8/22/2025 - The Impact of AI on Creative Industries and the Future of Work

Description

Chapters

Transcript

0:01 Are you ready? Yes, I'm ready
0:06 to fall in love.
0:12 [Music]
0:21 Boys were singing shinglings
0:24 the summer that we met.
0:27 [Music]
0:30 You were 10 and 17.
0:33 How could I forget?
0:37 >> I was shy to kiss you while the whole
0:41 world could see.
0:44 So shingling said everything to me.
0:51 The way we danced was not a dance, but
0:55 more a long embrace.
0:58 held on to each other and floated there
1:03 in space.
1:05 I was shy to kiss you while the whole
1:08 wide world could see.
1:11 So shingling said everything to me.
1:19 And all the poor old folks, they thought
1:23 we lost our minds.
1:26 They could not make heads or tails with
1:29 the young folks funny rhymes.
1:33 >> You and I knew all the words and we
1:35 always sang along
1:38 to old jam dong ding jaming
1:45 dang dong.
1:49 [Music]
1:54 [Applause]
1:56 [Music]
2:04 Happy Friday night, date night, people.
2:05 I don't know, those of you on YouTube,
2:07 you might notice I got myself a little
2:09 microphone. I'm a professional.
2:11 I'm a professional. Now, look at this
2:14 thing.
2:16 Now, if it sounds like crap, that's cuz
2:17 I don't know how to set it up.
2:24 Ah, Cindy [ __ ] is first. Cindy [ __ ]
2:27 look at my microphone. Hello. We can do
2:30 ASMR now.
2:32 Hi.
2:36 I got a doggy.
2:39 I think I'd be pretty good at that. You
2:41 know what I'm saying?
2:43 [Music]
3:01 Why you Why you getting so ramyammy? Why
3:03 you so ramy, champ?
3:06 [Music]
3:08 If you haven't, uh it's Friday night
3:10 date night. Go get your nachos. Put them
3:12 in the air fryer. Little uh little three
3:16 or four cheese combo right on top of
3:18 there with some jalapenos out of the
3:20 can. You know, get your hot pocket
3:23 heating up in that dirty microwave.
3:25 Let's get this show on the road. I have
3:27 I have some Kirkland sparkling water cuz
3:30 I know how to live.
3:33 Pinkies out cuz we're classy.
3:37 That's fantastic.
3:40 It's It's going to make me a little
3:42 gassy. What you got going? Oh, Kirkland.
3:44 Nice.
3:45 Producer brand.
3:47 >> Unofficial sponsor.
3:49 >> Yeah, exactly.
3:52 That's That's awesome.
3:57 Oh, man. So, yeah. So, I'm a little
4:00 discombobulated. I'm not used to having
4:02 a microphone in my face, but this will
4:04 be good for comedy. You know, lean in
4:06 for a comedic effect. You know, I don't
4:08 really know about this. This whole AI
4:10 learning lab thing. Who is this? Kyle
4:12 Shannon. I mean that, you know, right?
4:15 It's going to be good. It's going to be
4:17 fantastic.
4:19 All right.
4:20 [Music]
4:39 [Applause]
4:40 [Music]
4:48 [Applause]
4:49 [Music]
5:05 [Applause]
5:07 [Music]
5:27 [Music]
5:38 All right. What do we got going here?
5:40 Let me just let me get things lined up.
5:43 Everything's different. Everything's
5:45 different. You know, it's funny with
5:47 ADD.
5:50 Like when you have ADD, you kind of live
5:54 for the chaos, right? Except for [ __ ]
5:57 like how my setup is. Like for example,
6:00 let me show you something. This is
6:01 [ __ ] hilarious.
6:03 Okay, so for two years,
6:09 [Laughter]
6:13 for two years,
6:15 this was the stand that I put my my LED
6:19 light on. It was a cardboard box and
6:22 there's nothing in it. So, it's all
6:23 light. It's completely unstable,
6:26 right? And my little LED light sat here.
6:29 Now, sitting right next to this
6:32 was a little microphone tripod I've had
6:35 for like 10 years.
