WEBVTT Kind: captions Language: en Style: ::cue(c.colorCCFFFF) { color: rgb(204,255,255); } ::cue(c.colorFEFEFE) { color: rgb(254,254,254); } ::cue(c.colorFF66FF) { color: rgb(255,102,255); } ::cue(c.green) { color: green; } ::cue(c.yellow) { color: yellow; } ## 1 00:00:00.001 --> 00:00:02.486 ​Earlier this year, I asked GPT-3, 2 00:00:02.486 --> 00:00:05.489 ​the new text-generation ​artificial-intelligence system from OpenAI, 3 00:00:05.489 --> 00:00:07.491 ​to give me some ideas for videos. 4 00:00:07.491 --> 00:00:08.993 ​There were a couple of good ones in there, 5 00:00:08.993 --> 00:00:11.996 ​but the problem was that ​almost all of them were fictional. 6 00:00:11.996 --> 00:00:15.499 ​Videos on this channel are usually ​about real things in the real world, 7 00:00:15.499 --> 00:00:18.502 ​and no-one's managed to connect up ​a fact-checker to the AI yet. 8 00:00:18.502 --> 00:00:20.004 ​But that got me thinking: 9 00:00:20.004 --> 00:00:22.006 ​what about channels who aren't ​limited that way? 10 00:00:22.006 --> 00:00:24.508 ​What about channels which ​talk about broader topics, 11 00:00:24.508 --> 00:00:27.511 ​or their own opinions, ​or do ridiculous challenges? 12 00:00:27.511 --> 00:00:30.514 ​Could the AI come up with ​new ideas for them? 13 00:00:30.514 --> 00:00:31.815 ​So I called up some folks, 14 00:00:31.815 --> 00:00:36.020 ​I gave the AI a few recent video titles ​from their channels, and I asked it... 15 00:00:36.020 --> 00:00:37.521 ​come up with more like that. 16 00:00:37.521 --> 00:00:40.024 ​Now, I wasn't allowed to ​do live demonstrations, 17 00:00:40.024 --> 00:00:42.292 ​that's in the OpenAI terms of service ​I agreed to, 18 00:00:42.292 --> 00:00:43.527 ​but ahead of each call, 19 00:00:43.527 --> 00:00:46.530 ​I got GPT-3 to generate ​a hundred ideas 20 00:00:46.530 --> 00:00:48.032 ​and then I picked the ​most interesting ones. 21 00:00:48.032 --> 00:00:52.036 ​Or at least, the ones where the AI ​didn't completely go off the rails. 22 00:00:52.036 --> 00:00:57.541 ​If you look at my list, I might have ​20, 30, 40, 50, 100 ideas there, 23 00:00:57.541 --> 00:00:59.043 ​but I don't want to do any one of them. 24 00:00:59.043 --> 00:01:01.545 ​It's not easy to come up with good ideas. 25 00:01:01.545 --> 00:01:03.347 ​Mehdi Sadaghdar runs ElectroBOOM, 26 00:01:03.347 --> 00:01:06.550 ​a channel that has really technical ​videos about electronics, 27 00:01:06.550 --> 00:01:11.055 ​but which is, perhaps unfairly, best known ​for Mehdi repeatedly shocking himself. 28 00:01:11.055 --> 00:01:14.558 ​...that created [arc sound] 29 00:01:14.558 --> 00:01:17.061 ​[bleep] [arcing continues] 30 00:01:17.061 --> 00:01:20.064 ​One thing that he had in common ​with all the folks I talked to: 31 00:01:20.064 --> 00:01:22.066 ​they all have a list of ideas. 32 00:01:22.066 --> 00:01:27.071 ​It might be a whiteboard, a spreadsheet, ​a text file, but everyone had The List. 33 00:01:27.071 --> 00:01:30.107 ​I didn't have much hope that GPT-3 ​would work well with Mehdi's channel, 34 00:01:30.107 --> 00:01:33.577 ​because surely you can't just jam ​some electrical-sounding words together 35 00:01:33.577 --> 00:01:36.080 ​and get something sensible out of it. ​Right? 