mrbungie

27 ms·

481.
by mrbungie 4y ago
I've seen some really f**ed up text formatted files in the wild. In those cases I really see value in using something more "intelligent" (and declarative rather than imperative) rather than coding a non-trivial number of [pyt
482.
by mrbungie 4y ago
Or more probably it can be a combination of both (and possibly more) explanations. There is no way to know, all we can do is speculate. But to think that this is just chaos theory is as naive as to think these behaviours are fully planned.
483.
by mrbungie 4y ago
Of course you should trust the results and publish them anyway (or so should happen in an ideal world ;) ), but they may be disregarded as uninteresting or unworthy in their domain when reviewed. P-value and effect sizes by themselves are n
484.
by mrbungie 4y ago
A "dichotomy" that is explained by big bucks is not a dichotomy but mere business strategy from AWS's part.
485.
by mrbungie 4y ago
That would be even more intelligent than the majority (or even all, YMMV) of humans.
486.
by mrbungie 4y ago
As far as I've experienced, they're still quieter than a ICE car. Not sure about what's the real decibel difference or if it depends on legislation (my experience is mostly in Japan).
487.
by mrbungie 4y ago
Heavier and quieter. Recipe for disaster.
488.
by mrbungie 4y ago
Just look for the words "China IP law infrigement" in Google FFS, if that is trolling, well, then I'm trolling. About watermarks, I'm obviously talking about AI generated content watermarks, not general tiktok watermarks
489.
by mrbungie 4y ago
Probably it will be more like espionage and ensuring at least constant access to the litography equipment that China is lacking right now.
490.
by mrbungie 4y ago
Nope, I haven't and I also didn't mention or suggest any actual arms race. My point is that any policy regarding AI (or any policy for that matter) is going to be hard to enforce when including potential (economically) "rogue
491.
by mrbungie 4y ago
That is obviously going to happen in mainland China specially when it may affect the CCP's plans in some way (i.e. even "morally" in the case of Xi Jinping deepfakes). Are they going to enforce it for products/services s
492.
by mrbungie 4y ago
Good luck blocking such a big market. Money simply speaks louder.
493.
by mrbungie 4y ago
Prediction: Some nation-state that cares nothing about IP (i.e. China) wins the AI race in this domain making these regulations totally useless.
494.
by mrbungie 4y ago
Sabotage and propaganda are examples of making war without direct combat. I've always interpreted that as what Sun Tzu was referring to. I'm also sure that said position is not absolute, but rather just the idea that one should no
495.
by mrbungie 4y ago
Yeah sure, that's what should happen. I'm not saying otherwise, read the comment again.
496.
by mrbungie 4y ago
I think it's a good design decision that just lacks control during the app review process. There are apps that need full access to your photo gallery to be really useful (i.e. where limited pool of photos may have little sense in those
497.
by mrbungie 4y ago
Having diagrams (think free body diagrams in static mechanics, or a T-s diagram in thermodynamics) make a lot of non-trivial problems a lot simpler to communicate. And correctly understanding an unambiguous definition of a problem is a majo
498.
by mrbungie 4y ago
This is pretty pessimistic. I don't know what kind of expectations you have about LLMs. Less than 10 years have passed since the original Transformers paper and we're seeing tangible and useful software out of it. I've seen f
499.
by mrbungie 4y ago
I think that's totally going to be a problem in X years with devices like Neuralink. Maybe such devices will be expected to bear some kind of sign to indicate usage and to use them when interacting with others in social gatherings is
500.
by mrbungie 4y ago
Because I don't think that the model learned the literal memorization of chess moves. It must've at least compressed said information in some way way. And since the model is not biased to play chess on its structure nor sampling p
501.
by mrbungie 4y ago
Yet it achieves 1400. Add hard rules to stop it spewing out said moves and you have a 1400 ELO Player (most UIs won't even let you make illegal moves). It is difficult to say that is not impressive due to it being an emergent ability.
502.
by mrbungie 4y ago
I would argue that a decent human player also has "similar sequences of moves in its training set", due to reading plus playing or watching games. The illegal move argument is good though, and indicates no direct understanding of
503.
by mrbungie 4y ago
They announced a very similar set of features for Google Workspace 2 days ago: https://workspace.google.com/blog/product-announcements/gene...
504.
by mrbungie 4y ago
Focusing on individual product growth is just one strategy an organization may follow. And it's a strategy that if followed blindly will eventually make said organization be seen as unreliable to customers. I wouldn't pay Google f
505.
by mrbungie 4y ago
Yeah, sure. I'm not saying quantitative methods are not useful. Just that checking the field's SOTA and theory is an important first step that facilitates hypothesis making, even if those hypothesis are an alternative to what know
506.
by mrbungie 4y ago
Some (or maybe a lot depending on your environment and domain) data "scientists" focus more on analyzing/processing data than reviewing the state-of-the-art and main ideas/discussions about the domain in question. Quite
507.
by mrbungie 4y ago
Revolutions are never smooth and privileged groups threatened by new technologies or ideas always have reasons to push back and will do so fiercely, that's plain human nature. But if we block progress to protect the status quo of said
508.
by mrbungie 4y ago
Basically a neo-luddite kind of argument.
509.
by mrbungie 4y ago
Illegal != unavailable. Making them illegal is just going to restrict who will be able to train/harness them, and that just means controlling what people use compute for. The only way I see that happening is setting restrictions for co
510.
by mrbungie 4y ago
This. Maybe Search is not in MS's strenght, but damn me if they don't know how to play these kind of corporate (war)games.

More ›