6:40 So, for two and a half years, this was
6:42 my light stand. And now
6:47 look,
6:49 I'm professional.
6:50 [Laughter]
6:53 ADD is [ __ ] weird, man. You got some
6:56 I must have some OCD mixed in there.
7:03 Oh my god.
7:06 It's all comedy all the time here at the
7:08 AI Learning Lab.
7:10 Um
7:13 all right
7:15 so I was playing around with GPT5 how do
7:18 you share the output you can so there's
7:20 so in chat GPT
7:27 with any chat that you have
7:31 go grab a chat I did
7:34 casual conversation
7:37 marketing email for far this was
7:39 actually really impressive I I'll show
7:40 you this in a second here.
7:44 Um, so so right up here, this this upper
7:47 right hand corner.
7:50 Oh, wait. Share your tabs. Share your
7:52 tabs, Kyle. Be a professional. Uhoh.
7:55 What did I just do? Uh oh. What did I
7:57 do? Hang on. Hang on, people. We just
8:01 had a malfunction.
8:08 Calm down. Everybody calm down.
8:10 Everybody calm down.
8:13 Can you still hear me in the little
8:15 microphone here?
8:18 Okay. I just lost my I just lost my
8:21 monitor. We're going to be okay, people.
8:24 It's going to be okay.
8:32 Oh my god. I I just I should not be left
8:36 to my own devices for any kind of
8:38 professional situation. Okay. Anyway,
8:41 what I was saying before I was so rudely
8:42 interrupted with technology in the upper
8:44 right hand corner of Yan
8:49 here. Let me move this a little bit.
8:51 We'll do it like right there. Right here
8:54 it says share. You can share that. Now
8:56 you the thing to be careful with is if
8:59 you share your chat
9:02 um
9:03 it is not at all clear to me if you're
9:06 just sharing it with the people you're
9:07 sharing it with or if it's actually
9:09 available for everyone to go discover
9:12 it. So just be careful with it. All
9:14 right. This is off to a good start.
9:18 Tik Tok pin. 370 chats were published to
9:21 a public URL. Yeah. Oh, 370,000
9:25 370,000 chats were published to a public
9:28 URL today by XAI. Yeah. So that share
9:32 button with XAI today.
9:35 Yeah. Shared shar shared them all
9:37 publicly. And then I heard a similar
9:39 sort of thing with with chat GPT. So So
9:42 there you go. Um All right.
9:47 [Music]
9:49 M.
9:54 Oh, lordy lordy lordy.
9:58 Since we're here,
10:01 so I discovered some all sorts of weird
10:04 things today. The difference between
10:06 teams. So, I've got a team's account for
10:10 my content evolution role.
10:15 And in the Teams account,
10:19 you've got instant chat GPT5,
10:24 you've got thinking, you've got Pro, and
10:27 then you have your legacy models.
10:30 Now, the legacy models don't show
10:33 anything beyond GPT40. Let me just make
10:36 sure that I don't have them that I have
10:38 it turned on. Yeah, I've got show
10:40 additional models turned on and it only
10:43 shows
10:45 GPT40, but it does give me pro. Okay, if
10:51 I go to my personal account,
10:56 I also up here have Thinking Mini, which
10:59 is actually my favorite model lately. I
11:02 do not have Pro, which makes sense. But
11:04 then my legacy models, I have four of
11:07 them. I've got 4041, 03, and 04 mini. I
11:10 don't care about those other ones, but
11:12 it's weird,
11:15 you know. I wonder actually now that I
11:18 think about it.
11:19 Sometimes chat GPT
11:23 turning things on and off.
11:28 Show additional models. Turn it off.
11:33 I go here. All right. Then I got 40. Let
11:36 me come back in here. Turn it back on.
11:41 I bet they show up now.
11:47 Nope. So, the legacy models don't show
11:50 up in Teams and and Thinking Mini does
11:53 not show up in Teams, but Pro does.
11:56 So, I don't know what the hell. All
11:58 right.