36 00:01:36.080 --> 00:01:42.086 ​The very first one that it sent out ​was "charging phones from a Tesla coil". 37 00:01:42.086 --> 00:01:43.087 ​Uhh... 38 00:01:43.087 --> 00:01:46.090 ​I mean, if you have a super-powerful ​Tesla coil, yeah, that's possible. 39 00:01:46.090 --> 00:01:49.093 ​So if I was able to charge my phone, 40 00:01:49.093 --> 00:01:52.096 ​or at least try to charge my phone ​through Tesla coil, 41 00:01:52.096 --> 00:01:54.098 ​which is not a bad idea... 42 00:01:54.098 --> 00:01:59.603 ​although, yeah, I'd have to find ​a disposable phone for that one. 43 00:01:59.603 --> 00:02:04.274 ​The next one was ​"Lemon Lime Lightning: 2 MILLION VOLTS!" 44 00:02:07.111 --> 00:02:12.116 ​That sounds quite clickbaity ​...but maybe not possible. 45 00:02:12.116 --> 00:02:17.621 ​There's one that just says ​"24K Gold Plated Electrical Banana Award". 46 00:02:17.621 --> 00:02:21.125 ​I don't know what that is. 47 00:02:21.125 --> 00:02:23.127 ​That sounds like a good idea. 48 00:02:23.127 --> 00:02:28.132 ​It sounds like somebody tried to ​electroplate a banana with gold, 49 00:02:28.132 --> 00:02:32.636 ​and give it as an award to someone else? 50 00:02:32.636 --> 00:02:37.141 ​Not bad... these titles, I would say, ​from the clickbaity point of view, 51 00:02:37.141 --> 00:02:39.643 ​they are not actually bad, right? 52 00:02:39.643 --> 00:02:44.148 ​Let me write it down, ​it might be a good idea... 53 00:02:44.148 --> 00:02:50.154 ​"Using Tesla's Induction Motor Design ​to Test Solar Panels"? 54 00:02:50.154 --> 00:02:53.157 ​--buh? 55 00:02:53.157 --> 00:02:54.658 ​That feels like nonsense to me. 56 00:02:54.658 --> 00:02:58.929 ​That feels like a bunch of ​random words stuck together. 57 00:02:58.929 --> 00:03:04.168 ​"How To Build A ​3D Printed Plasma Speaker". 58 00:03:04.968 --> 00:03:06.670 ​...what? 59 00:03:06.670 --> 00:03:09.173 ​Let me think about... 60 00:03:09.173 --> 00:03:11.175 ​And those were the best ideas. 61 00:03:11.175 --> 00:03:15.179 ​GPT-3 wasn't great at coming up with ​things within the laws of physics. 62 00:03:15.179 --> 00:03:16.680 ​So what about for a channel ​that's more about 63 00:03:16.680 --> 00:03:19.183 ​exploring concepts and ​interesting questions 64 00:03:19.183 --> 00:03:21.185 ​rather than really ​specific science? 65 00:03:21.185 --> 00:03:24.188 ​We like to take big questions 66 00:03:24.188 --> 00:03:27.691 ​and then try and reduce them down ​to a silly little quest. 67 00:03:27.691 --> 00:03:31.195 ​That's Sabrina Cruz, one of the team ​behind the project Answer in Progress. 68 00:03:31.195 --> 00:03:35.199 ​Now I just need to get this ​steaming pile of garbage I call my code 69 00:03:35.199 --> 00:03:36.700 ​to actually function(!) 70 00:03:36.700 --> 00:03:41.705 ​...and her pitch sounded like ​exactly what GPT-3 should be good at. 71 00:03:41.705 --> 00:03:45.209 ​It nailed your video title style. 72 00:03:45.209 --> 00:03:46.210 ​Ooh! 73 00:03:46.210 --> 00:03:49.713 ​"why buying pets is a scam". 74 00:03:49.713 --> 00:03:51.715 ​Oh my god! 75 00:03:51.715 --> 00:03:55.719 ​Well, looks like I have a ​new video idea to work on. 76 00:03:55.719 --> 00:03:58.722 ​It definitely matches the ​correct level of clickbaitiness 77 00:03:58.722 --> 00:04:01.225 ​that we go in with, ​where it's just like, 78 00:04:01.225 --> 00:04:04.228 ​hey, I have a very bold hypothesis. 