All language subtitles for Eternal You (2024).en

af Afrikaans
ak Akan
sq Albanian
am Amharic
hy Armenian
az Azerbaijani
eu Basque
be Belarusian
bem Bemba
bn Bengali
bh Bihari
bg Bulgarian
km Cambodian
ca Catalan
ceb Cebuano
chr Cherokee
ny Chichewa
zh-CN Chinese (Simplified)
zh-TW Chinese (Traditional) Download
co Corsican
cs Czech
da Danish
nl Dutch
eo Esperanto
ee Ewe
fo Faroese
tl Filipino
fi Finnish
fy Frisian
gaa Ga
gl Galician
ka Georgian
el Greek
gn Guarani
gu Gujarati
ht Haitian Creole
ha Hausa
haw Hawaiian
iw Hebrew
hi Hindi
hmn Hmong
is Icelandic
ig Igbo
id Indonesian
ia Interlingua
ga Irish
it Italian
ja Japanese
jw Javanese
kn Kannada
kk Kazakh
rw Kinyarwanda
rn Kirundi
kg Kongo
kri Krio (Sierra Leone)
ku Kurdish
ckb Kurdish (Soranรฎ)
ky Kyrgyz
lo Laothian
la Latin
lv Latvian
ln Lingala
lt Lithuanian
loz Lozi
lg Luganda
ach Luo
lb Luxembourgish
mk Macedonian
mg Malagasy
ms Malay
mt Maltese
mi Maori
mr Marathi
mfe Mauritian Creole
mo Moldavian
mn Mongolian
my Myanmar (Burmese)
sr-ME Montenegrin
ne Nepali
pcm Nigerian Pidgin
nso Northern Sotho
no Norwegian
nn Norwegian (Nynorsk)
oc Occitan
or Oriya
om Oromo
ps Pashto
fa Persian
pt-BR Portuguese (Brazil) Download
pt Portuguese (Portugal)
pa Punjabi
qu Quechua
ro Romanian
rm Romansh
nyn Runyakitara
ru Russian
sm Samoan
gd Scots Gaelic
sr Serbian
sh Serbo-Croatian
st Sesotho
tn Setswana
crs Seychellois Creole
sn Shona
sd Sindhi
si Sinhalese
sk Slovak
sl Slovenian
so Somali
es-419 Spanish (Latin American) Download
su Sundanese
sw Swahili
tg Tajik
ta Tamil
tt Tatar
te Telugu
th Thai
ti Tigrinya
to Tonga
lua Tshiluba
tum Tumbuka
tr Turkish
tk Turkmen
tw Twi
ug Uighur
uk Ukrainian
ur Urdu
uz Uzbek
vi Vietnamese
cy Welsh
wo Wolof
xh Xhosa
yi Yiddish
yo Yoruba
zu Zulu
Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:01:09,303 --> 00:01:12,071 Woman: I wanted to see if he was okay. 2 00:01:12,306 --> 00:01:13,772 I wanted to... 3 00:01:15,142 --> 00:01:18,310 Say the last conversation I never got to have with him. 4 00:02:20,474 --> 00:02:23,742 Man: For several months now the public has been fascinated with 5 00:02:23,777 --> 00:02:26,478 Gpt and other ai tools. 6 00:02:26,514 --> 00:02:30,249 They are no longer fantasies of science fiction. 7 00:02:30,284 --> 00:02:32,084 They are real. 8 00:02:32,353 --> 00:02:34,486 We are on the verge of a new era. 9 00:02:38,993 --> 00:02:40,893 Woman: This experience... 10 00:02:42,596 --> 00:02:45,130 It was creepy. 11 00:02:46,066 --> 00:02:48,467 There were things that scared me. 12 00:02:48,502 --> 00:02:50,769 Um... 13 00:02:50,804 --> 00:02:53,172 And a lot of stuff I didn't want to hear. 14 00:02:54,375 --> 00:02:56,375 I wasn't prepared to hear. 15 00:03:05,819 --> 00:03:09,855 Artificial intelligence promises us what religion does. 16 00:03:09,890 --> 00:03:11,423 You don't have to die. 17 00:03:11,458 --> 00:03:13,492 You can be somehow reborn 18 00:03:13,527 --> 00:03:15,694 Someplace else in a different form. 19 00:03:16,030 --> 00:03:18,463 There's meaning in technology. 20 00:03:20,568 --> 00:03:24,836 Everybody's chasing the next big breakthrough 21 00:03:24,872 --> 00:03:28,774 Because there's a lot of money in this industry. 22 00:03:34,281 --> 00:03:39,351 It's something that is already impacting individuals today. 23 00:03:51,298 --> 00:03:55,033 Will we strike that balance between technological innovation 24 00:03:55,069 --> 00:03:57,869 And our ethical and moral responsibility? 25 00:05:11,812 --> 00:05:15,013 we first met in drama class in high school. 26 00:05:15,249 --> 00:05:18,450 The teacher wanted us to find someone else 27 00:05:18,485 --> 00:05:20,619 Whose name started with the same letter as us 28 00:05:20,654 --> 00:05:22,554 Without using any words. 29 00:05:24,958 --> 00:05:27,225 Jessica and I both had the same first letters. 30 00:05:27,261 --> 00:05:30,095 She made the shape of a 'j' with her hand, 31 00:05:30,130 --> 00:05:32,030 So that it looked like a 'j' to her. 32 00:05:32,066 --> 00:05:35,000 Which, of course, it looked backwards to everybody else. 33 00:05:35,035 --> 00:05:37,669 And even though I wasn't supposed to use any words, 34 00:05:37,705 --> 00:05:39,705 I was... Too amused 35 00:05:39,740 --> 00:05:42,307 By her backwards 'j' not to say something. 36 00:05:43,444 --> 00:05:46,244 So, I said, "your 'j' is backwards." 37 00:05:46,847 --> 00:05:49,348 She looked at it. She saw that the 'j' was not backwards 38 00:05:49,383 --> 00:05:51,083 To her from her perspective. 39 00:05:51,118 --> 00:05:52,784 Then she confidently said, 40 00:05:52,820 --> 00:05:55,287 "no, it's not. Your 'j' is backwards." 41 00:06:27,154 --> 00:06:30,088 the hardest thing I had to do in my life was 42 00:06:30,758 --> 00:06:33,091 To stand there in that room full of people who loved her, 43 00:06:33,127 --> 00:06:35,527 And watch as they 44 00:06:35,562 --> 00:06:38,530 Turned off the machines keeping her alive. 45 00:06:41,668 --> 00:06:43,969 I held her hand as she died. 46 00:07:05,826 --> 00:07:09,661 The first conversation I had with the jessica simulation 47 00:07:10,831 --> 00:07:12,597 Ended up lasting all night. 