|
Hello, as many of you know, the conference has been moved to a virtual conference because of the ongoing covid-19 situation. We want everyone to be safe and healthy, even as we discuss these important and exciting topics! Although we cannot meet in person, (I too miss the traditional outing to senor frogs!) we have arraigned an exciting line up of talks and content for everyone to enjoy at home. Slippers and tea are totally optional.  to better help everyone get the most out of this conference, we have organized the content into 5 different tracks:
In addition, we have a number of attendee guided breakout sessions. We still need more people to sign up though! Normally we'd have a sign up board in the main hallways, but since it's virtual, just let one of our organizers know on slack if you'd like to lead a breakout group. So far we have:
looking forward to your (virtual) attendance, and hoping we can learn and share so many ideas together! Registrations are now open!
|
#
?
Jul 6, 2020 18:32
|
|
|
|
| # ? May 4, 2024 08:28 |
|
|
hell yeah. sign up for my talk on my new ML-based classifier “Quicksort”. once you implement this in your code, you can put an ML bullet point in your marketing text. your boss will be impressed.
|
|
#
?
Jul 6, 2020 19:49
|
|
|
"what if i told you i could change the way you think about computers forever" the crowd murmurs suspiciously as i stand on stage with a smug smile slowly the word 'racism' appears on the screen only to be quickly crossed. the crowd explodes into applause and i go on to explain a machine learning system that is extremely racist
|
|
#
?
Jul 6, 2020 20:24
|
|
|
we even used a generative algorithm from your badge picture to create new people *500 headshots of lily white folks slowly scroll past*
|
|
#
?
Jul 6, 2020 21:59
|
|
|
ok so which track does my computational phrenology paper go in?
|
|
#
?
Jul 6, 2020 22:56
|
|
|
This thread hits way too close to home and I am extremely offended
|
|
#
?
Jul 7, 2020 00:44
|
|
|
Deep Dish Fuckfest posted:ok so which track does my computational phrenology paper go in? is the bump map of my tensor matrix hashed to the blockchain?
|
|
#
?
Jul 7, 2020 00:53
|
|
|
mrmcd posted:[*]transportation revolutions - some dweebs will read their tesla fanfic and present their speculative snuff films involving robot cars. The phrase 'trolly problem' will almost certainly make an appearance. As the founder and CEO of a self-driving car start up, I prefer to think of it as a trolly opportunity
|
|
#
?
Jul 7, 2020 01:06
|
|
|
Please come to me talk about human robot interaction where I present yet another “social robot” that is way dumber than Siri and costs $500. Ah dang sorry we shut down the company and the robot is useless now.
|
|
#
?
Jul 7, 2020 01:18
|
|
|
Asleep Style posted:As the founder and CEO of a self-driving car start up, I prefer to think of it as a trolly opportunity 
|
|
#
?
Jul 7, 2020 06:22
|
|
|
https://www.youtube.com/watch?v=nXARrMadTKk
|
|
#
?
Jul 7, 2020 07:57
|
|
|
I for one welcome our new robot overlords
|
|
#
?
Jul 7, 2020 10:24
|
|
|
Lol. I went to undergrad with Julia Galef and my friend was friends with her. Always funny when I'm reminded she's a ai cultist of some note now.
|
|
#
?
Jul 7, 2020 12:20
|
|
|
mrmcd posted:is the bump map of my tensor matrix hashed to the blockchain? yeah i guess we could add a blockchain so that you can add a fresh scan to it every year and if you want to get hired anywhere you need to provide at least 10 years worth of criminally-negative scans or something
|
|
#
?
Jul 7, 2020 13:15
|
|
|
as you can see, our AI has reached nearly human-level understanding of natural images: https://twitter.com/normalcatpics/status/1280166806699663360 we aim to phase out all human employees in favor of the AI by 2022
|
|
#
?
Jul 7, 2020 20:52
|
|
|
animist posted:as you can see, our AI has reached nearly human-level understanding of natural images: otoh i kind of like the ones that are obviously not either curated or transparently an instance of the dataset slightly distorted. the pseudo-psychedelic stuff coming out of ml is also the one thing that continues to convince me that it is a hugely significant development, if perturbing it fucks with your brain it is almost necessarily a model which bears some relationship to how your brain does things.
|
|
#
?
Jul 7, 2020 21:06
|
|
|
in my talk we will consider the companys responsibility to its shareholders and show that it would be immoral to not fire your workers and replace them with a haphazard bag of simulated neurons
|
|
#
?