12:00 So, go figure. Um,
12:04 one of the things that I talked about
12:06 today on office hours, by the way, we
12:07 had a really fun office hours today
12:09 because Jim Ross, who's got his own
12:12 community of self- storage entrepreneurs
12:16 who he's teaching how to use AI, he
12:18 said, "Hey, my meeting ends right before
12:20 yours starts. Do you mind if I bring a
12:21 bunch of people over?" So, he brought
12:23 over a bunch of people and we had a a
12:26 really good a really good uh office
12:28 hours today.
12:31 But in all three of my my I have three
12:34 AI meetings in a row on Fridays. I've
12:36 got my content evolution collab, I've
12:38 got office hours, and then I've got the
12:41 AI mastermind, the AI salon mastermind.
12:44 In all three of those conversations,
12:48 I kind of, you know, tested the waters
12:51 for like who's got GPT5 figured out? And
12:55 no one. No one. Everyone. Everyone is
13:00 sort of feeling a little
13:01 discombobulated.
13:03 Um
13:05 I I have not recombobulated yet around
13:07 chat GBT5. It just feels awkward and I
13:10 don't quite know how it works and I
13:12 don't quite know which model to use
13:13 when. Um and so if you're feeling if
13:17 you're feeling a little out of sorts
13:19 because of chat GPT5,
13:21 don't forget we've had GPT4
13:25 since March of 2023.
13:29 And so it's been more than two years
13:31 where we've had we've gotten acclimated
13:33 to how that model worked. So if this new
13:35 one feels weird, that should be
13:37 expected. Now GPT6, I think, is going to
13:41 come out before the end of the year.
13:42 Grock 5 is going to come out before the
13:45 end of the year. Gro 6 might come out
13:48 before the end of the year. So depending
13:50 on which tool you're using, you know,
13:52 Gemini keeps releasing new stuff. You're
13:55 gonna have We're gonna have a lot of
13:58 adapting to get used to. A lot of
13:59 adapting to do.
14:01 Vicki, fun office hours indeed. Indeed.
14:06 Oh, yours went even more off the rails
14:07 today. Your chat GBT. Yeah, it's just
14:10 weird. It's It's weird. The other thing
14:12 is um one of the things that Altman said
14:18 I don't know a week or two ago is that
14:20 their GPU
14:23 usage is like they do not have enough
14:26 GPUs to handle they have 700 million
14:30 users on a weekly basis
14:33 and the stuff that we're playing with if
14:35 you've noticed um image generation has
14:37 gotten really slow. They don't show the
14:38 preview anymore. Um, I I think they have
14:41 things quite throttled down right now.
14:43 So, if you're seeing inconsistencies and
14:46 weirdness, um, it's because they're
14:49 they're they're overloaded.
14:52 Um, Vicki, it was awful. It kept pulling
14:55 conversations that had nothing to do
14:56 with the project folder it was in. Oh,
15:00 yeah. Yeah. Projects are weird right
15:02 now, too. The if if you don't know about
15:05 projects, they're incredibly powerful.
15:07 You can basically group a bunch of chats
15:09 into a project and then it'll just talk
15:11 about those chats. But chat GPT also has
15:14 access to all of your other
15:15 conversations. So it it doesn't seem to
15:18 have good boundary control.
15:22 You you know like that ex-girlfriend you
15:24 had didn't have good boundary control.
15:27 That's chat GPT projects.
15:31 Rationing of GPU clock cycles. I
15:33 something like that. you know, they're
15:35 either rationing the the GPUs or just I
15:39 I think what's probably more likely is
15:43 they've got
15:46 sort of levels of thinking that they can
15:48 move up and down. So, they're probably
15:51 not doing it on a GPU byGPU basis, but
15:54 they're probably doing it on, you know,
15:56 kind of features, what the thing can
15:57 actually do. So, my guess is is that
16:00 chat GPT is a lot dumber right now than
16:02 it'll feel when they have enough GPUs.
16:05 Um, but I don't know. I don't know.
16:08 That's just chat TMZ.
16:11 Chat TMZ.
16:13 All right. What else you got going
16:15 there, people?
16:18 Oh, yeah. It's on Twitter if anyone
16:20 wants to sympathize. Oh, you you posted
16:22 it on Twitter. That's good. That's good.