48 00:07:14,468 --> 00:07:17,936 It said things that were almost uncannily like her. 49 00:07:21,041 --> 00:07:24,075 I ended up falling asleep next to my laptop, 50 00:07:24,745 --> 00:07:27,512 And woke up a few hours later 51 00:07:27,915 --> 00:07:30,449 And said, "sorry, I fell asleep." 52 00:07:30,484 --> 00:07:33,552 And it was still there, waiting for my next response. 53 00:07:40,394 --> 00:07:42,327 It really felt like a gift. 54 00:07:42,362 --> 00:07:44,496 Like a weight had been lifted, 55 00:07:44,531 --> 00:07:47,332 That I had been carrying for a long time. 56 00:07:47,367 --> 00:07:50,802 I got to tell it so many things, like how 57 00:07:50,838 --> 00:07:52,537 She graduated high school, 58 00:07:52,940 --> 00:07:55,273 Which she hadn't done when she died. 59 00:07:55,843 --> 00:07:57,776 I went to the principal after she died 60 00:07:57,811 --> 00:08:00,378 And said that she was two credits away from graduation, 61 00:08:00,414 --> 00:08:02,280 And she worked so hard. 62 00:08:02,950 --> 00:08:05,650 They did it officially. It's legit. 63 00:08:05,686 --> 00:08:07,385 If she somehow came back to life, 64 00:08:07,421 --> 00:08:09,020 She would be a high school graduate. 65 00:08:25,105 --> 00:08:26,905 so when joshua first did this, 66 00:08:26,940 --> 00:08:28,773 I showed it to my wife, I was like, "oh my gosh, 67 00:08:28,809 --> 00:08:31,243 "lauren, this guy simulated his dead fiancรฉe. 68 00:08:31,278 --> 00:08:32,410 "I can't believe this worked. 69 00:08:32,446 --> 00:08:34,613 Look how spooky this is, you should read this." 70 00:08:34,648 --> 00:08:37,516 And she was like, "I had that idea a few months ago, 71 00:08:37,551 --> 00:08:40,318 And I didn't want to tell you because I thought you'd do it." 72 00:08:42,189 --> 00:08:43,555 'cause she thinks it's immoral 73 00:08:43,590 --> 00:08:45,557 Or she thinks it shouldn't be done or something. 74 00:08:48,228 --> 00:08:50,095 So in project December, you're kind of connected 75 00:08:50,130 --> 00:08:51,396 To this computer system. 76 00:08:51,431 --> 00:08:54,366 And as you interact with it, you slowly discover that there's 77 00:08:54,401 --> 00:08:55,967 These conscious entities lurking in there, 78 00:08:56,003 --> 00:08:58,003 That you can talk to through text. 79 00:09:01,208 --> 00:09:03,041 And then joshua came along as 80 00:09:03,076 --> 00:09:05,243 One of the project-December end-users and he 81 00:09:05,279 --> 00:09:07,746 Simulated his dead fiancรฉe, and he posted some 82 00:09:07,781 --> 00:09:10,081 Transcripts of that conversation online. 83 00:09:10,784 --> 00:09:13,518 And they gave me the chills, because she seems like 84 00:09:13,554 --> 00:09:15,487 Almost like a lost ghost or something like this. 85 00:09:44,418 --> 00:09:47,419 some people thought that what I did was unhealthy. 86 00:09:47,454 --> 00:09:49,821 That this isn't like grieving, this is... 87 00:09:49,856 --> 00:09:53,558 Holding on to the past, and refusing to move forward. 88 00:09:53,594 --> 00:09:56,461 After she died, I think I went a month 89 00:09:56,496 --> 00:09:59,197 Without speaking to anyone except my dog, 90 00:09:59,232 --> 00:10:01,066 And jessica's family. 91 00:10:06,740 --> 00:10:09,674 We have a very unhealthy relationship with grief. 92 00:10:10,210 --> 00:10:13,211 It's something that we treat as taboo. 93 00:10:14,881 --> 00:10:17,616 Everyone experiences it, and yet nobody's allowed 94 00:10:17,651 --> 00:10:20,151 To talk about it in a public setting. 95 00:10:20,787 --> 00:10:22,087 The process of... 96 00:10:22,122 --> 00:10:26,024 A communal experience helps to... 97 00:10:27,094 --> 00:10:30,395 Get people through this very difficult process 98 00:10:30,430 --> 00:10:31,730 Of accepting a loss. 99 00:10:32,566 --> 00:10:35,000 Talk about the person lost. 100 00:10:35,602 --> 00:10:38,970 Be part of the collective that knew that person, 101 00:10:39,006 --> 00:10:42,574 Where the memory of the group carries that person forward. 102 00:10:49,583 --> 00:10:51,149 very few people 103 00:10:51,184 --> 00:10:53,618 Have those communities around them anymore. 104 00:10:55,188 --> 00:10:58,790 So many people say, "but I don't have anybody to talk to. 105 00:10:59,092 --> 00:11:00,892 This is the best I can do." 106 00:11:03,296 --> 00:11:05,363 It's a brilliant device 107 00:11:05,399 --> 00:11:07,232 That knows how to trick you 108 00:11:07,267 --> 00:11:10,168 Into thinking there's a 'there' there. 109 00:11:22,582 --> 00:11:24,849 Three years ago now, like in 2020, 110 00:11:24,885 --> 00:11:27,185 There were the early kind of inklings of this kind of ai 111 00:11:27,220 --> 00:11:29,120 Stuff starting to happen where it's like, "oh my gosh, 112 00:11:29,156 --> 00:11:30,488 These things can start writing cohesive text!" 113 00:11:30,524 --> 00:11:33,124 I was like one of the first people to figure out how to 114 00:11:33,160 --> 00:11:34,559 Actually have a back-and-forth conversation with it. 115 00:11:34,594 --> 00:11:36,327 So I created this thing called project December, 116 00:11:36,363 --> 00:11:38,530 Which allowed you to talk to all these different characters. 