|
Winkle-Daddy posted:oh look, the handy work of Summly, the "AI News Summarizing Product" built by some kids that Marissa gave like a billion dollars to. Working at yahoo under marissa was the fuckin worst. "guys, guys, don't you see how valuable tumblr is???" if by “built by some kids” you mean farmed out building an application around some algorithms licensed from SRI
|
# ? Sep 21, 2019 02:00 |
|
|
# ? May 27, 2024 03:54 |
|
Lysidas posted:just get a phd from CMU's ML department
|
# ? Sep 21, 2019 02:29 |
|
|
# ? Sep 21, 2019 05:01 |
|
https://twitter.com/_julesh_/status/1177533459843207168
|
# ? Sep 28, 2019 03:43 |
|
you think open sores has issues thanks to a culture of tribal knowledge and gatekeeping, then you discover people’s trash ML repos that are horribly broken out of the box
|
# ? Sep 28, 2019 05:58 |
|
yeah sure, a statistical model of the solar system is a pretty good predictor for eclipses. thats kindof what we've had since millennia (more or less successfully) idgi, whats his point?
|
# ? Sep 28, 2019 08:48 |
|
Krankenstyle posted:yeah sure, a statistical model of the solar system is a pretty good predictor for eclipses. thats kindof what we've had since millennia (more or less successfully) it is incredibly dumb since it is literally a curve fitting task for which machine learning is both overkill and pretty bad.
|
# ? Sep 28, 2019 11:05 |
|
Yes but curve fitting needs math while feeding data into a neural network only needs a git clone
|
# ? Sep 28, 2019 15:40 |
|
The jig is up for this hip hop indian brogrammer YouTuber charged loads of fans $199 for shoddy machine-learning course that copy-pasted other people's GitHub code https://www.theregister.co.uk/2019/09/27/youtube_ai_star/
|
# ? Sep 29, 2019 03:08 |
|
quote:Ray Phan, a senior computer-vision and deep-learning engineer at Hover, a 3D software startup based in San Francisco, told The Register that the lectures were confusing and contained mistakes. The online projects were too simplistic to be useful in the real world, and Raval was absent and unsupportive. Ah yes the senior AI engineer that needs to sign up for an intro to AI course
|
# ? Sep 29, 2019 05:36 |
|
the other day I was doing recatchpas for google and giving wrong answers as a form of protest against soulless automation guess what the robot on googles end decided: that I wasn’t being human enough and that me clicking on pictures with my mouse was generating unusual network traffic that the mighty google needed to be protected from
|
# ? Sep 29, 2019 06:50 |
|
Max Facetime posted:the other day I was doing recatchpas for google and giving wrong answers as a form of protest against soulless automation Wish the rest of us could be protected from your unusual network traffic too, OP.
|
# ? Sep 29, 2019 16:44 |
|
lancemantis posted:Ah yes the senior AI engineer that needs to sign up for an intro to AI course if it's the ray phan i'm thinking of, he was probably interviewed as an expert in the field, not one of the course's students since the course copied stuff from github, some of phan's code might even be in it
|
# ? Sep 29, 2019 17:29 |
|
DONT THREAD ON ME posted:i'd only do it if there were a way to prevent people from using it to deep fake other people. Deep fake the placard photo
|
# ? Sep 29, 2019 19:44 |
|
Is there an AI that analyzes pics and goes “beautiful/ugly”
|
# ? Sep 29, 2019 19:47 |
|
big scary monsters posted:have there been examples of adversarial approaches that limit access to the network they're trying to spoof? presumably a real adversary isn't going to hand you their model and give you a week's unrestricted cluster time running them against each other - what's the minimum attempts you can use to turn a gun into a turtle? there's a bunch of "black box attacks" that work this way, only seeing inputs and outputs often you don't even need those, though, because adversarial attacks usually transfer between networks. so you can just train your own model, and as long as it's even vaguely similar to whatever you're attacking, images that trick it will probably also trick your target
|
# ? Sep 29, 2019 22:53 |
|
DaTroof posted:if it's the ray phan i'm thinking of, he was probably interviewed as an expert in the field, not one of the course's students yeah ray phan at hover is a phd in computer vision but also no one should be a phd in computer vision, the field should be forgotten, the books burned and the earth where you bury the ashes should be salted.
|
# ? Sep 30, 2019 00:40 |
|
animist posted:there's a bunch of "black box attacks" that work this way, only seeing inputs and outputs it also helps that transfer learning is common in vision so you can expect whatever you’re dealing with to be some fine tuned head on an imagenet base
|
# ? Sep 30, 2019 03:43 |
|
Schadenboner posted:Wish the rest of us could be protected from your unusual network traffic too, OP. wow, how rude these are the bugs that I discovered while doing my protest: • after a while google started showing adversarially altered images, which is blatantly ableist against people with bad eyesight • it can get stuck in a “check also the newly appeared images” but all images are blank state • the unusual traffic decision occurs before checking if the solution was correct has nobody tested this poo poo??
