Browse content similar to US Special - Part Three. Check below for episodes and series from the same categories and more!
Line | From | To | |
---|---|---|---|
This is what happens when you let an artificial intelligence get | :00:00. | :00:42. | |
creative. In a future with mass unemployment, young people are | :00:43. | :00:46. | |
forced to sell blood. That's something I can do. You should see | :00:47. | :00:54. | |
the boy and shut up. This is a film whose script was written entirely by | :00:55. | :00:58. | |
an AI chatbot, one that's been trained to mimic science-fiction | :00:59. | :01:05. | |
movie screenplays. Well, I have to go to the skull. I don't know what | :01:06. | :01:13. | |
-- it's a fun watch, but only if you are in on the joke. Yes, it seems | :01:14. | :01:18. | |
writing a gripping or even coherent narrative is beyond AI at the | :01:19. | :01:25. | |
moment. But what if you were to give it the words and let it be the | :01:26. | :01:32. | |
director? What if you gave an AI complete creative control over the | :01:33. | :01:36. | |
costumes, the camera angles? What if you let it chews the cast? What | :01:37. | :01:44. | |
would happen then? -- choose. That was what Sachi and Sachi | :01:45. | :01:48. | |
commissioned this LA -based glad to find out. It is a real experiment. | :01:49. | :01:53. | |
Out on a limb kind of experiment for film. Now, they decided not to make | :01:54. | :01:58. | |
a full-length feature film, but instead to embark on a more | :01:59. | :02:03. | |
manageable project. A pop video, which is something that many new | :02:04. | :02:08. | |
directors cut their teeth on. A French dance group gave them | :02:09. | :02:11. | |
permission to interpret one of their songs. The lyrics would form the | :02:12. | :02:15. | |
script, but the interpretation would be completely down to the machine, | :02:16. | :02:23. | |
more accurately machines. A director's job is a lot of different | :02:24. | :02:26. | |
disciplines and specialties, so given that there is no one beat | :02:27. | :02:30. | |
artificial brain somewhere that can do that, we assembled the best and | :02:31. | :02:37. | |
most progressive pieces of television in a collect it to do | :02:38. | :02:41. | |
that job. And what a collective it was. The -- first up, IBM's Watson | :02:42. | :02:54. | |
was used. The humans could ask the next AI questions about what the | :02:55. | :02:58. | |
video should look like. For that, take to the director's chair. I came | :02:59. | :03:04. | |
up with a series of questions you might ask somebody like me, like, | :03:05. | :03:12. | |
who are the characters? If I said day and night that's non-sequitur | :03:13. | :03:16. | |
and random. But it is quite creative. I can't imagine a creative | :03:17. | :03:24. | |
type coming up with that. She is a Japanese Twitter chat bot that | :03:25. | :03:28. | |
appears to be a 17-year-old Japanese girl and was asked a whole lot of | :03:29. | :03:32. | |
questions about who the characters in the lyrics were, what actions | :03:33. | :03:34. | |
they should perform and what outfits they should wear. Do you have any | :03:35. | :03:39. | |
kind of understanding about how it came up with those answers? Day and | :03:40. | :03:46. | |
night. Night as the main character? From what we understand, this | :03:47. | :03:49. | |
particular AI has been programmed by having conversations with fans, it's | :03:50. | :03:56. | |
a Twitter handle. It has millions of conversations and it is relating in | :03:57. | :04:01. | |
a way that a 17-year-old girl relates on Twitter. Next, casting. | :04:02. | :04:07. | |
How does the director choose the right actor for a role? Well, it is | :04:08. | :04:11. | |
probably someone who feels emotionally connect to the work and | :04:12. | :04:16. | |
can transmit those emotions to the audience, but really it kind of | :04:17. | :04:20. | |
comes down to a gut feeling. That's something that computers don't have. | :04:21. | :04:26. | |
But what computers can do is measure the Performa's brain activity using | :04:27. | :04:32. | |
and EEG, like this. In a very unusual casting session, ten | :04:33. | :04:36. | |
unsuspecting actors were asked to perform the song while their facial | :04:37. | :04:41. | |
expressions and brainwaves were compared to that of the song's lead | :04:42. | :04:46. | |
singer when she performed it. The guy with the best emotional match | :04:47. | :04:52. | |
got the gig. They said, what will you do with that? We said the | :04:53. | :04:56. | |
director needs that to make the joys and whoever we choose has a matching | :04:57. | :05:01. | |
emotional intent to the scene. -- choice. Did they know at this point | :05:02. | :05:08. | |
that the director was and AI? No. When do they find out? Michael does | :05:09. | :05:16. | |
know now. In fact, when the cast him, we brought him down to the set | :05:17. | :05:19. | |
and explained this process. We said it would be a little weird because | :05:20. | :05:22. | |
we will give you directions based on artificial intelligence, as an | :05:23. | :05:26. | |
understanding of what this song is and what your actions are supposed | :05:27. | :05:31. | |
to be. On the day of the shoot even the camera drones followed flight | :05:32. | :05:35. | |
parts that were based on Watson's emotional analysis of the song. And | :05:36. | :05:40. | |
when it came to the editor, choosing the shots and cutting it together, | :05:41. | :05:47. | |
the AI build a brand-new editor, AI. This started by creating random | :05:48. | :05:50. | |
assortment of shots and then learned from feedback on the human sweet | :05:51. | :05:56. | |
shop they liked. And where. After many iterations the finished video | :05:57. | :05:59. | |
was given a visual treatment that was itself a type of neural art, | :06:00. | :06:08. | |
adding a day and night style to the -- character's scenes. How did turn | :06:09. | :06:12. | |