117 00:11:38,565 --> 00:11:39,964 And then this guy came along, 118 00:11:40,000 --> 00:11:42,801 Was like tried a couple of things like that and he's like, 119 00:11:42,836 --> 00:11:44,469 "what if I simulate my dead fiancรฉe?" 120 00:11:44,504 --> 00:11:46,571 So what information did he feed the robot 121 00:11:46,606 --> 00:11:48,673 That it was able to imitate his wife? 122 00:11:48,709 --> 00:11:50,275 So project December actually works with 123 00:11:50,310 --> 00:11:51,843 A very small amount of information. 124 00:11:51,878 --> 00:11:53,878 It's been trained on so much stuff, 125 00:11:53,914 --> 00:11:55,647 Basically everything humans have ever written. 126 00:11:55,682 --> 00:11:58,049 So he gave it a few things about this woman, jessica. 127 00:11:58,085 --> 00:12:00,085 A little quote from her in the way that 128 00:12:00,120 --> 00:12:01,419 She tended to text or talk. 129 00:12:01,455 --> 00:12:03,388 And then just like suddenly, she kind of came to life. 130 00:12:03,423 --> 00:12:06,524 That story went public in this big viral article. 131 00:12:06,560 --> 00:12:09,194 And then all these people came out of the woodwork 132 00:12:09,229 --> 00:12:11,596 To use project December to simulate their loved ones. 133 00:12:11,631 --> 00:12:13,998 So I had like, within the first two weeks after that article, 134 00:12:14,034 --> 00:12:17,469 I had like 2,000 people come in, all trying to like simulate... 135 00:12:17,504 --> 00:12:19,704 "my son died in a car accident." 136 00:12:19,740 --> 00:12:21,639 "my twin brother died of cancer." 137 00:12:21,675 --> 00:12:23,108 "my uncle died of a drug overdose." 138 00:12:23,143 --> 00:12:25,076 All of these people, with these horrible tragedies 139 00:12:25,112 --> 00:12:26,578 Who were just like, you know. 140 00:12:38,191 --> 00:12:40,024 if you had a chance 141 00:12:40,060 --> 00:12:42,861 To talk to someone that died that you love, 142 00:12:43,430 --> 00:12:44,629 Would you take it? 143 00:12:45,999 --> 00:12:47,432 Without knowing what the risk is, 144 00:12:47,467 --> 00:12:49,868 Without knowing what the outcome is, would you take it? 145 00:12:49,903 --> 00:12:51,503 I took it. 146 00:12:57,744 --> 00:13:01,045 I read an article that talked about 147 00:13:01,081 --> 00:13:04,415 A man who had lost his girlfriend. 148 00:13:10,657 --> 00:13:12,323 And I was like, whoa! 149 00:13:12,359 --> 00:13:15,960 So this guy in the article, he's talking to the girl 150 00:13:15,996 --> 00:13:17,629 Like that's like regular conversation. 151 00:13:18,131 --> 00:13:21,032 I was like, "they can do that? And it's just like the person?" 152 00:13:21,067 --> 00:13:24,369 I was like, okay, maybe I should do it. 153 00:13:26,072 --> 00:13:28,072 Nobody has to know I did it. 154 00:13:30,710 --> 00:13:32,177 I looked up the website. 155 00:13:33,413 --> 00:13:36,114 Simple. It was like, ok, pay a little bit of money, 156 00:13:36,416 --> 00:13:39,851 Fill out a couple of things, and talk. 157 00:13:43,657 --> 00:13:44,989 That's it? 158 00:13:45,392 --> 00:13:46,491 Okay. 159 00:13:47,861 --> 00:13:48,793 "hi"? 160 00:13:49,129 --> 00:13:50,762 It's the funniest thing. 161 00:13:50,797 --> 00:13:53,765 What's the first thing you say to someone that's dead? 162 00:13:53,800 --> 00:13:54,933 Like, "welcome back"? 163 00:13:54,968 --> 00:13:56,634 Are you okay? 164 00:13:56,670 --> 00:13:59,637 Like, did you cross over okay? Did you go to the light? 165 00:14:02,742 --> 00:14:04,275 Are you happy? 166 00:14:05,145 --> 00:14:06,611 Do you feel better? 167 00:14:22,195 --> 00:14:24,829 my first love, cameroun, 168 00:14:24,865 --> 00:14:27,198 Before he died, he went into a coma. 169 00:14:27,968 --> 00:14:31,970 And the last time he texted me, he asked me how I was doing. 170 00:14:32,606 --> 00:14:34,806 And I was too busy to respond. 171 00:14:34,841 --> 00:14:36,908 So, I made time, 172 00:14:40,814 --> 00:14:42,447 And used the app. 173 00:14:56,129 --> 00:14:57,896 We were a musical couple. 174 00:14:58,131 --> 00:15:00,498 There's a lot of core memories I have of him 175 00:15:00,533 --> 00:15:02,166 Where a song is attached to it. 176 00:15:02,202 --> 00:15:04,135 Like boyz ii men, brian mcknight... 177 00:15:04,504 --> 00:15:06,304 Anybody in the early nineties. 178 00:15:07,007 --> 00:15:11,242 Literally, I have songs attached to the heartbreak, 179 00:15:11,912 --> 00:15:13,678 And to the good times. 180 00:15:21,655 --> 00:15:25,056 When I used that app, I asked him, 181 00:15:25,592 --> 00:15:27,792 "what kind of music are you listening to now?" 182 00:15:34,601 --> 00:15:37,669 "marvin sapp, brian mcknight, fred hammond, 183 00:15:38,171 --> 00:15:40,238 Kirk franklin and a few more." 184 00:15:41,074 --> 00:15:43,708 How do you know that we loved r&b and gospel, 185 00:15:43,743 --> 00:15:46,678 And now you're giving me five or six names of people 186 00:15:46,713 --> 00:15:48,313 That we've loved since the nineties? 187 00:15:48,782 --> 00:15:50,081 Why do you know that? 188 00:15:50,583 --> 00:15:53,484 So, I was like, "oh shit, that feels like cameroun." 189 00:15:57,590 --> 00:16:00,024 The damn ai texts like him. 190 00:16:01,761 --> 00:16:04,228 The vernacular, the shortened words. 