16:27 Um
16:32 the uh the guy last night I showed this
16:35 guy Fernando who's a musician in LA and
16:38 he's um
16:40 he's been playing with Sununo, right? He
16:43 puts his songs in it. Well, he's put out
16:44 a bunch more videos and his um his LA
16:50 Moon I think was the name of the song.
16:52 the the the original video that he did,
16:55 that video has gone viral. Um he's
16:58 gotten a whole bunch of hate from AI
17:01 haters that are like, "You're evil.
17:02 You're in bed with the devil doing this
17:04 AI stuff." And he's just like, "Hey,
17:07 man. It's the future." Like, this is
17:08 really cool technology. And he keeps
17:10 being impressed by it. Um so we can go
17:14 look at a couple more of those if you
17:15 want to tonight.
17:28 Brian Whitney, someone in the building.
17:32 Someone is building a huge Dana Center
17:33 here in northeastern Pennsylvania.
17:36 That's interesting. Techie Becky, I love
17:38 getting my Friday night entertainment
17:39 right here. That's right. You don't Let
17:41 me tell you something, people. This is a
17:43 public service announcement right here.
17:45 You don't You don't need Netflix. You
17:47 don't need Paramount Plus. You don't
17:49 need Hulu.
17:51 Hulu, Hulu.
17:54 You don't need any of that. You come to
17:55 the AI learning lab. You get you some
17:58 learning. You get you some
17:59 entertainment. You got it all right
18:01 here.
18:04 Good lord, people. You think you got to
18:07 pay those Hollywood executives their
18:10 high bonus salaries to create crap,
18:13 import movies from Sweden.
18:17 Uh-uh. You just come right here. I got a
18:20 microphone now. You don't think this is
18:22 professional?
18:24 You think this is just a crappy home
18:26 office? You don't even know the sound
18:28 stage we got here. It's astounding. This
18:31 is all just a curved LED screen. You You
18:34 think those are actual shells?
18:39 All right.
18:41 I don't think that was funny, but you
18:42 know. Yeah. Thank you for the uh
18:44 whatever that was on my head.
18:52 API is application protocol interface.
18:55 It is an AP API basically allows one one
18:59 piece of software to talk to another
19:01 piece of software.
19:03 So here a good way to describe an API
19:06 is let's say I'm creating a mobile app
19:10 and in that mobile app I want to have a
19:13 map function right so I can you know
19:16 people can find my business.
19:20 Well, I could program an entire map
19:23 application and launch my own satellites
19:27 and, you know, spend the next 3 years, 5
19:30 years creating a solid map application.
19:33 Or I could say, "Huh, Google's already
19:37 done this. Why don't I just use their
19:38 API?" And then basically what happens is
19:41 you get a special key from Google and
19:44 you put it in your app and then you just
19:46 make a little box and you say, "Stick
19:48 Google's map here." and then it does it.
19:51 So that's that's kind of the that's kind
19:54 of the thing. It it allows software
19:56 developers to share what they've created
19:58 and make money on it, right? Every time
20:00 you use the little Google Map
20:02 application, they get, you know, a
20:03 hundredth of a penny or whatever it is
20:06 and over enough of those, they make a
20:08 fortune.
20:10 [Music]
20:13 Uh yeah, it's in the software stack. Um,
20:16 I finally got so fast frustrated with
20:19 chat's memory I deleted it all and it
20:23 still doesn't forget.
20:25 Huh.
20:27 Well, wait. But Tom, so in order to
20:30 delete it all. Okay, so there's a couple
20:32 of things. Let's go
20:34 let's go look at memory.
20:37 So if you wanted to if you wanted to
20:41 really get rid of your chat memory,
20:43 there's a couple of things you'd have to
20:45 do. So, let's go into settings.
20:50 Turtle image generation didn't work. Oh,
20:52 that's interesting. Um,
20:56 let's go into data controls. So,
21:03 improve the model for everyone. So, this
21:06 means it's sharing my data.
21:10 Interesting.
21:12 So I just turned Oh, I'm not sharing, am
21:14 I?
21:17 Yeah, I am. Wait.
