image retrieval. Most cases, I see CLIP being used for semantic segmentation, detection etc. Do these guys have similar results on these
561.
by sashank_1509 4y ago
the highest murder rate in US is 15.8 Louisiana while the highest in India is 6.5 (Patna). Most places in India have half or lesser murder rates compared to the average in US. India in general is just a much safer country, not in the same l
562.
by sashank_1509 4y ago
I wouldn’t be surprised if this is the case and I come from ML research which tends to have a very high standard of transparency and we still end up with nonsense results
563.
by sashank_1509 4y ago
Well if it’s about QED I’m always drawn to the paper : “There’s something rotten about the state of QED” by Oliver Consa ( https://arxiv.org/pdf/2110.02078.pdf ). Since I’m in CS, not physics could anyone let me know if