498 karma · joined June 20, 2022
Specialization and generalization of hardware is also cyclical same as centralized and decentralized services.
https://arxiv.org/abs/2301.10743
Another interesting research topic is the trusted generation of tasks for finetuning
https://arxiv.org/abs/2306.08568
And I suppose too running these at the edge is terribly interesting too, if you can find analyses of "quantization" this is a highly active research are, and results are pretty incredible since it cuts resources by huge factors and no one quite knows why.
This is one that's easy to dive into with consumer hardware, but don't know any great papers myself
Run locally: https://github.com/ggerganov/llama.cpp
Quantized models: https://huggingface.co/TheBloke
Explainability is under research, though I haven't seen any good solutions.
This nay arise from skeptics who are calling the things stochastic parrots, incapable of reason, without a world model, etc.
This peripheral is obviously the future, and not apparently a CueCat situation.
It's gonna take doing to convince people it's not that TV screen they stick on your face at the dentist.
At a glance,
Things that leave a bad taste in my mouth:
- Genres > Metal > "Numb" by Linkin Park
- New Releases isn't organized by genre, which means you breathe through a straw, and hope Google brings a release to your attention.
- The "feed" New Releases which is apparently robo-tailored includes a Slowdive album from 1996, and a bunch of corporate stuff I'll likely never be in the mood for.
- no acknowledgement of labels, collectives, or studios. all very real constructs in current music.
- presumably live music recs are out even though those were never accurate to begin with
- there wasn't any on boarding, i guess im just to assume google knows who i am. went through onboarding, it's just there to slurp up artist relations and obtain feedback on recent trends it's inefficient and irritating out of touch
I like epitaph.
- playlist decayed and dropped songs, YouTube is even worse about this
- Play Music had a new releases browser highlighting new album releases by genre, you could "walk" into them in AA, then they bled that off
- as it stands "Play" doesnt have music anymore, and so the only discovery is YouTube's recommendations
- YouTube's recommendations are somehow difficult to control, overly gamed, repetitive, coarse, and generic -- some of which are contradictory. Google wasn't even able to protect Gboard from exploits like this with their unsated thirst for big data automation dominance
- they haven't begun to cooperate with other vendors to operate with assistant, instead picking winners, again
- for a span battery was concerning, it's not right now
- for a span Google's failed record of protecting user privacy, even now Google+ echo's when the Photos flow Share workflow walks me into privacy violating features
- i have qualms about providing an ad personalization business real-time granular data about my mood (though it's unlikely since my tastes are too obscure, anyway)
- i have qualms about subjecting myself to the opinions of a business with large enough reach to be an effective political entity and lever of corporate activism against my neighbors in society (particularly for an inferior product)
Hope this helps.
Edit: also the playback history was pretty bad IIRC, I assume it lumps audio playback with shorts, and shorts can't be organized into playlist despite their efforts. I need music to be a commodity, not a gift/promise.
> ... developers with newly created personal Play Console accounts will soon be required to test their apps with at least 20 people for a minimum of two weeks before applying for access to production.
Does this mean that a developer has to recruit 20 users through a special workflow, or the users will simply see that they are 'testers' for a new app?
Here's, at least, some official words about how the current admin's advisors might be thinking about the considerations for adoption.
Another interesting bit aside from the administrative task of ensuring constructive adoption of AI would also be facilitating progression, defense, and contribution.
Some more perspective on what the federal government is going on:
- https://www.cisa.gov/sites/default/files/publications/STRv4_...
- https://github.com/Trusted-AI/adversarial-robustness-toolbox
- https://www.darpa.mil/program/explainable-artificial-intelli...
- https://www.nist.gov/system/files/documents/2023/03/30/AI%20...
https://github.com/meltyness/cef-tex
The build system is a VS file with god-knows-what hardcoded, doesn't properly reference CEF or Godot, and is therefore locked to Windows platform only (though Godot can theoretically run everywhere), and is probably full of a bunch of other footguns that I haven't even thought of. Got 5 stars on GitHub just from tag discovery, and the wishful thinking of some Godot-ers that I'll ever have time to un-suck it. It's also got R-rated comments, a video demonstration, and a short written log of the weird hurdles I had to surmount.
There's a chance the perpetrator has simply compromised a legitimate copyright authority's account.
It's definitely disturbing that "basically anyone" has the capability to participate in this sort of censorship though.
Perhaps though the current frontier is the software->application abstraction that can be explored through public git repositories and Github "Topics".
Of course the application for this set would mirror the popular brain drains, short-form content production, video game mods/cheats/piracy, deep fakes and pranks, etc.
Just as the OS and operation of a few commercial packages you had access to in the 80s seemed limitless, so is the landscape of FOSS packages.
I mean for sure I agree with their conclusion that Cisco has neglected innovation but they don't really expound on Cisco's real missteps clearly.
I guess I recall some supporting facts; closure of NH optics plant, Webex splintered to Zoom, but there's just a political undercurrent here to maintain there's some conspiracy and irrevocable bond between the US and Cisco as leading infrastructure hardware manufacturer, that seems to be the focus of the article. Hence, propaganda.
Python had an easy to use CLI and supportive docs that made it easy to throw things together.
Case-in-point, the language literally includes functions like:
dir
That facilitate treating library objects as simple directories of functionality.Default arithmetic library supports APA.
There's just a lot of stuff that's frictionless in Python, where Ruby has sprawling and unintuitive syntax, and mandatory programming constructs.