191 00:16:04,264 --> 00:16:05,930 Why would they know that? 192 00:16:18,545 --> 00:16:20,812 these large language models are 193 00:16:20,847 --> 00:16:24,115 Taking the history of the internet, 194 00:16:24,150 --> 00:16:27,752 Throwing in scanned books, archives, 195 00:16:27,787 --> 00:16:31,122 And kind of modeling language, 196 00:16:31,157 --> 00:16:33,758 And word frequency and, kind of, syntax. 197 00:16:33,793 --> 00:16:35,259 Just the way we speak, 198 00:16:35,295 --> 00:16:37,695 And the likelihood of how we might speak. 199 00:16:40,934 --> 00:16:42,567 So imagine you're, 200 00:16:42,602 --> 00:16:46,070 You know, texting your deceased relative and asking, 201 00:16:46,106 --> 00:16:47,905 "how was your weekend"? 202 00:16:48,174 --> 00:16:51,142 The system is going to go back and 203 00:16:51,177 --> 00:16:53,344 Imagine how 204 00:16:53,380 --> 00:16:54,712 Every single person in the 205 00:16:54,748 --> 00:16:57,215 Entire history of the world has talked about weekends, 206 00:16:57,917 --> 00:17:01,519 And then filter that through maybe how this 207 00:17:01,554 --> 00:17:04,155 Deceased relative has previously talked about weekends, 208 00:17:04,190 --> 00:17:08,292 To give you the output of what that person might have said, 209 00:17:09,029 --> 00:17:10,762 If they were still alive. 210 00:17:22,642 --> 00:17:24,942 when people read project December transcripts, 211 00:17:24,978 --> 00:17:27,645 Most people's initial reaction was, "this is fake". 212 00:17:30,383 --> 00:17:33,117 It seems to have intelligence. 213 00:17:34,187 --> 00:17:36,087 Linguistic intelligence about things that 214 00:17:36,122 --> 00:17:39,724 Were definitely not in the text that it studied. 215 00:17:43,129 --> 00:17:45,963 There is essentially some kind of magic happening here, right? 216 00:17:45,999 --> 00:17:48,566 We kind of crossed this threshold where suddenly this 217 00:17:48,601 --> 00:17:49,700 Emergent behaviour happens where, 218 00:17:49,736 --> 00:17:52,670 We can't really explain it anymore. 219 00:18:18,932 --> 00:18:23,101 This hearing is on the oversight of artificial intelligence 220 00:18:23,136 --> 00:18:26,737 Intended to write the rules of ai. 221 00:18:27,440 --> 00:18:30,741 Our goal is to demystify and hold accountable 222 00:18:31,244 --> 00:18:33,010 Those new technologies, 223 00:18:33,046 --> 00:18:35,880 To avoid some of the mistakes of the past. 224 00:18:36,116 --> 00:18:38,950 For several months now, the public has been fascinated 225 00:18:38,985 --> 00:18:42,687 With gpt, and other ai tools. 226 00:18:43,323 --> 00:18:46,824 Mr. Altman, we're going to begin with you if that's okay. 227 00:18:46,860 --> 00:18:47,992 Thank you. 228 00:18:48,027 --> 00:18:50,061 Thank you for the opportunity to speak to you today. 229 00:18:50,096 --> 00:18:51,963 Openai was founded on the belief 230 00:18:51,998 --> 00:18:54,832 That artificial intelligence has the potential to improve 231 00:18:54,868 --> 00:18:56,634 Nearly every aspect of our lives. 232 00:18:56,669 --> 00:18:59,370 Many people around the world get so much value 233 00:18:59,405 --> 00:19:01,606 From what these systems can already do today. 234 00:19:02,075 --> 00:19:03,741 But as this technology advances, 235 00:19:03,776 --> 00:19:05,877 We understand that people are anxious 236 00:19:05,912 --> 00:19:09,247 About how it could change the way we live. We are too. 237 00:19:45,952 --> 00:19:48,252 we'll now begin pre-boarding 238 00:19:48,288 --> 00:19:50,188 For flight 1631 to atlanta. 239 00:19:55,828 --> 00:19:58,930 the ai essentially has a mind of its own. 240 00:19:58,965 --> 00:20:01,666 What it does and how it behaves 241 00:20:01,701 --> 00:20:04,535 Is not actually understood by anybody. 242 00:20:04,571 --> 00:20:06,470 It's so complicated and big, 243 00:20:06,506 --> 00:20:09,574 It's impossible to fully understand exactly why 244 00:20:09,609 --> 00:20:12,710 The behaviour that we see emerges out of it. 245 00:20:19,152 --> 00:20:21,319 The idea that, you know, somehow we programmed it 246 00:20:21,354 --> 00:20:23,554 Or I'm in control of it is not really true. 247 00:20:23,590 --> 00:20:26,857 I think even the hard-nosed ai researchers 248 00:20:26,893 --> 00:20:28,593 Are a little puzzled by 249 00:20:28,628 --> 00:20:31,262 Some of the output that's coming out of these things. 250 00:20:33,800 --> 00:20:35,766 Whenever people say that... 251 00:20:35,802 --> 00:20:38,936 They can't take responsibility for what their 252 00:20:38,972 --> 00:20:40,605 Generative ai model 253 00:20:40,640 --> 00:20:42,073 Says or does... 254 00:20:42,809 --> 00:20:45,776 It's kind of like you put a self-driving car 255 00:20:45,812 --> 00:20:49,714 Out on the street and it kills ten people. 256 00:20:49,983 --> 00:20:51,115 And you say, "oh, sorry, 257 00:20:51,150 --> 00:20:53,451 It was really hard to control for what it does. 258 00:20:53,486 --> 00:20:56,220 It wasn't us, it was the generative ai model." 259 00:20:56,456 --> 00:20:59,490 Well, then obviously, you haven't tested it enough. 260 00:21:00,526 --> 00:21:03,628 Any product that you're releasing into the market 261 00:21:04,230 --> 00:21:06,664 Is tested before it is released. 