htrp

35 ms·

871.
by htrp 2y ago
Great point, this is super insidious. there probably is a seedy underbelly of pay for play placement, link sharing, and all the other influencer equivalents of shady SEO tactics
872.
by htrp 2y ago
You most likely won't be an astronaut or a movie star, just like you most likely won't make it as an influencer.
873.
by htrp 2y ago
The expectation is that you'll have to have dynamically generated benchmarks with better eval at some point given the potential for brute forcing the private validation set.
874.
by htrp 2y ago
unclear why they couldn't have just done what normal people do and rented datacenter/warehouse space
875.
by htrp 2y ago
disappointing that large mega Corp does the exact same thing broke developers do to get around rate limits
876.
by htrp 2y ago
Academia is also usually a single person trying to get stuff working so they can publish a paper to get tenure/a job. Good SWE practices are 100% not taught in school.
877.
by htrp 2y ago
Not Sebastian (who I assume is the OP), but his blog/substack is also a great resource https://magazine.sebastianraschka.com/
878.
by htrp 2y ago
This! And a reminder that Phi1.5 comes in at 2.8GB. Getting a performant model even while using aggressive quantization is going to require more RAM. iPhone 16 ProAI may end up having notebook levels of ram (12 / 16GB)
879.
by htrp 2y ago
> Hiring on merit will be a permanent policy at Scale. > We believe that people should be judged by the content of their character — and, as colleagues, be additionally judged by their talent, skills, and work ethic. > There is a m
880.
by htrp 2y ago
Please tell me you guys called it "SuperCELL"
881.
by htrp 2y ago
> Their work on folding and ai could very well be a business worth in the hundreds of billions and they know it as well as many other bets. And we'll be writing case studies of how they squandered billions of R&D to help found o
882.
by htrp 2y ago
> Literally the only thing that can stop Google now is the fact they keep bringing Microsoft and Oracle flunkies into leadership positions. You mean the thing that's already stopped them? If they had seriously invested into the TPU
883.
by htrp 2y ago
I like to think of it as being focused on their core mission, and not trend chasing
884.
by htrp 2y ago
would love a blog post here of all the challenges (or lack thereof) they faced to get this to work
885.
by htrp 2y ago
I assume this is to do with misaligned corporate incentives.
886.
by htrp 2y ago
just like Google they fired the entire speech flow teams
887.
by htrp 2y ago
One thing that is underappreciated in the startup mythos is just closing up shop and trying again.
888.
by htrp 2y ago
https://www.nytimes.com/2024/06/11/style/ai-search-slop.html
889.
by htrp 2y ago
X to doubt.
890.
by htrp 2y ago
> Our foundation models are fine-tuned for users’ everyday activities, and can dynamically specialize themselves on-the-fly for the task at hand. We utilize adapters, small neural network modules that can be plugged into various layers o
891.
by htrp 2y ago
> Apple Intelligence is free for users, and will be available in beta as part of iOS 18, iPadOS 18, and macOS Sequoia this fall in U.S. English. Some features, software platforms, and additional languages will come over the course of the
892.
by htrp 2y ago
No life commitments and a social network in a new community built around their officemates
893.
by htrp 2y ago
https://science.nasa.gov/mission/insight/ > The Interior Exploration using Seismic Investigations, Geodesy and Heat Transport (InSight) was a NASA Discovery Program mission that placed a single geophysical land
894.
by htrp 2y ago
> Note it isn’t free to handle cash so comparing 3%-ish fees to 0% is wrong since cash is attractive to steal for both employees and others and requires time safes, armed guards, etc. Vastly important point here.
895.
by htrp 2y ago
Use standard cloud images > Something something incorrect cpu architecture for your Macbook
896.
by htrp 2y ago
>Getting my models dunked on by people who can't open MS Outlook more than 3 tries out of 5, however, have a remarkable depth and insight into their chosen domain of expertise. It's rather humbling. The people who have lasted i
897.
by htrp 2y ago
> The Dead Privacy Theory says approximately all private data is not private; but rather is accessible on whim by any data scientist, SWE, analyst, or db admin with access to the database. I like this so much I'm definitely stealing
898.
by htrp 2y ago
China.....
899.
by htrp 2y ago
You can't outspend the industry labs given the compute inflation in transformer architectures (unless you are ridiculously well connected in the venture/sovereign funding communities). And realistically, do we need another GPT4 ev
900.
by htrp 2y ago
This is exciting because it is an architecture that had so much promise, but we could never solve the gradient/parallelization problems better than transformers. This code will allow people yo experiment and see if it is a viable archi

More ›