79 00:04:04.228 --> 00:04:06.730 ​Let me do something that ​barely argues it! 80 00:04:06.730 --> 00:04:10.267 ​All right, well, it was downhill ​from there, 'cos it was... 81 00:04:10.267 --> 00:04:11.268 ​Oh no. 82 00:04:11.268 --> 00:04:15.239 ​"how to destroy your career with 1 email". 83 00:04:15.239 --> 00:04:18.242 ​I mean, isn't that relatable content(?) 84 00:04:18.242 --> 00:04:22.246 ​Aw geez. ​Do you remember that one... 85 00:04:22.246 --> 00:04:25.249 ​Google did a prank one year for April Fools? 86 00:04:25.249 --> 00:04:26.250 ​Oh my god. 87 00:04:26.250 --> 00:04:27.751 ​With the minion emojis? 88 00:04:27.751 --> 00:04:30.254 ​Now that's an email that ​could ruin your career. 89 00:04:30.254 --> 00:04:32.756 ​Oh my god. Right! 90 00:04:32.756 --> 00:04:35.259 ​So what we've just ​done there is workshopped a story. 91 00:04:35.259 --> 00:04:36.260 ​Boom. 92 00:04:36.260 --> 00:04:40.764 ​Okay, I think it'd be optimistic to say ​either of those could be a full video, 93 00:04:40.764 --> 00:04:43.767 ​but there's enough there that it might ​spark some bigger idea. 94 00:04:43.767 --> 00:04:45.769 ​"how to fix your posture". 95 00:04:45.769 --> 00:04:46.770 ​...sorry. 96 00:04:46.770 --> 00:04:51.775 ​Which, incidentally, yeah, I'm leaning ​over a laptop at the minute... 97 00:04:51.775 --> 00:04:54.278 ​No, that's literally the video ​I'm working on right now. 98 00:04:54.278 --> 00:04:55.279 ​No kidding! 99 00:04:55.279 --> 00:04:56.280 ​Yes it is. 100 00:04:56.280 --> 00:04:59.783 ​I'm working on a video ​about why my back hurts so badly! 101 00:04:59.783 --> 00:05:03.287 ​Wow. Okay. Um... 102 00:05:03.287 --> 00:05:07.557 ​The singularity is going to happen ​like next week, I assume. 103 00:05:07.557 --> 00:05:09.793 ​That sounds like a success, 104 00:05:09.793 --> 00:05:12.296 ​but there's a bit of... almost ​cold reading going on here? 105 00:05:12.296 --> 00:05:15.299 ​Because the AI sent out a hundred ideas, ​I picked the best ones, 106 00:05:15.299 --> 00:05:19.803 ​and then Sabrina connected one of those ​to a better idea she already had. 107 00:05:19.803 --> 00:05:22.806 ​Even the title she gave, ​"why my back hurts so badly", 108 00:05:22.806 --> 00:05:25.809 ​is way better than anything ​the AI came up with. 109 00:05:25.809 --> 00:05:29.813 ​So the best of the best ​were plausible video ideas. 110 00:05:29.813 --> 00:05:33.817 ​But I did show Sabrina more of the full list, ​and, well, most of them weren't winners. 111 00:05:33.817 --> 00:05:37.321 ​These are all formatted so that ​they would feel like 112 00:05:37.321 --> 00:05:42.826 ​it would at least come across the table ​when we're thinking about titles, 113 00:05:42.826 --> 00:05:45.829 ​but a vast majority are just, ​like, that sounds good. 114 00:05:45.829 --> 00:05:47.331 ​Let me think about it for a second... 115 00:05:47.331 --> 00:05:48.332 ​Right. 116 00:05:48.332 --> 00:05:51.335 ​Wait, what does that even mean? 117 00:05:51.335 --> 00:05:56.840 ​I think they've really captured the subtle ​negativity that just overflows my channel. 118 00:05:56.840 --> 00:06:00.110 ​Maybe score half a victory for GPT-3 there. 