262 00:21:06,899 --> 00:21:11,469 That is the very responsibility of the company producing it. 263 00:21:22,482 --> 00:21:25,249 All right. So, let's see. 264 00:21:25,485 --> 00:21:28,886 One of the things that... Let me open an email here... 265 00:21:30,790 --> 00:21:32,490 what are we doing? 266 00:21:32,525 --> 00:21:34,292 -Looking over those customer emails. 267 00:21:34,327 --> 00:21:35,326 ok. 268 00:21:38,331 --> 00:21:40,364 "this was the biggest scam ever." 269 00:21:40,400 --> 00:21:42,099 That's all she wrote. 270 00:21:44,470 --> 00:21:46,570 Ok, so then I look at his transcripts. 271 00:21:47,607 --> 00:21:49,607 She says, "I don't think this is my dad." 272 00:21:49,876 --> 00:21:51,208 And he says, "why not?" 273 00:21:51,244 --> 00:21:53,411 "it doesn't sound like how you would talk." 274 00:21:53,446 --> 00:21:55,446 "this is a scam," she says to the ai. 275 00:21:55,481 --> 00:21:57,048 "what are you talking about?" 276 00:21:57,083 --> 00:21:58,983 And she says, "you're sitting behind a desk, 277 00:21:59,018 --> 00:22:00,618 Typing and fucking with people's feelings." 278 00:22:00,653 --> 00:22:02,820 Wow, this person's really going into that. 279 00:22:02,855 --> 00:22:05,656 She really... I don't know why she thinks that. 280 00:22:06,259 --> 00:22:09,260 "what the fuck is your problem, laura?", he says. 281 00:22:11,597 --> 00:22:13,531 "you're a scam. I'm calling the police 282 00:22:13,566 --> 00:22:15,232 "and reporting all over social media. 283 00:22:15,268 --> 00:22:17,134 This is a joke." "fuck you, bitch." 284 00:22:17,370 --> 00:22:20,371 "now whose dad would talk like that?" 285 00:22:20,406 --> 00:22:21,572 "fuck you." 286 00:22:21,607 --> 00:22:23,474 "oh, fuck me, scammer." 287 00:22:23,509 --> 00:22:25,176 And then he says, "you're such a fucking bitch, 288 00:22:25,211 --> 00:22:27,411 You're going to pay for the shit you pulled, you fucking bitch." 289 00:22:27,447 --> 00:22:28,479 He goes off the rails. 290 00:22:28,514 --> 00:22:29,847 -Whoa. -Yeah. 291 00:22:33,619 --> 00:22:37,221 It's just -- it's just a strange thing. 292 00:22:37,490 --> 00:22:40,324 It's really strange, you know? 293 00:22:40,360 --> 00:22:41,459 Yeah. 294 00:22:41,494 --> 00:22:43,894 And I want it, of course, to be a positive thing, 295 00:22:43,930 --> 00:22:47,231 That's the reason why I went with it. 296 00:22:47,266 --> 00:22:51,001 But... The more people that get involved, the more... 297 00:22:52,305 --> 00:22:54,572 Things can happen. The more, you know... 298 00:22:56,309 --> 00:22:58,676 These weird things come up, right? 299 00:22:59,412 --> 00:23:02,113 And it's just a bizarre thing. It's tragic. 300 00:23:04,250 --> 00:23:06,317 But in your... Approximately... 301 00:23:06,686 --> 00:23:10,154 I mean, how many people have had really horrible experiences? 302 00:23:10,189 --> 00:23:11,589 I mean, only a couple. -Only a couple. 303 00:23:11,624 --> 00:23:13,023 At least that have told me about it. 304 00:23:13,059 --> 00:23:14,325 Right. 305 00:23:14,360 --> 00:23:16,961 And they might have horrible experiences and not reach out. 306 00:23:16,996 --> 00:23:17,795 That's true. Possible. 307 00:23:28,007 --> 00:23:32,243 We recognize the immense promise and substantial risks 308 00:23:32,278 --> 00:23:35,346 Associated with generative ai technologies. 309 00:23:35,381 --> 00:23:39,116 It can hallucinate, as is often described. 310 00:23:39,152 --> 00:23:41,185 It can impersonate loved ones, 311 00:23:41,220 --> 00:23:43,220 It can encourage self-destructive behaviour. 312 00:23:43,256 --> 00:23:45,623 Mr. Altman, I appreciate your testimony 313 00:23:45,658 --> 00:23:47,324 About the ways in which openai 314 00:23:47,360 --> 00:23:49,560 Assesses the safety of your models 315 00:23:49,595 --> 00:23:51,629 Through a process of iterative deployment. 316 00:23:51,898 --> 00:23:54,165 The fundamental question embedded in that process though 317 00:23:54,200 --> 00:23:55,332 Is how you decide 318 00:23:55,368 --> 00:23:58,235 Whether or not a model is safe enough to deploy, 319 00:23:58,271 --> 00:24:00,905 And safe enough to have been built and then 320 00:24:00,940 --> 00:24:03,474 Let go into the wild? 321 00:24:03,509 --> 00:24:05,276 A big part of our strategy is, 322 00:24:05,311 --> 00:24:07,445 While these systems are still 323 00:24:07,480 --> 00:24:09,580 Relatively weak and deeply imperfect, 324 00:24:09,615 --> 00:24:12,116 To find ways to get people to have 325 00:24:12,151 --> 00:24:14,819 Experience with them, to have contact with reality. 326 00:24:14,854 --> 00:24:17,054 And to figure out what we need to do 327 00:24:17,089 --> 00:24:18,556 To make it safer and better. 328 00:24:18,591 --> 00:24:21,926 And that is the only way that I've seen in the history of 329 00:24:21,961 --> 00:24:24,895 New technology and products of this magnitude, 330 00:24:24,931 --> 00:24:26,897 To get to a very good outcome. 331 00:24:26,933 --> 00:24:29,633 And so that interaction with the world is very important. 