The Future of Artificial Intelligence: Crash Course AI #20 - By CrashCourse
Transcript
00:0-1 | Hi everyone . I'm Jibril and welcome to the final | |
00:02 | episode of Crash course Ai we've covered a lot of | |
00:06 | ground together from the basics of neural networks to game | |
00:09 | playing language modelling in algorithmic bias . We've even experimented | |
00:13 | with code in labs and as we've been learning about | |
00:16 | different parts of artificial intelligence as a field . There | |
00:19 | are a couple of themes that keep coming up . | |
00:21 | 1st . Ai is in more places than ever before | |
00:24 | . The machine learning professor Andrew ng says that artificial | |
00:27 | intelligence is the new electricity . This is a pretty | |
00:31 | bold claim but lots of governments are taking it seriously | |
00:34 | and playing to grow education , research and development in | |
00:37 | Ai china's plan alone calls for over 100 billion U | |
00:40 | . S . Dollars in funding over the next 10 | |
00:43 | years . Second AI is awesome . It can help | |
00:46 | make our lives easier and sort of gives us superpowers | |
00:49 | , who knows what we can accomplish with the help | |
00:51 | of machine learning and Ai . And third Ai doesn't | |
00:55 | work that well yet . I still can't ask my | |
00:58 | phone or any smart device to do much and we're | |
01:00 | far away from personal robot butlers . So what's next | |
01:05 | ? What's the future of Ai ? Yeah . Mhm | |
01:15 | . One way to think about the future of AI | |
01:18 | is to consider milestones AI hasn't reached yet . Current | |
01:21 | soccer robots aren't quite ready to take on human professionals | |
01:25 | and series still has a lot of trouble understanding exactly | |
01:28 | what I'm saying for every AI system , we can | |
01:31 | try and list what abilities would take the current technology | |
01:34 | to the next level . In 2014 , for example | |
01:37 | , the Society of automotive Engineers attempt to do just | |
01:40 | that for self driving cars there to find five levels | |
01:43 | of automation for each additional level . They expected that | |
01:46 | the Ai controlling the car can do more without human | |
01:49 | help . At level one , cruise control automatically accelerates | |
01:52 | and decelerates to keep the car at a constant speed | |
01:56 | . But everything else is on the human driver at | |
01:58 | level three , the car is basically on its own | |
02:01 | , it's driving , monitoring its surroundings , navigating and | |
02:04 | so on . But a human driver will need to | |
02:06 | take over is something goes really wrong , like really | |
02:09 | bad weather or a downed power line . And at | |
02:12 | level five the human driver can just sit back , | |
02:15 | have a smoothie and watch crash course ai while the | |
02:17 | car takes him to work through rush hour traffic . | |
02:20 | And obviously we don't have cars with the technology to | |
02:23 | do all this yet . But these levels are a | |
02:25 | way to evaluate how far we've come and how far | |
02:28 | research still has to go . We can even think | |
02:30 | about other AI is using levels of automation , like | |
02:33 | for example , maybe we have level one Ai assistants | |
02:36 | right now that can set alarms for us , but | |
02:38 | we still need to double check their work . But | |
02:41 | what are levels two through five , what milestones would | |
02:44 | need to be achieved for an Ai to be as | |
02:46 | good as a human assistant ? What will be milestones | |
02:49 | for computer vision or recommend our systems or any topics | |
02:52 | discussed in this course , We'd love to read your | |
02:54 | ideas in the comments . Sometimes it's useful to think | |
02:57 | about different kinds of ai on their own as we | |
03:00 | make progress on each very difficult problem . But sometimes | |
03:03 | people try and imagine an ultimate Ai for all applications | |
03:07 | and artificial general intelligence or A . G . I | |
03:11 | understand why there's such an emphasis on being general . | |
03:14 | It can be helpful to remember where all this Ai | |
03:17 | stuff first started and for that let's go to the | |
03:20 | thought bubble . Alan turing was a british mathematician who | |
03:23 | helped break german enigma codes during World War two and | |
03:26 | helped define the mathematical theory behind computers in his paper | |
03:29 | computing machinery and intelligence . From 1950 he introduced the | |
03:33 | now famous turing test or the imitation game turing , | |
03:37 | proposed an adaptation of a guessing game . In his | |
03:39 | version there's an interrogator in one room and a human | |