119 00:06:00.110 --> 00:06:03.347 ​What about if we go from ​factual exploration to opinion? 120 00:06:03.347 --> 00:06:07.851 ​I have a big Google Document of ​videos I would like to do, 121 00:06:07.851 --> 00:06:11.855 ​once the current ones ​that'll take forever are done. 122 00:06:11.855 --> 00:06:14.358 ​And that list always gets longer ​and none of them get done, 123 00:06:14.358 --> 00:06:18.662 ​cos I'm always like, oh but this one should, ​now I've had a new one! 124 00:06:18.662 --> 00:06:21.865 ​So, one day I'd like to get to that list, 125 00:06:21.865 --> 00:06:23.867 ​but it's just continuously growing. 126 00:06:23.867 --> 00:06:26.370 ​Harry Brewis is better known as hbomberguy, 127 00:06:26.370 --> 00:06:27.871 ​and he's one of very few ​video essayists I know 128 00:06:27.871 --> 00:06:30.374 ​who can somehow keep my interest long enough 129 00:06:30.374 --> 00:06:34.878 ​that I'll watch a feature-length piece ​about a show I've never even seen. 130 00:06:34.878 --> 00:06:38.382 ​But when your release schedule is ​a two-hour video every few months, 131 00:06:38.382 --> 00:06:41.885 ​idea generation isn't really ​your production bottleneck. 132 00:06:41.885 --> 00:06:45.389 ​But I wanted to see if the system could ​at least come up with some suggestions. 133 00:06:45.389 --> 00:06:47.891 ​It got obsessed with video games. 134 00:06:47.891 --> 00:06:49.393 ​Well, yeah, that's gonna happen(!) 135 00:06:49.393 --> 00:06:53.397 ​So I switched which videos of yours ​it was trying to riff off, 136 00:06:53.397 --> 00:06:57.901 ​and if there was one word even remotely ​about video games in there, 137 00:06:57.901 --> 00:06:59.903 ​everything was about video games. 138 00:06:59.903 --> 00:07:03.407 ​It's the only thing that it can ​make sense of out of all of this! 139 00:07:03.407 --> 00:07:07.411 ​"The Last Of Us Is Disappointing, And Here's ​Why". 140 00:07:07.411 --> 00:07:09.413 ​That's really good! ​Oh my goodness. 141 00:07:09.413 --> 00:07:14.418 ​That's actually the perfect video title ​for a video about The Last of Us, 142 00:07:14.418 --> 00:07:16.920 ​because people who don't like ​The Last of Us 143 00:07:16.920 --> 00:07:19.423 ​will be like, ah, finally, validation, 144 00:07:19.423 --> 00:07:21.925 ​and people who do like it will be like, ​ah, I've got to see this. 145 00:07:21.925 --> 00:07:24.428 ​I mean, it did just nick your format. 146 00:07:24.428 --> 00:07:26.430 ​It locked on to all your formats and went, 147 00:07:26.430 --> 00:07:29.433 ​"I know what to do with this, ​just sub in different names." 148 00:07:29.433 --> 00:07:30.934 ​I should have said something like, 149 00:07:30.934 --> 00:07:33.937 ​"ah, that's pointlessly contrarian, ​who would name a video that?" 150 00:07:33.937 --> 00:07:38.442 ​Oh god, I need to save that. ​That might come in handy! 151 00:07:38.442 --> 00:07:42.212 ​We also have "Mass Effect 3 Was A Mistake, ​and Here's Why". 152 00:07:42.212 --> 00:07:47.451 ​...I mean, not correct, but close, yeah. 153 00:07:47.451 --> 00:07:50.954 ​That's what I was thinking, because ​obviously the opinion has to match up with... 154 00:07:50.954 --> 00:07:54.458 ​It has to be contrarian enough! 