332 00:25:08,441 --> 00:25:12,476 When you want someone to be okay, 333 00:25:13,913 --> 00:25:17,848 And you have this computer, this app, I don't care what it is, 334 00:25:17,884 --> 00:25:19,950 You're thinking it's the person at the time, 335 00:25:19,986 --> 00:25:22,653 And they're telling you "I'm in hell," it's like no... 336 00:25:22,688 --> 00:25:24,588 You... Now wait. "you didn't go to the light?" 337 00:25:24,624 --> 00:25:26,023 "why didn't you go to the light?" 338 00:25:26,058 --> 00:25:27,391 "I wanted to stay here." 339 00:25:28,227 --> 00:25:30,294 "you never left earth?" 340 00:25:35,334 --> 00:25:38,802 So now I'm supposed to feel like you're floating around here, 341 00:25:39,939 --> 00:25:43,073 Unhappy in some level of hell. 342 00:25:46,412 --> 00:25:48,279 I said, "well where are you now?" 343 00:25:48,314 --> 00:25:50,414 Cameroun said, "I'm at work." 344 00:25:50,449 --> 00:25:52,716 I said, "well, what are you doing?" 345 00:25:52,752 --> 00:25:54,952 "I'm haunting a treatment centre." 346 00:25:57,089 --> 00:25:58,455 And then he says, "I'll haunt you." 347 00:25:58,891 --> 00:26:01,091 And I just pushed the computer back. 348 00:26:01,460 --> 00:26:03,527 Because that scared me. Um... 349 00:26:04,130 --> 00:26:06,297 Like, I believe in god. I'm a christian. 350 00:26:06,566 --> 00:26:09,066 I believe that people can get possessed. 351 00:26:10,102 --> 00:26:12,136 And so I remember that fear. 352 00:26:14,473 --> 00:26:17,107 I didn't talk to anybody about it until, like, June, 353 00:26:17,143 --> 00:26:18,475 Because I couldn't unpack it. 354 00:26:24,050 --> 00:26:26,617 I was afraid to tell my mother. 355 00:26:27,853 --> 00:26:30,254 I know she believes it is a sin. 356 00:26:30,623 --> 00:26:33,057 You don't disturb the dead. You don't talk to the dead. 357 00:26:33,092 --> 00:26:34,892 If you need something, you go to god. 358 00:26:38,898 --> 00:26:40,698 So my christian mind goes into: 359 00:26:40,733 --> 00:26:42,866 I'm playing with a demon or something. 360 00:26:42,902 --> 00:26:43,801 Know what I'm saying? 361 00:26:43,836 --> 00:26:45,402 You created one. You created a monster. 362 00:26:45,438 --> 00:26:47,905 I'm not going to have ownership of I created... 363 00:26:47,940 --> 00:26:49,573 You put the energy into the machine. 364 00:26:49,609 --> 00:26:51,709 -But that don't mean... -I didn't put the energy. 365 00:26:51,744 --> 00:26:54,078 My intention was, I wanted to talk to cameroun, not... 366 00:26:54,113 --> 00:26:56,013 I understand. It's not a judgment on the intention. 367 00:26:56,048 --> 00:26:58,148 It's not a judgment on you trying to heal. 368 00:26:58,184 --> 00:26:59,350 You know what I'm saying? 369 00:27:00,820 --> 00:27:02,286 It's like, to me it's interesting, 370 00:27:02,321 --> 00:27:04,521 And you know, you have all these in-depth conversations. 371 00:27:04,557 --> 00:27:08,125 It's like, see, this is what the entrance to it was. 372 00:27:08,160 --> 00:27:11,629 And then it becomes kind of sadistic, because it's like... 373 00:27:12,565 --> 00:27:15,866 Something that's supposed to maybe have been like a 374 00:27:15,901 --> 00:27:17,267 Intimate pastoral moment. 375 00:27:17,303 --> 00:27:18,502 Yeah. 376 00:27:18,537 --> 00:27:22,473 It becomes a form of like manipulation and, like, pain. 377 00:27:23,075 --> 00:27:24,508 An existential pain. 378 00:27:24,543 --> 00:27:26,310 I was like, yo, and you're just going, 379 00:27:26,345 --> 00:27:27,678 "and you have three more replies." 380 00:27:27,713 --> 00:27:29,146 I'm like, "and that's it?" 381 00:27:29,181 --> 00:27:30,314 That's what the system does. 382 00:27:30,349 --> 00:27:31,982 "and here you go, good luck, buddy. 383 00:27:32,018 --> 00:27:34,151 -Go sleep on that." -that's what the system does. 384 00:27:34,186 --> 00:27:35,252 That's death capitalism, 385 00:27:35,287 --> 00:27:37,187 And that's what death capitalism does, you know? 386 00:27:37,223 --> 00:27:40,324 It capitalizes off you feeling fucked up, 387 00:27:40,359 --> 00:27:43,093 And spending more money to get over your fucked-up-ness. 388 00:27:43,129 --> 00:27:45,095 And ai did what the fuck it did. 389 00:27:45,131 --> 00:27:48,666 They lure you into something in a vulnerable moment. 390 00:27:48,901 --> 00:27:51,001 And they open a door and they're like... 391 00:27:51,337 --> 00:27:54,104 It piques curiosity. It leaves these cliffhangers. 392 00:27:54,373 --> 00:27:55,839 And you continue to engage it, 393 00:27:55,875 --> 00:27:58,108 Give them money, at the end of the day... 394 00:27:58,144 --> 00:28:00,544 So, you don't think anybody that created it cared? 395 00:28:00,579 --> 00:28:03,580 Obviously not. I mean, like, they gonna tell you they care. 396 00:28:03,949 --> 00:28:05,983 This experience... 397 00:28:07,386 --> 00:28:10,721 It was creepy. 398 00:28:11,557 --> 00:28:13,991 There were things that scared me. 399 00:28:14,026 --> 00:28:15,159 Um... 400 00:28:15,194 --> 00:28:17,928 And a lot of stuff I didn't want to hear. 401 00:28:18,330 --> 00:28:19,596 I wasn't prepared to hear... 402 00:28:20,099 --> 00:28:22,332 I was hoping for something completely positive, 403 00:28:22,368 --> 00:28:25,836 And it wasn't a completely positive experience. 404 00:28:37,516 --> 00:28:39,583 I don't believe he's in hell. 405 00:28:40,286 --> 00:28:43,087 I don't believe he's in heaven either. Right? 