make3

40 ms·

811.
by make3 5y ago
their moat is that their app is now everywhere, which is worth something. So many other services have easy to copy apps, but it takes years for other apps to beat them. See Facebook, instagram, etc. A good app is really worth something.
812.
by make3 5y ago
I work in this domain, and you can make these things say anything with a little probing, even stuff like "Hitler was right to kill all the Jews, I wish he was still alive today." They likely don't want to have "OpenAI GP
813.
by make3 5y ago
that sucks.
814.
by make3 5y ago
do you have sources for this, it's really not obvious for me, sounds a bit like a very confidently said hot take a little bit (without a source)
815.
by make3 5y ago
recharging a crime phone in front of a camera lol
816.
by make3 5y ago
that's why they're pre-cached in theory, so they don't have to know. if you're saying you can't trust the closed source implementation, then yeah
817.
by make3 5y ago
do you know how face id works? or this is just a.. gut feeling
818.
by make3 5y ago
strange, so they're not using the same voice model depending on the page. This model sounds more "mechanical", even though it's definitely better here
819.
by make3 5y ago
why is mktemp insecure?
820.
by make3 5y ago
not pytorch anymore. there's support now
821.
by make3 5y ago
74% median decoding accuracy. Up-to 93% best performance. Is 74% decoding accuracy good in a sentence of a very small length? do we know if the misses are at least semantically close or random? It's very weird and suspicious to me that
822.
by make3 5y ago
Your message implies that you trust that Google will use this message in a clean way. They won't; they have a history of using algorithms to detect TOS abuses in a very Gung-ho way without any sort of functional appeal process.
823.
by make3 5y ago
Wtf you're coworkers. The relationship will always be, at its core, transactional. People should be nice, but nice doesn't mean the relationship isn't transactional. Someone nice can still fire you at a moment's notice.
824.
by make3 5y ago
I feel like students are already aware of this. The question for me in a country that almost (90%) fully funds students going to college, is about whether we fund them as much as students going for degrees that will lead to jobs.
825.
by make3 5y ago
markov chains model n grams.. this is really much better
826.
by make3 5y ago
Well, my understanding is that your point is that it's not useful and only regurgitates code it already saw. That's super false, it's useful tool that generates code in a very context sensitive manner. Not always perfect, and
827.
by make3 5y ago
It can, that does not mean that it will, in any case other than people actively probing it for that.
828.
by make3 5y ago
this is the worse take I've seen so far
829.
by make3 5y ago
Controlling the generation for code quality will be extremely hard. The only thing I see is that they could filter their dataset so that some bad proxy of code quality is taken into account, something like the number of stars (which is clea
830.
by make3 5y ago
Are you involved in the project ? Can I ask what your source is? Great if it's the case.
831.
by make3 5y ago
Are you saying you read that it didn't have a hidden layer? My point is that having such a huge dataset would not be extremely useful without using a deep neural net (of at least one hidden layer)
832.
by make3 5y ago
They say that the dataset is hundreds of Gigs worths of games, so the net must still be really pretty big. Though definitely not directly comparable, dataset of GPT2-xl is 8 million web-pages. What I mean to say is that this is clearly dee
833.
by make3 5y ago
but.. they're using a pretty big neural net themselves (NNEU) as far as I can tell? With datasets of hundreds of Gigs. Doesn't this remove the significance of the matchup, that was supposed to be about Deep Learning vs more tradi
834.
by make3 5y ago
He still shows up as the image of a serial killer to me
835.
by make3 5y ago
this comment is a knee jerk reaction, it's pretty obvious why it's appropriate, please make more of an effort (spider silk is associated to strenght, regular silk is weak & for appearance)
836.
by make3 5y ago
isn't that very close to just rebranding a phone that already exists though
837.
by make3 5y ago
Take a graphics class online. The great thing is that learning this has a very fast reward loop. With every small thing you learn, you can try it and it looks great right away. So: Probably take a graphics class, with an intro to shaders an
838.
by make3 5y ago
The last point about making a phone completely from scratch was a negative one for me, the idea of making such a complicated thing completely from scratch felt like it was bound to be really bad and full of bugs compared to other companies
839.
by make3 5y ago
What I meant by "forced to learn something useful" is what you put in a more clear way by being forced to generalize.
840.
by make3 5y ago
"You do realize we can train a neural network to perform this task" I'm doing a master's in deep learning for NLP and I'm not sure we can. Language modelling can't do this because grammatical yet semantically i

More ›