155 00:07:54.458 --> 00:07:56.460 ​One that came from nowhere: 156 00:07:56.460 --> 00:08:03.967 ​"How Bethesda Broke Me, and Here's Why I'll ​Keep Buying Their Games Anyway" 157 00:08:03.967 --> 00:08:04.968 ​That's so specific! 158 00:08:04.968 --> 00:08:07.971 ​I wonder why ​it got that from my videos? 159 00:08:07.971 --> 00:08:09.473 ​That's really good. 160 00:08:09.473 --> 00:08:12.476 ​This one I don't understand? You might: 161 00:08:12.476 --> 00:08:17.481 ​"On Quantic Dream, David Cage, ​and the Uncanny Valley." 162 00:08:17.481 --> 00:08:18.482 ​That's really good, actually. 163 00:08:18.482 --> 00:08:19.483 ​That's incredible. 164 00:08:19.483 --> 00:08:23.987 ​Quantic Dream are... they're basically trying ​to be films instead of games in a lot of ways. 165 00:08:23.987 --> 00:08:27.491 ​There's a lot of advanced motion capture ​stuff and facial animation. 166 00:08:27.491 --> 00:08:29.493 ​So there is a lot of ​uncanny valley stuff in there. 167 00:08:29.493 --> 00:08:31.495 ​That's just a straight-up good video title. 168 00:08:31.495 --> 00:08:32.496 ​Okay! 169 00:08:32.496 --> 00:08:34.998 ​I was going to jokingly say ​I should make one of these, 170 00:08:34.998 --> 00:08:37.501 ​but that's probably the one ​I would make so far. 171 00:08:37.501 --> 00:08:41.004 ​A lot of my stuff is just whatever I thought ​would make for an interesting concept, 172 00:08:41.004 --> 00:08:46.009 ​or I would find challenging to explain, ​I think that's really fun to try and do. 173 00:08:46.009 --> 00:08:50.013 ​If I wanted to make a video along those lines, ​that would be the title for it. 174 00:08:50.013 --> 00:08:52.516 ​But I probably wouldn't make ​those specific videos. 175 00:08:52.516 --> 00:08:54.518 ​Some of those ideas were actually good. 176 00:08:54.518 --> 00:08:57.320 ​Like, at least two of those I could make. 177 00:08:57.320 --> 00:09:01.024 ​And now I kind of want to, just so... 178 00:09:01.024 --> 00:09:03.527 ​A video essayist not only ​has to stand behind their ideas, 179 00:09:03.527 --> 00:09:06.830 ​but also put in exhaustive research ​and be passionate about a subject. 180 00:09:06.830 --> 00:09:08.031 ​And while it could be useful ​to have a system 181 00:09:08.031 --> 00:09:11.535 ​that'd take a video outline and ​have it autocomplete in a certain style 182 00:09:11.535 --> 00:09:14.037 ​and make the arguments you want it to, 183 00:09:14.037 --> 00:09:16.039 ​text generation just isn't there yet. 184 00:09:16.039 --> 00:09:19.543 ​Which leaves us with channels ​that are about experiences. 185 00:09:19.543 --> 00:09:22.546 ​Folks who aren't creating factual ​or documentary or opinion videos, 186 00:09:22.546 --> 00:09:24.548 ​just plain ol' regular entertainment. 187 00:09:24.548 --> 00:09:27.551 ​Now, I did try to get in touch ​with a few channels who do 188 00:09:27.551 --> 00:09:31.555 ​beauty, or makeup challenges, ​but I didn't get any replies. 189 00:09:31.555 --> 00:09:33.056 ​They're in such different worlds 190 00:09:33.056 --> 00:09:35.058 ​that I just don't have ​any contacts in common. 