03:43 | and a machine in the other . The interrogator talks | |
03:46 | to the hidden players and tries to figure out which | |
03:48 | is human and which is a machine turing . Even | |
03:51 | gave a series of talking points like please write me | |
03:53 | a sonnet on the subject of the forth bridge at | |
03:56 | 34,957 and 70,764 . Do you play chess ? I | |
04:03 | have K . At K . One and no other | |
04:05 | pieces . You only have K . At K . | |
04:07 | Six . In our at our one it's your move | |
04:11 | . What do you play ? The goal of the | |
04:13 | imitation game ? Was a testing machines . Intelligence about | |
04:16 | any human theme from math to poetry . We wouldn't | |
04:19 | just judge how really robots fake human skin books as | |
04:23 | turning put it . We do not wish to penalize | |
04:25 | the machine for its inability to shine in beauty competitions | |
04:28 | , nor to penalize a man for losing in a | |
04:31 | race against an aero plane . This idea suggests a | |
04:34 | unified goal for Ai , an artificial general intelligence . | |
04:38 | But over the last 70 years AI researchers focus on | |
04:41 | sub fields like computer vision , knowledge representation , economic | |
04:45 | markets , planning and so on . Thanks doc Global | |
04:49 | . And even though we're not sure if an artificial | |
04:51 | general intelligence as possible , many communities are doing interdisciplinary | |
04:55 | research and many air researchers are taking baby steps to | |
04:59 | combine specialized sub fields . This involves projects like teaching | |
05:02 | a robot to understand language or teaching an AI system | |
05:05 | that models the stock market to read the news and | |
05:08 | better understand market fluctuations to be clear . Most of | |
05:11 | Ai is still science fiction . We're nowhere near Blade | |
05:14 | Runner her or any similar movies . Before we get | |
05:18 | too excited about combining everything we've built to achieve A | |
05:21 | . G . I . We should remember that . | |
05:22 | We still don't know how to make specialized A . | |
05:25 | I . S . For most problems . Some sub | |
05:27 | fields are making progress more quickly than others and we're | |
05:30 | seeing A . I . Systems pop up in lots | |
05:31 | of places with awesome potential to understand how A . | |
05:34 | I . Might be able to change our lives . | |
05:36 | Ai professors Yolanda Gil and Bart Selman put together the | |
05:40 | computing research associations AI road map for the next 20 | |
05:43 | years . They predict Ai reducing health care costs , | |
05:46 | personalizing education as celebrating scientific discoveries , helping national defense | |
05:51 | and more part of the reason they expect so much | |
05:54 | progress is that more people than ever , including us | |
05:57 | , are learning how to build A . I . | |
05:58 | Systems . And all these problems have lots of data | |
06:02 | to train new algorithms . It used to be hard | |
06:04 | to collect . Training data , going to libraries to | |
06:06 | copy facts and transcribe books . But now a lot | |
06:10 | of data is already digital . If you wanna know | |
06:12 | what's happening on the other side of the planet , | |
06:14 | you can download newspapers or grab tweets from the twitter | |
06:17 | . Api interested in hyper local weather prediction . You | |
06:21 | can combine free data from a weather service with local | |
06:23 | weather stations to know when the water your plants . | |
06:26 | And if you feed that data into a robot gardener | |
06:29 | , you could build a fully automated weather knowing plant | |
06:32 | growing food making garden make your communities around the globe | |
06:36 | are combining data ai and cheap hardware to create the | |
06:39 | future and personalized Ai technologies . Well imagining an ai | |
06:43 | human utopia is exciting . We need to be realistic | |
06:46 | to and many industries automation doesn't only enhance human activities | |
06:51 | . It can replace humans entirely . Truck delivery and | |
06:54 | tractor drivers are some of the most common jobs in | |
06:57 | the US as of 2014 . If self driving vehicles | |
07:01 | revolutionized transportation in the near future will all those people | |
07:05 | lose their jobs ? We can't know for sure . | |
07:07 | But google . Prizewinning computer science professor Moshe Vardi points | |
07:11 | out that this is already the trend in some industries | |
07:14 | . For example . U . S . Manufacturing output | |
07:16 | will likely keep rising but manufacturing jobs have been decreasing | |
07:21 | a lot plus computers use energy and that means we're | |
07:24 | not getting any benefits from A . I . For | |