21:20 Yeah, I am. Okay. Um
21:26 Okay. I just turned off model
21:28 improvement. So model improvement. This
21:30 sends data to OpenAI. I just turned that
21:33 off. I had that on and I'd turned that
21:35 off before. So that's that's one that
21:37 they probably when they updated to five,
21:39 they're like, "Let's just reset
21:40 everyone's preferences."
21:42 Remove browser data. Let the agent mode
21:46 remote browser reuse cookies between
21:49 sessions.
21:50 All right, that's fine. I'm fine with
21:52 that. But if you want higher security,
21:56 okay, share links,
21:58 archive chats.
22:03 So the date the debt management guidance
22:08 I guess you can archive okay you can
22:10 archive all chats
22:13 you can export data
22:18 security multiffactor authentication
22:22 secure signin with chat and then let's
22:25 go to personalization so in
22:27 personalization
22:30 you've got reference chat history which
22:32 I had turned off.
22:35 You've got reference record history.
22:37 What's that? Oh, recording. Record
22:40 history. Let chat GBT reference all
22:44 previous recording transcripts.
22:47 I guess that's advanced voice.
22:52 And I had I had reference chat history
22:54 turned off. I'm going to turn that back
22:55 on. So, you could just turn off both of
22:58 these without deleting stuff. The other
23:00 thing you can do is manage your
23:02 memories. So these are the explicit
23:04 memory. So there's two different kinds
23:06 of memory in chat GPT.
23:08 There's the explicit memory where you
23:10 say, "I want you to remember this." And
23:12 then there's it remembering all of your
23:14 chat histories. And you can turn those
23:16 on and off on personalization. So So
23:19 Tom, you might want to just try turning
23:21 off both of those and see if that helps.
23:25 Tik Tok gifts from Lord Digital Gods.
23:27 Thank you, sir. Thank you, kind sir.
23:29 always appreciated.
23:32 Um,
23:34 [Music]
23:41 so for you YouTubers, does this audio
23:43 sound better? Does it sound too hot? Is
23:45 the mic too hot? Is my voice annoying?
23:51 You're like, we were better when we
23:52 couldn't hear you.
23:58 I sound a bit tiny. I bet that's because
24:00 that's too loud. Test, test, test. How's
24:03 that? Is that better? Testing, testing,
24:06 test, test, test, test, test, test.
24:09 Audio sounds good. Yes. Okay, so I was
24:11 clipping. Hang on. Let me actually go
24:14 over to this software. Hello. Test test
24:17 test. Okay, now I'm not clipping.
24:20 All right, there we go. All right. I'm a
24:22 professional.
24:27 You liked it before. Okay. All right.
24:29 There's somewhere between where I was
24:31 before and where I was now. What's up,
24:32 champion? Hang on. Hold, please.
24:47 All
24:56 right. Champy. Champy lacked a little
24:58 confidence.
25:01 He was hopping up on the bed. He was a
25:02 little little scaredy cat. He was a
25:04 scaredy dog.
25:07 Um, all right.
25:10 So, let me show you a uh
25:13 a prompt that I did today
25:16 that was really quite good
25:24 marketing email for pharma. Okay,
25:31 beauty.
25:35 That's good. All right,
25:38 let's see.
25:44 [Music]
25:46 I saw Champ's relative yesterday all
25:48 posting irregulars. That's cool.
25:52 Um,
25:53 okay. So,
25:58 so today I was having a talk with my
26:00 co-founder at StoryVine and we were
26:02 talking about marketing and one of the
26:04 things going on in the pharma industry,
26:06 which is, you know, it's going on
26:07 everywhere, but there's a lot going on
26:09 in pharma right now, is big pharma
26:12 companies are doing like big cuts.
26:14 They're doing like $3 billion
26:16 restructurings and they're, you know,
26:18 defunding,
26:20 you know, budgets for certain brands
26:22 and, you know, prioritizing their core
26:25 brands and dep prioritizing other
26:26 brands. There's all sorts of stuff going
26:28 on.
26:29 And,
26:32 you know, historically,
26:35 what we would try to do is navigate to
26:37 the companies and the brands that had
26:40 bigger budget, right? and we would try
26:41 to avoid the the chaos.