191 00:09:35.058 --> 00:09:37.561 ​I did give the AI a list ​of makeup challenges, 192 00:09:37.561 --> 00:09:40.564 ​and if anyone wants to try ​the No Water Makeup Challenge, 193 00:09:40.564 --> 00:09:42.365 ​the Curtain Makeup Challenge, 194 00:09:42.365 --> 00:09:45.569 ​or the If I Only Had More Time ​Makeup Challenge, go right ahead, 195 00:09:45.569 --> 00:09:47.571 ​I just don't know what those are. 196 00:09:47.571 --> 00:09:50.073 ​But there was one more ​person I talked to. 197 00:09:50.073 --> 00:09:53.376 ​And he's an expert on getting ​the title and the idea right. 198 00:09:53.376 --> 00:09:56.580 ​A perfect idea is an idea ​that is clickable 199 00:09:56.580 --> 00:09:59.849 ​but also highly highly entertaining ​throughout the entire video. 200 00:09:59.849 --> 00:10:04.588 ​And a lot of people are either good at ​picking great videos or great clickbait, 201 00:10:04.588 --> 00:10:07.090 ​and very very very few people ​can pick both. 202 00:10:07.090 --> 00:10:08.091 ​It's very hard. 203 00:10:08.091 --> 00:10:12.596 ​Because some things that are ​very clickable might be a moment. 204 00:10:12.596 --> 00:10:16.366 ​Like, getting punched by the ​strongest man in the world, right? 205 00:10:16.366 --> 00:10:19.102 ​That's an interesting video ​but that's only ten seconds. 206 00:10:19.102 --> 00:10:20.370 ​It's not a 15-minute video. 207 00:10:20.370 --> 00:10:23.607 ​And if you try to make it a 15-minute ​video, then it's a bad idea. 208 00:10:23.607 --> 00:10:25.408 ​Not like, you just take things ​and stretch it out, 209 00:10:25.408 --> 00:10:27.611 ​it has to be a really entertaining video ​at that length, 210 00:10:27.611 --> 00:10:29.112 ​while still being clickable. 211 00:10:29.112 --> 00:10:30.113 ​That is the perfect idea. 212 00:10:30.113 --> 00:10:32.616 ​Jimmy Donaldson is better known as MrBeast, 213 00:10:32.616 --> 00:10:35.619 ​and fifty million subscribers ​watch his channel that's a mixture 214 00:10:35.619 --> 00:10:38.622 ​of ridiculous stunts and ​hyperactive philanthropy. 215 00:10:38.622 --> 00:10:42.125 ​As we talked, it was really clear ​just how much thought goes into 216 00:10:42.125 --> 00:10:46.129 ​the titles, the ideas, ​and every frame of every video. 217 00:10:46.129 --> 00:10:48.131 ​The AI... does not have that context. 218 00:10:48.131 --> 00:10:50.133 ​It's just putting words together. 219 00:10:50.133 --> 00:10:52.636 ​The very first thing it came out with was: 220 00:10:52.636 --> 00:10:55.639 ​"I Paid People To Take Selfies With Me". 221 00:10:55.639 --> 00:10:59.142 ​Yeah, I can see where it would get ​that inspiration from, 222 00:10:59.142 --> 00:11:02.145 ​but there's no, like, 'oomph'. ​What's the virality? 223 00:11:02.145 --> 00:11:03.947 ​"I Bought A Haunted House". 224 00:11:03.947 --> 00:11:06.916 ​I can see where it got it from ​and the inspiration, 225 00:11:06.916 --> 00:11:09.152 ​but what it's missing is a twist. 226 00:11:09.152 --> 00:11:10.153 ​Back to the selfie one, right? 227 00:11:10.153 --> 00:11:14.157 ​If it was like "I paid people to ​take selfies with me", 228 00:11:14.157 --> 00:11:17.160 ​and I looked like Shrek ​but I had like a knife in my hand, 229 00:11:17.160 --> 00:11:18.662 ​then it's kind of interesting, 230 00:11:18.662 --> 00:11:20.664 ​but obviously that's not ​what it's thinking. 