07:26 | free massive amounts of machines running . These algorithms can | |
07:29 | have a substantial carbon footprint on top of that . | |
07:32 | As we've discussed , you have to be pretty careful | |
07:35 | when it comes to trusting AI systems because they often | |
07:38 | end up with all kinds of biases you may not | |
07:40 | want . So we have to consider the benefits of | |
07:43 | massive AI deployment with the costs and the now famous | |
07:46 | story from a few years ago Target figured out a | |
07:48 | woman was pregnant based on her shopping history and they | |
07:51 | sent her maternity coupons but she was still in high | |
07:54 | school so her family saw the mill even though she | |
07:57 | hadn't told them do we want our data being used | |
08:00 | like this and potentially revealing personal details or what about | |
08:04 | the government ? Should it be allowed to track people | |
08:06 | with facial recognition , installing cameras at intersections ? When | |
08:10 | we provide companies with location data from our phones , | |
08:12 | we could help them build better traffic models so we | |
08:15 | can get to places faster . Cities could improve bus | |
08:18 | routes . But it also means someone is always watching | |
08:22 | you . Ai could also track your friends and family | |
08:24 | where you shopped eight and who you hung out with | |
08:27 | . If statistics have shown that people who leave home | |
08:29 | late at night are more likely to commit a crime | |
08:31 | and a I know that you left even though it's | |
08:34 | just for late night cookie though . Should it call | |
08:37 | the police to watch you ? Just in case . | |
08:39 | So we can go down any number of scary thought | |
08:42 | experiments . And there's a lot to consider when it | |
08:45 | comes to the future of Ai . AI is a | |
08:47 | really new tool and it's great that so many people | |
08:49 | have access to it . But that also means there | |
08:52 | are very few laws of protections about what they can | |
08:55 | and can't do innovations in a . I have awesome | |
08:58 | potential to make positive changes . But there are also | |
09:01 | plenty of risks , especially if the technology advances faster | |
09:05 | than the average person's understanding of it . It's probably | |
09:07 | the most accurate to say that the future is complicated | |
09:11 | and the most important thing we can do is to | |
09:13 | be educated and involved in A . I as the | |
09:16 | field changes which we're doing right now and crash course | |
09:19 | AI labs , we use some of the same machine | |
09:21 | learning technologies that the biggest companies using their products and | |
09:26 | that the universities rely on for cutting edge research . | |
09:29 | So when we see a company or government rolling out | |
09:31 | a new technology , we know what questions to ask | |
09:35 | , Where do they get their data ? Is this | |
09:37 | even a situation where we want a I to help | |
09:39 | humans ? Is this the right tool to use ? | |
09:42 | What privacy are we giving up for this ? Cool | |
09:45 | . New feature . Is anyone auditing this model ? | |
09:48 | Is this A I really doing what the developers hoped | |
09:51 | it would were also hopefully walking away from crash Course | |
09:53 | AI with some basic tools to build different kinds of | |
09:56 | AI from handwriting recognition to recommend . Our systems were | |
10:00 | excited to see what future you decide to build if | |
10:03 | you want to learn more about AI will include more | |
10:06 | free learning resources . In the description . In the | |
10:08 | meantime I've been making some pretty good progress that john | |
10:11 | green pot . Oh john green but yeah john green | |
10:19 | , but tell the audience what is this pizza ? | |
10:23 | See , not just donuts and bagels anymore . I | |
10:26 | want to thank you all for watching Crash Course AI | |
10:28 | and as they say in john green box hometown , | |
10:31 | Don't forget to be awesome . Crash Course AI is | |
10:40 | produced in association with PBS Digital Studios . If you | |
10:43 | want to help keep crash course free for everyone forever | |
10:45 | , you can join our community on Patreon and if | |
10:48 | you want to keep up to date with my prototyping | |
10:50 | adventures , check out my channel below |
Summarizer
DESCRIPTION:
OVERVIEW:
The Future of Artificial Intelligence: Crash Course AI #20 is a free educational video by CrashCourse.
This page not only allows students and teachers view The Future of Artificial Intelligence: Crash Course AI #20 videos but also find engaging Sample Questions, Apps, Pins, Worksheets, Books related to the following topics.