jerpint

30 ms·

451.
by jerpint 4y ago
Yes I agree. My understanding is you have in training dataset the original code, diff + commit message. So you train the LM to: Input: code+commit output: diff
452.
by jerpint 4y ago
I see chatGPT as a first iteration, and it’s glaringly obvious what works and what doesn’t. I’m hopeful future iterations (not just from openAI) will make it overall a net improvement, things like adding real sources and better handling thi
453.
by jerpint 4y ago
Can confirm it doesn’t work on brave on mobile
454.
by jerpint 4y ago
That’s unlikely to be how the system works, chatGPT by design is a probabilistic model meaning it’s very unlikely it’ll spit out the same output twice. The detectors probably use a vanilla language model that just recognizes the patterns in
455.
by jerpint 4y ago
Faith in the internet is restored
456.
by jerpint 4y ago
The most frustrating for me is the way the “offline” feature works. On my android, even though I had downloaded songs/podcasts/etc. I had to dig in settings and set it explicitly to offline mode to enable offline playback. Another
457.
by jerpint 4y ago
What happened to the polar bears
458.
by jerpint 4y ago
I think it would be helpful in the blog post to list the series of changes happening to the files before the animations
459.
by jerpint 4y ago
I saw the video and instantly recognized the manim style, it looks great!
460.
by jerpint 4y ago
I agree nothing in that article was a compelling argument
461.
by jerpint 4y ago
What a brilliant way to augment a dataset
462.
by jerpint 4y ago
ChatGPT can easily become much more impress with small improvements; Giving it access to other APIs / the web, handling images + voice, accessing databases etc.
463.
by jerpint 4y ago
One thing that could be useful for this is writing code locally and testing it on a remote server. I assume the simple act of editing reflects the change on the other machine? No need to commit, push, then pull to do a one-liner experimenta
464.
by jerpint 4y ago
I was thinking the same thing
465.
by jerpint 4y ago
It can also serve as a replacement for IR sensors in homes and keep thieves at bay
466.
by jerpint 4y ago
Once you use infinite scrolling you can’t go back
467.
by jerpint 4y ago
Does using this mean sending all of your potentially private data via an api call to openAI?
468.
by jerpint 4y ago
Is your comment self-referential?
469.
by jerpint 4y ago
Education today seems broken, perhaps this will be a way to fix a broken system. They don’t teach abacus use in schools anymore as well as how to use log tables
470.
by jerpint 4y ago
I wonder if they will train a DaVinci-004 for premium and keep 003 for regular
471.
by jerpint 4y ago
It’s more that without backpropagation, you no longer need to store your forward activations across many layers to compute the backwards pass which usually is dependent on a forward pass. When a network is hundreds of layers, and batches ar
472.
by jerpint 4y ago
Looks like it’s been fixed: https://twitter.com/BBCBreaking/status/1613174866047340545
473.
by jerpint 4y ago
Username is irl_chad …
474.
by jerpint 4y ago
Do you know how to do this? Also what would be the best software for live looping?
475.
by jerpint 4y ago
Is there an open source implementation?
476.
by jerpint 4y ago
Probably not, since you need to compute the activations of unknown inputs and there could be infinitely many variations of them
477.
by jerpint 4y ago
I love complex behaviours like these, I might try to implement it myself!
478.
by jerpint 4y ago
I’ve tried the jupytext approach and had problems when files get out of sync (which inevitably happens). So far, my best Jupyter/vim solutions are either to use a jupyter extension that enables vim bindings or use jupyter in vscode wit
479.
by jerpint 4y ago
VScode vim emulation is great! But when you just want to fire an editor to do a quick change, vim is your friend. It’s also reliably present on almost any server you’ll ssh into
480.
by jerpint 4y ago
Merry xmas to you too!

More ›