Jul 8, 2020 02:57
|
|
|
We hold these truths to be self-evident, that all ai are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are racism, bias and the pursuit of profit to the detriment of the world.
|
|
#
?
Jul 8, 2020 08:39
|
|
|
thank you for coming to my talk. slides and code will be available at github afterwards
|
|
#
?
Jul 8, 2020 08:40
|
|
|
hello good to see so many people in chat. today I’m going to be talking about how we trained the wrong model on the wrong data, with hilarious consequences
|
|
#
?
Jul 8, 2020 08:56
|
|
|
Soricidus posted:hello good to see so many people in chat. today I’m going to be talking about how we trained the wrong model on the wrong data, with hilarious consequences did black people get disproportionately get impacted by your model?
|
|
#
?
Jul 8, 2020 09:17
|
|
|
which conference room has the ais with hot evil sexy voices who insult me and my squish meat body?
|
|
#
?
Jul 8, 2020 15:11
|
|
|
mrmcd posted:
if they tell me how to build my own Cherry 2000 then sign me the gently caress up 
|
|
#
?
Jul 8, 2020 19:43
|
|
|
sexbots is a terrible idea, like robots needed yet another reason to resent humanity
|
|
#
?
Jul 9, 2020 07:55
|
|
|
well, no sandwiches for him
|
|
#
?
Jul 9, 2020 08:03
|
|
|
mrmcd posted:
lol
|
|
#
?
Jul 9, 2020 08:06
|
|
|
https://github.com/cmdr2/stable-diffusion-ui
|
|
#
?
Jan 16, 2023 11:16
|
|
|
ive found this website where people are obv training their ai on much less broad sets of images, and the results are so significantly better than regular ai sets, which seem to be able to give broad understanding of anything you prompt it with, but can't do any particular thing well
|
|
#
?
Jan 16, 2023 11:30
|
|
|
i already posted the ai thread mate
|
|
#
?
Jan 16, 2023 11:49
|
|
|
echinopsis posted:ive found this website where people are obv training their ai on much less broad sets of images, and the results are so significantly better than regular ai sets, which seem to be able to give broad understanding of anything you prompt it with, but can't do any particular thing well finally, ai images for my highly specific fetish
|
|
#
?
Jan 16, 2023 19:45
|
|
|
yeah it’s kinda disturbing imo seems like it’s not great at photorealistic stuff, maybe not surprising, but at recreating stuff that looks like digital art? some of those models have been trained on sets of images of one person for example, and if one can accept stylistic rather than photorealistic output, the results are pretty drat good. and yes it’s easy to find flaws but it’s still relatively early days too and just to be clear when I say good I mean convincingly high quality. conceptually it’s quite disturbing, especially the creepiness of using sets of images of individuals, presumably without their permission, to generate these models.
|
|
#
?
Jan 16, 2023 20:40
|
|
|
i like the "do scifi franchise in the style of X" ones. those are pretty fun and seem to work well.
|
|
#
?
Jan 16, 2023 20:46
|
|
|
i’m intrigued to work out how to train up a model, I wanna put everything rodtronics in and see what it spits out.
|
|
#
?
Jan 16, 2023 20:47
|
|
|
Shaggar posted:i like the "do scifi franchise in the style of X" ones. those are pretty fun and seem to work well. yeah, I think aiming for photorealistic enters uncanny valley too easily, and the flaws are glaring, but when it comes to recreating digital art, that problem isn’t there and it does extremely well someone mentioned something once about how maybe with media from 2022 onwards it may be very difficult to discern if AI was used or not and I think that will be increasingly true. a lot of the high quality output, if you took it back in time 5 years, would be considered high quality digital art. the same can’t be said about AI generated art from even 2021. the genie is out of the bottle (christina aguleria)
|
|
#
?
Jan 16, 2023 20:52
|
|
|
 page 219
|
|
#
?
Jan 17, 2023 05:26
|
|
|
*waddles up to the mic, adjusts beard caught in my conference pass lanyard* hm yes this is more of a comment than a question, nice presentation but didn't plan9 offer this back in the '70s??
|
|
#
?
Jan 27, 2023 14:44
|
|
|
lol
|
|
#
?
Jan 31, 2023 10:54
|
|
|
|
| # ? May 4, 2024 08:28 |
|
|
I’m pretty sure leant did this with eurisko
|
|
#
?
Jan 31, 2023 11:30
|
|