231 00:11:20.664 --> 00:11:23.166 ​Or, with the haunted house, 232 00:11:23.166 --> 00:11:27.671 ​I bought a haunted house and offered people ​a million dollars to play in it. 233 00:11:27.671 --> 00:11:30.674 ​Yeah, iIt feels like there should ​be an "and" in there somehwere. 234 00:11:30.674 --> 00:11:36.179 ​It's missing that little twist, where ​you apply that, and boom, now it's viral. 235 00:11:36.179 --> 00:11:39.182 ​It also had one which feels like ​an entire series on your channel, 236 00:11:39.182 --> 00:11:43.687 ​just put in a blender so it's just mush ​and no-one wants to watch it, which is: 237 00:11:43.687 --> 00:11:46.189 ​"I Paid People To Do ​Weird Things On Camera". 238 00:11:46.189 --> 00:11:47.691 ​I'm like, that's... 239 00:11:47.691 --> 00:11:49.693 ​That's my channel in a nutshell! 240 00:11:49.693 --> 00:11:50.694 ​It's exposing me! 241 00:11:50.694 --> 00:11:55.198 ​A lot of GPT-3's ideas felt like someone ​trying to imitate the MrBeast channel 242 00:11:55.198 --> 00:11:58.702 ​without really understanding ​what humans like to watch. 243 00:11:58.702 --> 00:12:00.003 ​So while it didn't go well, 244 00:12:00.003 --> 00:12:03.006 ​it did perform a little bit ​better than I expected. 245 00:12:03.006 --> 00:12:05.475 ​And I think it was Jimmy ​that gave the best summary: 246 00:12:05.475 --> 00:12:07.210 ​This could be very very effective 247 00:12:07.210 --> 00:12:09.979 ​in giving you something close ​to what you'd want, 248 00:12:09.979 --> 00:12:13.717 ​and then you just use that as ​inspiration to derive something good. 249 00:12:13.717 --> 00:12:15.218 ​AI works best as a co-pilot. 250 00:12:15.218 --> 00:12:16.720 ​At least, for now. 251 00:12:16.720 --> 00:12:18.722 ​Because one of the clichés about ​artifical intelligence is that 252 00:12:18.722 --> 00:12:22.225 ​as soon as it can do something, ​that doesn't count as AI any more. 253 00:12:22.225 --> 00:12:24.227 ​Playing chess? ​It's just brute-forcing numbers. 254 00:12:24.227 --> 00:12:25.228 ​Categorising images? 255 00:12:25.228 --> 00:12:28.732 ​There's no thought there, ​it's just training a model. 256 00:12:28.732 --> 00:12:32.235 ​Coming up with a list of plausible ideas, ​with an okay but not great hit-rate? 257 00:12:32.235 --> 00:12:36.239 ​That's something that would have been ​impossible just a couple of years ago. 258 00:12:36.239 --> 00:12:37.240 ​But now? 259 00:12:37.240 --> 00:12:38.742 ​It's just putting ​some words together. 260 00:12:38.742 --> 00:12:41.511 ​In a few years' time, ​who knows what else 261 00:12:41.511 --> 00:12:44.247 ​it's going to be easy to dismiss ​in the same way? 262 00:12:44.247 --> 00:12:45.248 ​Let me just stop tape... 263 00:12:45.248 --> 00:12:47.751 ​Wait, here, let me give you a little ​clip to go at the end of the video? 264 00:12:47.751 --> 00:12:48.752 ​Yeah, yeah, go for it. 265 00:12:48.752 --> 00:12:51.254 ​If you enjoyed this video ​subscribe to Tom Scott! 266 00:12:51.254 --> 00:12:52.756 ​He's too nice to say it. 267 00:12:52.756 --> 00:12:53.757 ​Hit that sub button. 268 00:12:53.757 --> 00:12:54.758 ​It'll make him really happy. 269 00:12:54.758 --> 00:12:55.759 ​Right there, tap it. 270 00:12:55.759 --> 00:12:56.760 ​All right, now you can go. 271 00:12:56.760 --> 00:13:00.930 ​I have literally never said that ​in my life, thank you!