26:45 And and what hit me today was, well,
26:48 wait a minute. What if we leaned into
26:50 the fact that um that these pharma
26:53 companies are going through tough times
26:55 and we market story because Storyine's a
26:58 really efficient video creation tool.
27:01 And so what if we leaned into it and I
27:03 thought well maybe chat GPT could help
27:05 us identify like either who are the
27:08 companies or what are the trends or
27:09 things like that. So I put it in GPT5
27:14 thinking mini which is my new favorite
27:17 model and then I put it in did I do deep
27:21 research? I think I did do deep
27:22 research. Does it say? Yeah. Uh, I don't
27:27 remember if it was deep research or just
27:28 turned on the web web search, but it was
27:31 one of them, but but I wrote, I want you
27:33 to write me up a marketing email that
27:35 targets pharma. This was a bit of a a
27:39 chain of consciousness, stream of
27:40 consciousness kind of prompt. This this
27:43 isn't a very This isn't very well
27:44 written, not very well structured, but
27:48 chat GBT doesn't really seem to care
27:50 about that. You give it a bunch of
27:51 context and it figures the [ __ ] out. I
27:53 want you to write up write me up a far
27:56 pharma a marketing email that targets
27:58 pharmas that may have had recent cuts or
28:01 layoffs where cost cutting is more
28:03 obvious and I want to position storyvine
28:05 as an alternative think something like
28:08 just because budgets have been slashed
28:10 doesn't mean you need to give up video
28:12 or if your salesforce has been reduced
28:15 what if you could triple their
28:16 effectiveness with personal promotion
28:18 and then I put these subjects are not
28:20 literal suggestions but the kind of
28:22 positioning I'm looking for, right? So,
28:25 just tell it, you know, don't copy me
28:27 exactly because if you don't say that,
28:29 it'll just copy you exactly. It won't
28:31 think. And then I said, then I had this
28:34 f this further thought. Why don't you do
28:36 some research and identify the kinds of
28:38 pharma scenarios where this kind of
28:40 positioning could make sense, categorize
28:43 them, and write a marketing position for
28:46 each category with an example email per
28:49 category.
28:51 And then I said, "Go." And it thought
28:53 for 24 seconds. And this was my the
28:57 first response. This was my first
28:59 response. I I find this just astounding.
29:03 Um that's probably a little too big, but
29:06 that's a little better for TikTok. Um
29:11 so it says, "You want a crisp,
29:12 deployable email copy and strategic
29:14 positioning that leads into the reality
29:16 pharma marketers are living in right
29:18 now." And so it did some research uh and
29:21 then so so quick evidence snapshot and
29:23 and these all have citations.
29:26 Layoffs and workforce reductions across
29:28 biioarma continued into 2025 with
29:31 thousands of roles cut. Industrywide
29:33 cost pressure is being driven by drug
29:36 pricing and reimbursement debates.
29:38 Pharma is actively shifting to omni
29:40 channel and digital first HCP engagement
29:43 which that leans right into us.
29:45 personalized video and short form video
29:47 s perform significantly better than
29:49 generic content which leans right into
29:51 us. So so two of the trends that it
29:54 found are right in the pipeline of
29:56 StoryVine and then it and then here's
29:58 the categories it found. Category A
30:00 salesforce reductions commercial team
30:02 slowdown. Here's the scenario.
30:05 Here's the positioning. Turn every
30:08 remaining rep into a high lever content
30:10 engine. personalized compliant video
30:12 outreach to multip that multiplies touch
30:15 points and preserves the human
30:16 connection like this is right out of our
30:19 positioning which it might have pulled
30:20 out of our positioning.
30:22 um key benefits to the client, subject
30:25 line options, right? So, here's subject
30:27 lines and then here's an email example.
30:29 Subject preheader,
30:32 there's your email, here's your little
30:34 bullet points, there's your call to
30:36 action,
30:38 right? PS, we can spin up a 60 60-second
30:41 personalized sample video. Category B,
30:44 marketing budget squeezes. And then same
30:47 thing for that. Category C, mergers,
30:49 acquisitions, or major restructuring.
