AI Rewind: A Year of Amazing Machine Learning Papers

This video shows a curated list of the latest breakthroughs in AI by release date with a clear video explanation, link to a more in-depth article, and code. The videos include great commentary on the future of AI by many important people in the field such as Fei-Fei Li, Luis Lamb, Gary Marcus, and more.

Enjoy the video!

Chapters:

Hey! Tap the Thumbs Up button and Subscribe. You'll learn a lot of cool stuff in 2021, I promise. 2020, A year in review Where do you want AI to go?

References:

• Complete article:
• GitHub repo:

Follow me for more AI content:

• Instagram: 
• LinkedIn: 
• Twitter: 
• Facebook: 
• Medium: 
• Youtube: 

Video Transcript

00:00where do you want ai to go what what00:02would make you happy00:03what is the objective here the objective00:05function for those of us who are00:06building ai00:08um or trying to inform it from other00:10fields00:11what would count in success where do you00:12want to take it to00:17[Music]00:50[Music]00:57[Music]01:18[Music]01:43[Music]01:54[Music]02:39[Music]03:24[Music]03:30[Music]03:51[Music]04:16[Music]04:29[Music]04:50but it's a trick no no it's much more04:52than that uh whosoever be he were they04:54shall have the power04:57whatever man it's a trick please be my05:00guest05:02come on really yeah oh this is gonna be05:05beautiful clint you've had a tough week05:07we won't hold it against you if you05:08can't get it up you may have seen this05:10before05:10right yeah all right05:39[Music]06:24[Music]06:30[Music]06:51[Music]07:16[Music]07:27[Music]08:13[Music]08:17researchers and artists around the world08:19have been using style game 2 to do all08:20kinds of interesting work08:22at nvidia we've been pushing the08:23frontier of using gans to synthesize08:25extremely realistic images08:28the catch is that you need to train them08:30on extremely large quantities of data08:32so to tackle this challenge we've08:33invented a new approach for stylegan 208:35that we call adaptive discriminator08:37augmentation or ada08:38this lets us reduce the number of08:40training images by a factor of 10 a08:42factor of 20 or more while still getting08:44great results08:56[Music]09:06as a scientist i want to09:10push the scientific knowledge and09:12principles of09:13ai further and further and the first09:15thing is start with09:17some really fundamental laws and09:21principles of09:22ai i keep saying that i still feel our09:24ai09:25era is pre-newtonian physics that we are09:28still studying phenomenology09:30and engineering but there is going to be09:34a moment09:34or a set of moments where we're starting09:37to understand09:38the principles of intelligence and that09:41is09:42the scientists in me we're in the head09:45of a citizen i guess09:47and directing stanford's human center ai09:51institute09:52i want this to be a technology that can09:55an idealistic way really better human09:58conditions09:58right it's so profound it's so10:02horizontal10:03it's so um it has so much human and10:07societal impact10:10it can be very very bad and can be very10:12very good10:13so there i would like to see a10:16framework of this technology being10:19developed and deployed10:21in the most benevolent way10:25what comes to mind very often so i hang10:27out because of the family i'm born into10:30with people who work at state10:32departments who deal with international10:34laws military laws etc10:36and there's a there's a there's a big10:38concern here with militarization of ai10:40right10:41danny just mentioned human decision10:43making humans in the loop why people are10:45trying to negotiate10:46treaties limiting should we limit should10:49they always be a human10:50in the loop when we when we build10:52dangerous machines10:54you know kill a robot as well as is10:56happening as we all know10:58in the defense world and and as will11:01happen so what about the growing threats11:03of militarization of ai which is ongoing11:06and what should we11:07uh should we do anything about it or11:10pessimistic can we do anything about it11:12or is that11:15genie out of the box i want the costs11:18and the benefits of ai to be11:21evenly distributed across society both11:24in the united states11:26and globally and i want the public to11:29trust that that's what's being done11:31and i think that's impossible without11:34changes to law11:36just impossible um i like one of the11:38sayings of michael raving11:40from harvard and hebrew university said11:42that it's great that11:44uh computer science has not been around11:46for 2000 years11:48and we are at a stage where very very11:50important results11:51occur in front of our eyes i also like a11:54saying by11:56alan turing in his mind paper where he11:58said we can only see a short distance12:01ahead12:02but we can see plenty there that needs12:04to be done12:05and as fei-fei said as scientists i want12:08ai to advance but since ai is having an12:11impact12:12as big as physics had in the 20th12:15century12:15engineering heading and 20th century all12:18the ethical issues12:19all the biases and all the social12:21implications that margaret and others12:23and here in francesca12:25have been studying they are key as12:28responsibility of our12:29ourselves as scientists and in terms of12:32uh12:32technical technically speaking what we12:35are trying to do12:36is to conciliate traditions of ai so12:39that we cannot see12:41and as you said gary before you've been12:43saying since the12:44late 90s we have to converge we have to12:47look for convergency12:49you cited the only colonial cited other12:51prominent scientists today12:53we need a way of seeing that several12:56techniques can contribute to this12:57endeavor of making ai12:59fairer ai less bias and ai13:03to make something very positive for us13:06as humanity we need uh as scientists to13:09see13:09our fields in a very uh human humanistic13:13way13:14so that not only the technical stuff13:16advances but we also have to13:18be guided by serious and by effective13:20ethical principles13:22laws and norms as ryan said it's hard to13:25do to do that at the moment we are at13:27the beginning of uh13:29an ai cambrian explosion as several13:32people mention here but we need to be13:34very aware13:35of the social ethical and global13:38implications13:39that ai has these days we have to be13:41concerned about the north south13:44divide about the different cultures in13:46order to regulate it properly13:49we cannot see it only from a single13:52cultural perspective so13:53that's what i want to see ai researchers13:56doing to be they have to be concerned13:58about the technical results14:00the outstanding results ai has been14:02showing but also14:03we have to be to care about other people14:06about14:06other people's other countries and over14:09and over all14:10for the global health of the planet14:12thank you very much14:13and merry christmas happy new year to14:16everyone thank you gary and vincent for14:18the14:18brilliant debate you brought today and14:20for the scientists14:22i am reminded of the old african proverb14:24that i'm sure you all know which is it14:25takes a village to raise a child14:27clearly it will take a village as we've14:29seen today14:30to raise an ai that is ethical robust14:32and trustworthy