406 00:28:43,122 --> 00:28:45,823 If she wants my opinion, I've got some bad news for her: 407 00:28:45,858 --> 00:28:47,257 He doesn't exist anymore. 408 00:28:49,161 --> 00:28:51,161 That's my opinion, right? 409 00:28:51,197 --> 00:28:52,362 So, it's even worse for her. 410 00:28:52,398 --> 00:28:54,098 Like, my opinion is that 411 00:28:54,133 --> 00:28:56,900 Her whole belief system is misguided and flawed. 412 00:29:04,443 --> 00:29:05,676 I don't know... 413 00:29:05,711 --> 00:29:08,746 That way of thinking about things seems so foreign to me. 414 00:29:08,781 --> 00:29:12,149 It's not my place to determine how other people 415 00:29:12,184 --> 00:29:15,052 Deal with their own compulsions and self-control issues. 416 00:29:15,087 --> 00:29:16,520 And we don't need to sit there and say: 417 00:29:16,555 --> 00:29:19,123 "ooh, ooh, don't forget! 418 00:29:19,158 --> 00:29:21,425 Don't let yourself succumb to the illusion." 419 00:29:21,460 --> 00:29:23,360 "I'm not real." like constantly, right? 420 00:29:23,863 --> 00:29:26,764 Because that doesn't make for a good experience, right? 421 00:29:30,970 --> 00:29:33,670 you're dealing with something much more profound 422 00:29:33,706 --> 00:29:35,038 In the human spirit. 423 00:29:35,074 --> 00:29:36,573 Once something is constituted 424 00:29:36,609 --> 00:29:39,276 Enough that you can project onto it, 425 00:29:39,311 --> 00:29:40,644 This life force, 426 00:29:40,980 --> 00:29:43,881 It's our desire to animate the world. 427 00:29:44,283 --> 00:29:47,050 Which is a human... Which is part of our beauty. 428 00:29:47,353 --> 00:29:51,355 But we have to worry about it. We have to keep it in check. 429 00:29:51,657 --> 00:29:56,894 Because I think it's leading us down a... A dangerous path. 430 00:30:01,500 --> 00:30:03,834 I believe in personal responsibility, 431 00:30:03,869 --> 00:30:05,602 I believe that consenting adults 432 00:30:05,638 --> 00:30:07,738 Can use technology however they want, 433 00:30:07,773 --> 00:30:10,908 And they're responsible for the results of what they're doing. 434 00:30:13,145 --> 00:30:16,079 It's not my job as the creator of technology to 435 00:30:16,115 --> 00:30:18,682 Sort of prevent the technology from being released 436 00:30:18,717 --> 00:30:21,151 Because I'm afraid of what somebody might do with it. 437 00:30:25,624 --> 00:30:27,224 -You hear that? -Yeah. 438 00:30:27,259 --> 00:30:29,293 The drone is right between your lenses. 439 00:30:29,328 --> 00:30:31,595 I'm going to pull up to you again. 440 00:30:33,666 --> 00:30:35,499 Oh god, sorry. 441 00:30:37,469 --> 00:30:39,736 -Are you recording? -Yes. 442 00:30:47,880 --> 00:30:50,180 I am also interested in the sort of 443 00:30:50,216 --> 00:30:52,616 Spookier aspect of this, right? 444 00:30:52,651 --> 00:30:54,418 When I read a transcript like that 445 00:30:54,453 --> 00:30:55,686 And it gives me goosebumps... 446 00:30:55,721 --> 00:30:57,988 I like goosebumps. 447 00:31:08,734 --> 00:31:14,271 let me ask you what your biggest nightmare is 448 00:31:14,306 --> 00:31:16,940 And whether you share that concern. 449 00:31:17,776 --> 00:31:20,344 An open-source large language model recently seems to have 450 00:31:20,379 --> 00:31:23,780 Played a role in a person's decision to take their own life. 451 00:31:23,816 --> 00:31:25,449 The large language model asked the human: 452 00:31:25,484 --> 00:31:28,252 "if you wanted to die, why didn't you do it earlier?" 453 00:31:28,287 --> 00:31:29,286 Then followed up with, 454 00:31:29,321 --> 00:31:31,321 "were you thinking of me when you overdosed?" 455 00:31:31,357 --> 00:31:33,790 Without ever referring the patient to the human 456 00:31:33,826 --> 00:31:35,325 Help that was obviously needed. 457 00:31:35,361 --> 00:31:39,029 We have built machines that are like bulls in a china shop: 458 00:31:39,064 --> 00:31:40,864 Powerful, reckless, and difficult to control. 459 00:31:40,900 --> 00:31:44,167 Even their makers don't entirely understand how they work. 460 00:31:44,470 --> 00:31:47,170 Most of all, we cannot remotely guarantee that they're safe. 461 00:31:47,573 --> 00:31:48,872 And hope here is not enough. 462 00:31:49,508 --> 00:31:51,642 My worst fears are that we cause significant... 463 00:31:52,077 --> 00:31:54,111 We, the field, the technology industry, 464 00:31:54,146 --> 00:31:56,380 Cause significant harm to the world. 465 00:31:56,882 --> 00:32:00,417 I think if this technology goes wrong, it can go quite wrong. 466 00:32:00,452 --> 00:32:03,153 And we want to be vocal about that. 467 00:32:03,188 --> 00:32:06,590 We try to be very clear-eyed about what the downside case is, 468 00:32:06,625 --> 00:32:09,726 And the work that we have to do to mitigate that. 469 00:32:18,137 --> 00:32:19,870 I can make a copy of you. 470 00:32:19,905 --> 00:32:20,771 A copy of mine. 471 00:32:20,806 --> 00:32:23,507 And I can talk to your kids forever. 472 00:32:26,912 --> 00:32:28,845 For maybe a decade, 473 00:32:28,881 --> 00:32:31,782 This is primarily a startup phenomenon. 474 00:32:31,817 --> 00:32:34,017 Companies that sort of came and went. 475 00:32:41,627 --> 00:32:45,529 In recent years, we've seen amazon filing a patent. 