out? Good and bad, really. The result is a pop your that's not | :06:13. | :06:19. | |
going to win any. In fact, the band, whose videos are usually very highly | :06:20. | :06:23. | |
produced, have distanced themselves from the whole thing. And I'm not | :06:24. | :06:27. | |
really convinced that and AI has actually been the director here. -- | :06:28. | :06:33. | |
an AI. Are you? Got a hold different lot of sAI to do different jobs, | :06:34. | :06:39. | |
including asking what this guy should wear. But then the people at | :06:40. | :06:44. | |
the lab I'd really saying otherwise. Are you happy with the results? Yes. | :06:45. | :06:50. | |
The exercise we went through the vibe a lot of great learning. The | :06:51. | :06:55. | |
output itself with the video, it could be better, as a human. But for | :06:56. | :07:03. | |
a first attempt a worthy effort. As the director, behind the scenes I'd | :07:04. | :07:07. | |
like, that super intimidating, because it won't be long before this | :07:08. | :07:10. | |
technology evolves where that editor that we've created becomes | :07:11. | :07:19. | |
absolutely able to interpret like, oh, that shot's running and it has | :07:20. | :07:22. | |
emotion, I should use that. And if that cuts to the concept of this | :07:23. | :07:25. | |
emotional expression it will understand how one crates with | :07:26. | :07:32. | |
another. All of a sudden we will have really emotional cutting from | :07:33. | :07:34. | |
artificial intelligence. It's going to happen. | :07:35. | :07:44. | |
Like most American cities, LA is a driving city, but it can take ages | :07:45. | :07:53. | |
to get across town because of, well, take a look at this. So if you're a | :07:54. | :07:58. | |
parent who has to take their kids somewhere, two or three hours before | :07:59. | :08:01. | |
picking them up again, it might not be worth your while coming home in | :08:02. | :08:06. | |
the meantime. So what you do if you are a parent who is too busy to | :08:07. | :08:11. | |
ferry their kids around the city? Kate Russell has taken a pop, skip | :08:12. | :08:13. | |
and drive to check this out. He would be a teenager? Lots of | :08:14. | :08:23. | |
homework, hectic after-school activity schedule, Dean ferried | :08:24. | :08:29. | |
around by time poor parents. -- being ferried. Hang on a minute, | :08:30. | :08:33. | |
that looks pretty good for the teenager at least! Juggling work | :08:34. | :08:39. | |
commitments with school runs and extra curricular activities can be a | :08:40. | :08:44. | |
nightmare. For some parents are taxi is the only option. But how did know | :08:45. | :08:48. | |
your children are safe getting into a car with a stranger. That is the | :08:49. | :08:56. | |
question Hop, Skip, Drive set out to answer. Three of us created the | :08:57. | :09:01. | |
company. We all have kids, we have six kids between us. Every kid | :09:02. | :09:06. | |
activity known to man is participated in. We just couldn't | :09:07. | :09:11. | |
manage it any more. Carly loves to dance, and she's pretty good at it. | :09:12. | :09:15. | |
We originally lived somewhere different. Then we moved here and we | :09:16. | :09:22. | |
couldn't get her to class on time, coming from work, all the way back | :09:23. | :09:26. | |
here and so we put in a local dance studio, which wasn't the same. Hop, | :09:27. | :09:31. | |
Skip, Drive says it runs background checks on all its drivers, including | :09:32. | :09:36. | |
fingerprinting. They also told us applicants must have at least five | :09:37. | :09:40. | |
years childcare experience. Really it is caregiving. So for my son, | :09:41. | :09:47. | |
Jackson, the caregiver will Parker, go into the school, sign him out, | :09:48. | :09:50. | |
there is a booster seat for him because he is not eight and I can | :09:51. | :09:54. | |
track the ride as a parent the whole way. Rides are scheduled ahead of | :09:55. | :09:59. | |
time, with clearly branded cars and drivers and have a secret password | :10:00. | :10:03. | |
to give to the children on pickup, so everyone can be sure they are | :10:04. | :10:06. | |
getting in the right car. What happens when they pick you up? First | :10:07. | :10:10. | |
I get a text message on my phone and then they come up to the door, they | :10:11. | :10:16. | |
tell me the password and we go in the car and make conversation. For | :10:17. | :10:21. | |
the drivers, many of them mothers themselves, it's a much more | :10:22. | :10:23. | |
rewarding job than just driving a taxi. How is it driving lots of | :10:24. | :10:29. | |
different kids? What approach to you use? It depends. It depends on the | :10:30. | :10:36. | |
child. My main objective is just too obviously be friendly and some kids | :10:37. | :10:41. | |
just get in and they have their headphones on and they have their | :10:42. | :10:46. | |
music trading, you know, and they just need to chill out. -- music | :10:47. | :10:51. | |
playing. Because they have just come from school or at times. At other | :10:52. | :10:56. | |
times the kids are really chatty, so I just have to take the emotional | :10:57. | :11:04. | |
temperature. Now Carly can go to the dance studio she loves and Jennifer | :11:05. | :11:08. | |
can manage her life more easily. I don't have to leave work early every | :11:09. | :11:12. | |
day, because you know that's good I need to earn to pay for dams. Hop, | :11:13. | :11:18. | |
Skip, Drive plans a careful expansion into more regions in the | :11:19. | :11:21. | |
future. They also added carpooling to reduce the cost and this should | :11:22. | :11:26. | |
help ease congestion around school drop-off points as well. | :11:27. | :11:31. | |
That's it for the shortcut of Click for this week. Much more in the | :11:32. | :11:37. | |
full-length version on iPlayer online and on Twitter. Thanks for | :11:38. | :11:42. | |
watching. See you soon. | :11:43. | :11:43. |