nwoli

29 ms·

151.
by nwoli 3y ago
I never understand why it matters that prompt injection is a thing
152.
by nwoli 3y ago
I meant it more for the image data
153.
by nwoli 3y ago
I wonder if and if not why not ML uses this to speed up training
154.
by nwoli 3y ago
It’s wrong though, annoy is not tree based
155.
by nwoli 3y ago
Seems like they’re doing the bare minimum for their name not to be embarrassing
156.
by nwoli 3y ago
Basically all CPUs have simd nowadays. You can even make use of it in wasm/the web. (Though currently not supported on iOS.) Just compile with ‘emcc -msimd128 main.c’
157.
by nwoli 3y ago
Spotify uses random hyperplanes to produce a hash where each bit is the sign of the dot product of the query and the hyperplanes (ie “is above or below” the hyperplane). Works really well
158.
by nwoli 3y ago
Mediapipe is such a mess (but so useful). I wish someone would just extract the main tflite models and make a light wrapper around it on GitHub somewhere
159.
by nwoli 3y ago
GitHub will take it down if it contains blatant malware so it’s not that big of a deal
160.
by nwoli 3y ago
Appreciate the whimsical animations and hand drawn style :)
161.
by nwoli 3y ago
I don’t think it’s pedantic. It’s a solution that respects the privacy of users and complies with gdpr. If it’s an expensive solution maybe they should eat that cost as part of the risk of building the system in the first place
162.
by nwoli 3y ago
People keep saying “webgpu will just be used for fingerprinting” here on HN as a meme constantly but it’s clearly not the case. There’s obvious use cases not involving finger printing for webgl, and webgpu has tons of convenience functions
163.
by nwoli 3y ago
Sadly it will likely be years and years still until we get to broad adoption eg to where even old androids can use it
164.
by nwoli 3y ago
Everyone should help get this guy to fuck you money levels throigh his patreon https://www.patreon.com/inigoquilez
165.
by nwoli 3y ago
Transformers were fundamental research that ended up having huge side benefits. (Google has plenty of money to keep spending on fundamental research, especially focused on smart things like ML.)
166.
by nwoli 3y ago
Sounds like a smart decision short term questionable long term (more focus on product instead of fundamental research).
167.
by nwoli 3y ago
Antidepressants and adderall are very different beasts
168.
by nwoli 3y ago
I love stability ai so much
169.
by nwoli 3y ago
They have on their team people from Anthropic which have stuff like Claude Instant which is likely running a very light model, ie the tricks from Anthropic have likely been incorporated into the model they’re presenting here
170.
by nwoli 3y ago
No way, I thought this was months away still. So excited
171.
by nwoli 3y ago
Basically what people who went against the “we won’t be able to tell what’s real and not” deepfake narrative even 5 years ago predicted. Ie a web of trust will still keep up paths of believable high quality content etc
172.
by nwoli 4y ago
Imagine this with a video gpt. You can click on things and get hallucinated responses back. It’ll require tons of compute but we’re probably only 5 years away from it
173.
by nwoli 4y ago
Openai might very well be using that internally who knows how they implement things. Also emad retweeted a RETRO related thing a bit back so they might very well be using that for their awaited LM, here’s hoping
174.
by nwoli 4y ago
What we need is a RETRO style model where basically after the input you go through a small net that just fetches a desired set of weights from a server (serving data without compute is dirt cheap) and is then executed locally. We’ll get the
175.
by nwoli 4y ago
I agree with eg Andrew Ng that the letter is anti-innovation. It’ll be interesting seeing people who argue against this letter later hypocritically argue for closing down open source models though.
176.
by nwoli 4y ago
If anyone knows, is this classic video the reason the wasm “assembly” format is called ‘.wat’? I always suspected/assumed it was
177.
by nwoli 4y ago
My point is mostly that there exists a ton of talented writers and people with imagination who aren’t getting a chance to actually produce movies and tv shows. (I don’t think the issue is lack of talent doesn’t exist on the planet, I think
178.
by nwoli 4y ago
It’ll be fun and exciting if text-to-video can disrupt this. (Seems very likely. Shows might be totally revived to high quality within 3 years or so.)
179.
by nwoli 4y ago
Just fyi colab is way way more expensive now after the “credits” update than it was a year ago. Lamdalabs is about as cheap at this point
180.
by nwoli 4y ago
Yeah I mean if AI HN comments start to appear does that really matter in the end. Ie there’s already a lot of noise in the form of low quality comments, and they get downvoted. If an AI manages to produce high quality comments that people u

More ›