ml_basics

9 ms·

31.
by ml_basics 3y ago
Pretty cool landing page, listen to the audio at the top of the page which gives a high level explanation and demonstrates the method
32.
Soundstorm: Efficient Parallel Audio Generation(google-research.github.io)
4 points by ml_basics 3y ago|2 comments
33.
by ml_basics 3y ago
Check out the scale, those are very tall peaks in the centre at 21km!
34.
by ml_basics 3y ago
Congrats to all the people behind this! To the naysayers that point out that OpenAssistant is not as good as ChatGPT: all of LLM development is still in the early days; the work being done by teams like OA is helping to shape the future dev
35.
by ml_basics 3y ago
Yep totally agreed. One of the things I'm excited to see going forward is multimodal models (trained e.g. on text + video + audio + images). I'm sure there's a lot more to it than this, but maybe one factor that makes humans
36.
by ml_basics 3y ago
> Where's the predicted society where automation allows us all to work for two hours a day, and spend the rest at leisure? Others have already commented on the fact that quality of life has generally increased. I think it's als
37.
by ml_basics 4y ago
Which ones do you mean? From my coarse understanding, seems like most places that didn't industrialise got colonised by ones that did. Maybe that's not the case with the development of the internet, but still it's the case th
38.
by ml_basics 4y ago
> Especially in a chatbot use case with cumulative prompting, which is the best use case for such a large context vs. the default cheaper 8k window. Depends on what is up with the images and how they translate into tokens. I really have
39.
by ml_basics 4y ago
What's the lifespan of an LLM going to be in the next few years? Seems like at the current pace, cutting edge models will become obsolete pretty quickly. Since model training is very expensive, this means the LLM space has some paralle
40.
by ml_basics 4y ago
From the paper: > Given both the competitive landscape and the safety implications of large-scale models like GPT-4, this report contains no further details about the architecture (including model size), hardware, training compute, dat
41.
by ml_basics 4y ago
> Some buses in London also have "Hail and Ride" zones, where a bus can be stopped with a wave on the hand at any point on the route. Unfortunately in my experience growing up in London, such zones really operated with informal
42.
by ml_basics 4y ago
There's probably a bias given that this is a website with global reach. For any company posting, almost all readers are not living nearby.
43.
by ml_basics 4y ago
Very cool. Worth mentioning though that the highlighted figures (1.12 tok/s for OPT-175B for "FlexGen with Compression") are for inputs of 512 tokens and outputs of 32 tokens. Since decoder-only transformer memory requirement
44.
by ml_basics 4y ago
Interesting idea RE PGP key as middle name... just it would suck to then later on lose access to your private key!
45.
by ml_basics 4y ago
Nice blog post. I wonder if it's possible to speed up the search with something like https://github.com/google-research/google-research/tree/mast... Also kind of surprising that something like this is no
46.
by ml_basics 4y ago
If anyone can find this I'd be very grateful, would be cool to read his comments
47.
by ml_basics 4y ago
Yes these are so good! Highly recommend.
48.
by ml_basics 4y ago
Is this the post you mean? http://itila.blogspot.com/2016/04/appendix-two-open-letter-t...
49.
by ml_basics 4y ago
Great book by a great teacher. I read this during my masters and was my route into machine learning. David did many great things but sadly passed away too young in 2016, surely would have continued to have a great positive impact in the wor
50.
by ml_basics 4y ago
Does that allay OPs concerns though? What happens if tracking-based ads are significantly more profitable than private ones (probably the case, based on the huge revenues of Google/Meta)? How high would the paid account fees need to be
51.
by ml_basics 4y ago
Super cool to see advances in large language models making their way into customer facing products, even if they are still not perfect. I hope that this forces the sleeping giant (Google) to wake up and start shipping more customer facing t
52.
by ml_basics 4y ago
It would be great if when clicking through the link, the relevant text could be highlighted in the webpage, similar to the featured snippets in Google search. E.g. when searching "What were the causes of the swiss civil war?" Goog
53.
Pronunciation of English "wh" (who, what, where,...)(en.wikipedia.org)
2 points by ml_basics 4y ago|1 comments
54.
by ml_basics 4y ago
> a deteriorating company culture Could you elaborate on this? Genuine question, because how Apple is so successful is a complete mystery to me. I worked there during the pandemic and my experience was that there is no company culture, i
55.
by ml_basics 4y ago
> There was also the issue that some EU countries(i.e Italy) were giving italian/EU citizenship to fresh immigrants from Africa so that they can travel to the UK. Source for this please? As far as I understood it is extremely diffic
56.
by ml_basics 4y ago
> Most bootcampers are gone, even ones that were performing well. What are bootcampers, does this just mean recent hires? Or people who came in with no specialist skills who rotate until they find a team?
57.
by ml_basics 4y ago
What's the name of your company?
58.
Transformer object detectors are less robust than CNNs(lakeraai.medium.com)
3 points by ml_basics 4y ago|0 comments
59.
by ml_basics 4y ago
GOFAI = "Good old fashioned AI" for those not familiar with the acronym
60.
by ml_basics 4y ago
Awesome! I wonder how long it will be until there is an open source implementation compatible with Stable Diffusion

More ›