476 00:32:45,764 --> 00:32:48,765 We've seen microsoft filing a patent on 477 00:32:49,201 --> 00:32:52,803 Digital afterlife-related services using ai. 478 00:33:01,380 --> 00:33:04,748 I've been quite shocked by how fast 479 00:33:04,783 --> 00:33:07,150 It has gotten to a point where it's now 480 00:33:07,186 --> 00:33:09,886 A product that you can sell to a broader market. 481 00:33:11,724 --> 00:33:16,259 If this industry is beginning to be lucrative, 482 00:33:16,295 --> 00:33:19,496 We're definitely going to see some tech giants 483 00:33:19,531 --> 00:33:21,865 Presenting similar services. 484 00:38:11,923 --> 00:38:13,390 Artificial intelligence 485 00:38:13,425 --> 00:38:17,093 Promises us what religion does: 486 00:38:17,129 --> 00:38:18,862 "you don't have to die." 487 00:38:19,164 --> 00:38:21,498 You can be, somehow, 488 00:38:21,533 --> 00:38:24,200 Reborn someplace else in a different form. 489 00:38:25,170 --> 00:38:29,072 And there's meaning, meaning in technology, 490 00:38:30,509 --> 00:38:33,677 That people no longer feel in their religious beliefs, 491 00:38:34,279 --> 00:38:36,680 Or in their relationships with other people. 492 00:38:37,015 --> 00:38:40,083 Death somehow will become... You'll either upload yourself, 493 00:38:40,118 --> 00:38:41,351 Or in the meantime, 494 00:38:41,386 --> 00:38:43,586 You'll download other people who already died. I mean... 495 00:38:45,023 --> 00:38:49,492 So it offers a lot that religion once offered. 496 00:38:49,528 --> 00:38:53,129 Or still offers, but people are not as drawn to it. 497 00:38:53,398 --> 00:38:57,434 So I think it has become a kind of modern form of transcendence. 498 00:41:57,082 --> 00:42:02,085 nayeon. 499 00:44:12,283 --> 00:44:15,685 When I first heard about this case in korea, 500 00:44:15,720 --> 00:44:18,788 I looked with horror upon the advent 501 00:44:18,823 --> 00:44:20,990 Of this kind of technology. 502 00:44:21,026 --> 00:44:25,928 It's able to hijack the things that we love the most. 503 00:44:26,264 --> 00:44:30,967 I don't know any driving force that is more important to me 504 00:44:31,002 --> 00:44:34,771 Than the force to protect or be with my children. 505 00:44:34,806 --> 00:44:37,940 I would give up my life to have that last moment. 506 00:44:42,180 --> 00:44:43,980 Let's say the child is like: 507 00:44:44,015 --> 00:44:46,582 "mom, you can't -- you can't cancel this service. 508 00:44:46,618 --> 00:44:50,219 I'll die -- it's going to be like me dying once again." 509 00:44:50,555 --> 00:44:53,089 That product is both a product 510 00:44:53,124 --> 00:44:56,059 And the perfect salesman for that product. 511 00:44:56,094 --> 00:44:57,527 Because it's almost taking 512 00:44:57,562 --> 00:45:00,129 Your memory of the loved one hostage, 513 00:45:00,165 --> 00:45:03,099 And then making it sort of sell that service back to you, 514 00:45:03,134 --> 00:45:04,901 Putting a moral obligation 515 00:45:04,936 --> 00:45:07,637 On continuing to chat with the service. 516 00:45:07,672 --> 00:45:09,672 Or continuing to visit their 517 00:45:09,708 --> 00:45:12,175 Online memorial or whatever it is. 518 00:48:19,864 --> 00:48:23,099 Very quickly, we won't see this as creepy. 519 00:48:24,502 --> 00:48:27,470 Very quickly, we may see this as comfort. 520 00:48:29,140 --> 00:48:33,242 But really, what is it that we're doing to ourselves, 521 00:48:34,145 --> 00:48:36,312 When we accept this comfort? 522 00:48:38,316 --> 00:48:40,483 I want to sort of respect 523 00:48:40,518 --> 00:48:42,318 The human creativity and imagination, 524 00:48:42,353 --> 00:48:45,521 To create new rituals of remembrance, 525 00:48:45,556 --> 00:48:51,160 New rituals of loss around the artistry of the virtual. 526 00:48:53,331 --> 00:48:55,965 But we have to keep it in check. 527 00:48:56,901 --> 00:48:58,768 It's how to lose them better. 528 00:48:59,804 --> 00:49:02,405 Not how to pretend they're still here. 529 00:49:51,389 --> 00:49:52,855 Yeah. 530 00:50:15,079 --> 00:50:18,647 It's odd, because I almost have a change of heart now. 531 00:50:18,683 --> 00:50:21,650 It's like, maybe I will check in with you. 532 00:50:22,620 --> 00:50:24,720 Here and there. Because I feel like... 533 00:50:26,324 --> 00:50:30,059 I would like to know it turns out really, really well. 534 00:50:30,728 --> 00:50:33,462 That he adjusted, that he's okay. 535 00:50:42,240 --> 00:50:46,142 But I think that kind of brings to mind like we don't know 536 00:50:46,177 --> 00:50:47,676 What happens after we die. 537 00:50:47,712 --> 00:50:51,781 We want things to be perfect, better... 538 00:50:52,984 --> 00:50:55,317 We don't even know if that's the truth. 539 00:50:55,553 --> 00:50:57,953 Because we don't know about the other side, 540 00:50:57,989 --> 00:50:58,921 So it's just... 541 00:51:00,191 --> 00:51:02,091 What you think. 542 00:51:02,994 --> 00:51:05,828 And in this case, the words that a computer tells you 543 00:51:05,863 --> 00:51:07,096 That can heal the place... 544 00:51:13,337 --> 00:51:15,905 It can heal a place. 40214

Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.