jerpint

36 ms·

541.
by jerpint 4y ago
No it’s really just tabular csv data, like a typical spreadsheet would hold. These datasets are rarely ever outperformed by deep learning compared to standard ML.
542.
by jerpint 4y ago
Most AI today is generally some form of machine learning / deep learning. When it comes to deep learning, python is by far the most popular language (libraries like PyTorch and tensorflow have a thriving ecosystem)
543.
Mnist VAE Clock(jerpint.io)
1 points by jerpint 4y ago|1 comments
544.
by jerpint 4y ago
Digital MNIST clock whose digits change by traversing a VAE latent space.
545.
by jerpint 4y ago
Third (better) option: have a regulating body have a separate, undisclosed test set. If you can't beat it, you can't deploy your model. If you can beat it, you still need to have your models peer reviewed and scrutinized
546.
by jerpint 4y ago
It would be really fun to have two language models talk to each other on a fake phone conversation
547.
by jerpint 4y ago
When that much money is at stake, it's hard to keep incentives aligned
548.
by jerpint 4y ago
Isn't the whole point of open source, so long as licenses and attributions are respected, that anyone is free to do whatever they please with these models and their redistribution?
549.
by jerpint 4y ago
I had seen a video demoing an attack like this some while ago and I started "wiping" keypads with my fingers so they're all "warm"
550.
by jerpint 4y ago
This is very informative, I never knew people "did this at home"
551.
by jerpint 4y ago
This was my same impression when trying out helix. Really nice and shiny, but all the shortcuts I've spent time creating over the years are gone and not easily reconfigured.
552.
by jerpint 4y ago
What is the value add compared to other free alternatives?
553.
by jerpint 4y ago
I actually didn't know this, and gives me all the more reason to be done
554.
by jerpint 4y ago
From experience, I found deploying with onnx and fastapi much more enjoyable than deploying with torchserve. Curious to see how easy they will make browser deployment and how it compares to tensorflow.js
555.
by jerpint 4y ago
I was a happy pixel and pixel 2 owner. After a pixel 5, I'm done with the pixel series. It doesn't feel as nice as it used to and overall feels finicky. Unlimited storage to google photos is great though.
556.
by jerpint 4y ago
Sounds like something to write a song about
557.
by jerpint 4y ago
187 comments in this thread (as of writing), 0 on thea original blog source. Someone should write a bot that tells a user "hey! Your post is being discussed on HN $here"
558.
by jerpint 4y ago
Everytime I need a refresher on transformers, I read the same author's post on transformers. Looking forward to this one!
559.
by jerpint 4y ago
I personally enjoyed the human aspect of the narrative
560.
by jerpint 4y ago
I wonder how constant everything else is kept, e.g. The seed. It's interesting that all the poses seem to align
561.
by jerpint 4y ago
I think this was the original vision behind teslabot. FSD is too hard, lets focus on AGI instead
562.
by jerpint 4y ago
I really like working inside a terminal, and I like being able to navigate seemlessly between code tabs and terminal tabs as though they are one. I can do this easily in vim+tmux, I can't do this easily in IDEs where the mouse is consi
563.
by jerpint 4y ago
I wonder what the solution will be in north america
564.
by jerpint 4y ago
Incredible! That conveyer belt design... Wow
565.
by jerpint 4y ago
I recorded myself speaking French and was able to translate decently well on my laptop. Very impressive!
566.
by jerpint 4y ago
It was CLIP, which also uses GPT-3 to embed text
567.
by jerpint 4y ago
Collect some data (and mushrooms at the same time) and I'm sure you can make a killer app :)
568.
by jerpint 4y ago
It probably would, remote sensing problems usually work best on hyper spectral data
569.
by jerpint 4y ago
Works on brave (after scrolling past a few ads)
570.
by jerpint 4y ago
This is very impressive. My go-to for spotting fake faces online has been to look for different poses of the same face. I'm going to have to find a new tell

More ›