soraki_soladead

6 ms·

31.
by soraki_soladead 4y ago
Quantity of datasets doesn’t seem like the right metric. The library just needs the datasets you care about and both libraries have the popular ones. What’s more important is integration and if you’re training custom TF models then tfds wil
32.
by soraki_soladead 4y ago
According to this the SOTA program requires the income to make rental payments beyond the one year: https://www.nyc.gov/site/hra/help/sota.page “SOTA is only provided to households whom DSS has determined wil
33.
by soraki_soladead 4y ago
As mentioned, many of these are very common in other large cities (EDIT: fine, in the US). I'm going to call out this one because its frequently mentioned and seems harder to tackle than the other problems: > Mentally ill people in
34.
by soraki_soladead 4y ago
Second hand may not have been the best phrasing on my part, I admit. What I mean is that the model only has textual knowledge in its dataset to infer what “basketball” means. It’s never seen/heard a game, even if through someone else’s
35.
by soraki_soladead 4y ago
I didn’t read it as being a religious take. They appear to be referring more to embodiment (edit: alternatively, online/continual learning) which these models do not posses. When we start persisting recurrent states beyond the current
36.
by soraki_soladead 4y ago
> Some theorists even predict another 9/11 since that event transpired after the last Trillion dollar accounting error by the Pentagon. Dropping “conspiracy” from “conspiracy theorists” is a nice touch but it doesn’t lend much to su
37.
by soraki_soladead 4y ago
Is sci-hub still adding papers? I thought it was stalled because of a court case in India. Regardless, if its on arxiv why use sci-hub?
38.
by soraki_soladead 4y ago
That’s correct. As I understand it, it only becomes a problem when Tesla’s board says it is.
39.
by soraki_soladead 4y ago
I’m not convinced there’s a difference.
40.
by soraki_soladead 4y ago
> Anyway, just my personal opinion. It's become a very 'us vs them', lines-in-the-sand argument these days, and it'd be great if the conversation could be less heated and more philosophical. It's heated and less
41.
by soraki_soladead 4y ago
I agree. I truly miss the graph-mode API, especially coming from Theano before TF, but it wasn’t as beginner friendly and Google wanted to capture market share for their cloud. At least with jax the core library isn’t adopting any of the fr
42.
by soraki_soladead 4y ago
Nope. It’s a different namespace and the numpy methods just wrap existing TF methods: https://github.com/tensorflow/tensorflow/blob/v2.10.0/tensor...
43.
by soraki_soladead 4y ago
> I think it's relatively uncommon to go back and compare specifically what's happening in the brain with a given ML architecture. Less common but not unheard of. Here's one example, primarily on focused on vision: http:&
44.
by soraki_soladead 4y ago
You don't refute my point. In fact, you strengthen it by providing no evidence why this should be a requirement of modern logos for software companies. You list a bunch of things a logo should be useable for in your mind, otherwise its
45.
by soraki_soladead 4y ago
> it won't look good in black & white That sounds like a concern that stopped being relevant for many software companies a decade ago at least. These days app icons and hero images are more important than whether you can fax or
46.
by soraki_soladead 4y ago
Many more people can put together a manual of step-by-step pictures than can program.
47.
by soraki_soladead 4y ago
I’m referring to post-training pruning not smaller models. This is already well-studied but it’s not as useful as it could be on current hardware. (Deep learning currently works better with the extra parameters at training time). Retrieval
48.
by soraki_soladead 4y ago
> GPT-3 already uses about twice as many parameters This isn't accurate. GPT-3 has 175B parameters. The human brain has ~175B cells (neurons, glia, etc.) The analog to GPT-3's parameter count would be synapses, not neurons, whe
49.
by soraki_soladead 4y ago
True but new species will emerge. There have be numerous mass extinction events on Earth.
50.
by soraki_soladead 4y ago
If the model is learning the nature of code and synthesizing new code then it may not violate some licenses. When people do this its usually fine. However, if the model is simply compressing code it sees into the model weights, just memoriz
51.
by soraki_soladead 4y ago
That isn't what post-scarcity means. It doesn't mean there is zero scarcity. It means basic needs are easily met: https://en.wikipedia.org/wiki/Post-scarcity_economy You're making claims about how space
52.
by soraki_soladead 4y ago
One concrete disadvantage of _extreme_ inequality, specifically in a democracy like the US, is the erosion of your rights when they are inconvenient for those with more wealth/soft-power than you. When wealth inequality is significant,
53.
by soraki_soladead 4y ago
My interpretation is that it's fun to use so they use it a lot but not altogether useful (eta: and/or necessary): they didn't miss it on work projects.
54.
by soraki_soladead 4y ago
That seems like a distinction without a difference.
55.
by soraki_soladead 4y ago
That assumes that people left to their own free will and time would end up in a dismal state as the default. I’m not sure. I’d love to see a proper analysis but there’s lots of evidence both ways. Some people with excess leisure would end u
56.
by soraki_soladead 4y ago
If someone says, "Cars may be great at transportation to and from work but we should be cautious to use them as a counterbalance for a trebuchet." I don't think it's disingenuous to take issue with the second part of the
57.
by soraki_soladead 4y ago
I completely agree. However, here are some things a more useful study would cover: - Describe the task creation process, why they are appropriate for measuring X, etc. Preferably drawing from similar studies in people instead of ad hoc. Lot
58.
by soraki_soladead 4y ago
Many of the perceived failures of DALL-E 2 are subjective and ambiguous (by the author's own interpretation in some cases!) and speak to the author's negative biases which are well known outside of this publication. To be clear, I
59.
by soraki_soladead 4y ago
Gary Marcus here again to throw cold water on anything deep learning. The negative tone when not succeeding in tasks it wasn't designed or advertised for continues to be disappointing. > To the extent that the goal is to develop art
60.
by soraki_soladead 4y ago
There is a TOC with the PDF and Chrome's PDF reader shows it in the sidebar. Other PDF readers should do the same.

More ›