|
# ? Oct 1, 2019 16:41 |
|
lancemantis posted:it also helps that transfer learning is common in vision so you can expect whatever you’re dealing with to be some fine tuned head on an imagenet base fun fact: there are more than 120 breeds of dog in imagenet, out of 1000 categories total. they're wayy overrepresented compared to everything else in there. so, over the past 10 years, humanity has collectively spent billions of dollars in R&D on... building the perfect dog show judge.
|
# ? Oct 1, 2019 17:04 |
|
The thing that grinds my gears is how they don't specify what percentage of a square needs to have a traffic light in it before I make a click/no-click decision on teh box. I mean, is "Click all boxes where >=25/33/50/67/75% of the box's area is filled by a traffic light (or sub-assembly thereof)" really so loving hard? E: Also poor documentation on whether or not pedestrian crossing lights (not signs) attached to traffic lights are to be considered as traffic lights for the purposes of the above.
|
# ? Oct 1, 2019 17:15 |
|
Cybernetic Vermin posted:https://arxiv.org/abs/1608.08225v4 , which does the very basic needful demonstrating that e.g. encoding multiplication requires exponentially wide hidden layers, which should reasonably be interpreted as "cannot be learned with current methods". animist posted:okay, so, first thing you gotta understand is what people mean by "deep neural networks".
|
# ? Oct 1, 2019 17:18 |
|
help?? why’s my deep neural network only multiplying correctly some of the time?! here’s my network: C++ code:
|
# ? Oct 1, 2019 17:25 |
|
|
# ? Oct 1, 2019 17:32 |
|
if you're feeling fancy,
|
# ? Oct 1, 2019 17:33 |
|
gently caress me for recognizing this
|
# ? Oct 1, 2019 18:40 |
|
drat i wish it was possible to teach computers to do multiplications
|
# ? Oct 1, 2019 19:06 |
|
Pinterest Mom posted:if you're feeling fancy, is this the calculations for damage in Path of Exile
|
# ? Oct 1, 2019 19:09 |
|
Question: was looking for a thing that was going around not too long ago. It was like a neural network of morphing faces (not sure if it was that or a straight render). Looked like the wall of faces from doom. Anyone know what I’m talking about? Driving me crazy I can’t find it
|
# ? Oct 1, 2019 20:57 |
|
multiplying two inputs with *each other* without using some specialized blocks is the hard bit. multiplying inputs *by a constant* and adding them up with some clamping function applied is the basic building block.
|
# ? Oct 1, 2019 21:52 |
|
Alan Smithee posted:Question: was looking for a thing that was going around not too long ago. It was like a neural network of morphing faces (not sure if it was that or a straight render). Looked like the wall of faces from doom. Anyone know what I’m talking about? Driving me crazy I can’t find it this? https://www.youtube.com/watch?v=XOxxPcy5Gr4
|
# ? Oct 1, 2019 22:50 |
|
I don’t believe so. From what I remember it was like a big puddle of faces, almost like this queen album cover With the faces changing of course And the doom face wall I was talking about Alan Smithee fucked around with this message at 23:12 on Oct 1, 2019 |
# ? Oct 1, 2019 23:09 |
|
Captain Foo posted:is this the calculations for damage in Path of Exile no it's Loss
|
# ? Oct 1, 2019 23:23 |
|
Alan Smithee posted:I don’t believe so. From what I remember it was like a big puddle of faces, almost like this queen album cover oh i found it
|
# ? Oct 1, 2019 23:24 |
|
More like Human unlearning
|
# ? Oct 1, 2019 23:32 |
|
Cybernetic Vermin posted:multiplying two inputs with *each other* without using some specialized blocks is the hard bit. multiplying inputs *by a constant* and adding them up with some clamping function applied is the basic building block. well these legos suck!!! *stomps around in a circle angrily*
|
# ? Oct 2, 2019 04:13 |
|
https://twitter.com/WIRED/status/1181437300414275584
|
# ? Oct 10, 2019 00:38 |
|
i.e. a state of the art for 1996 reimplementation of pagerank
|
# ? Oct 10, 2019 10:15 |
|
using something called "random sampling", a concept from quantum chromodynamics,
|
# ? Oct 10, 2019 15:58 |
|
|
# ? May 27, 2024 03:54 |
|
Algorithm uses math. Quantum mechanics uses math. Cool!
|
# ? Oct 11, 2019 04:11 |