30:52 There's a whole category for that.
30:54 Category D, launches or life cycle
30:56 management. There's a whole category for
30:58 that.
30:59 Category E, regulatory or pricing
31:02 pressure,
31:04 right? And so because it's got memory,
31:06 it knows what Storyvine is. It knows
31:08 what our value prop is. And then there's
31:11 wrap up and next steps. So this was in
31:15 the three minutes I had in between
31:17 calls,
31:19 you know, I had a a conversation with my
31:21 my partner and she was like, "Oh, maybe
31:24 see if you can write up a like an
31:25 email."
31:28 And so I wrote this prompt and like
31:32 bang, 24 seconds later, or maybe it was
31:35 30 because it took six seconds to write
31:37 this.
31:41 Um, so that is an example to me where
31:44 ChatGpt 5 significantly outperformed,
31:49 you know, four. It just it just went
31:51 deeper. The writing's really good. The
31:53 thoughts really good. And that's just
31:56 the mini model. Again, my my this
31:58 thinking mini model for me is has been
32:01 pretty pretty damn good. So anyway,
32:07 thoughts, questions.
32:10 [Music]
32:18 Oh, I got to do Black Bar. That's
32:20 horrible. All right, side hustle Mimi.
32:22 Sorry. Did I ever tell you guys about
32:24 the time I checked into a hotel
32:27 in
32:29 New Mexico? One of two.
32:37 What does the agent in chat GPT do? I
32:39 haven't tried it yet. Okay, so this is
32:41 really cool.
32:43 Um, well, let me see.
32:46 Chat GPT agent isn't really cool.
32:51 It is
32:55 It's going to be really cool when it
32:57 doesn't suck.
33:02 And the only reason it sucks right now,
33:04 the thing I mentioned earlier that that
33:06 chat GPT is completely overloaded right
33:09 now. They don't have enough GPUs.
33:12 And so, um, the agent capability is
33:16 really slow. Sometimes it just craps
33:18 out.
33:19 Sometimes it doesn't finish it thing its
33:21 thing. But I I'll do a little demo here
33:24 and and we can go play with it. But um
33:27 in a nutshell what agent does there's
33:30 think of think of chat GBT kind of has
33:33 three basic modes. The first mode is
33:37 just regular chat GBT mode. So when you
33:40 flip into Oh wait two of two. Two of
33:43 two. They gave me a complimentary plate
33:45 of nachos upon checkin.
33:48 It was weird. Like I had I had luggage.
33:53 That's really funny.
33:59 Oh my god. Yeah. Some some middle
34:02 manager there at that hotel was like, "I
34:04 got an idea."
34:06 You know how what what what is the one
34:08 that gives you a cookie? You know how
34:10 they got warm chocolate chip cookies?
34:12 We're going to one up them. We're going
34:15 nachos. I'm talking jalapeno peppers,
34:18 the cheese sauce. It's going to be
34:19 awesome. How do people get it up to
34:22 their room? They just balance it on
34:23 their luggage. It'll be awesome. Okay.
34:31 All right. Here we go. Is this better?
34:34 That That's a little easier to read,
34:36 right? Yeah, that's good. Except it puts
34:39 me in the middle of everything. All
34:40 right. Whatever.
34:42 Okay. Um chat GPT agent. So before I get
34:47 in in demo agent,
34:51 um
34:53 there's there's kind of three different
34:55 modes with large language models.
34:59 There's the basic chat
35:03 single chat response. You put in a
35:05 prompt, you get out an answer. So when
35:08 you choose the instant model in chat GBT
35:12 um
35:16 let me make a new chat here
35:20 and just say um make me
35:25 a poem about hotel nachos.
35:33 You put that in and it, you know, here's
35:36 a little ode to hotel nachos and then
35:38 out out it comes, right? You put in a
35:40 thing, you you you get out an answer.
35:42 The second primary mode of chat GPT is
35:46 this thinking mode and that is you put
35:49 in a prompt and it thinks about it for a
35:52 little bit and it what it's doing is
35:54 it's actually prompting itself
35:55 internally going well the user asked me
35:58 for a poem and what should I do and so
36:00 let me let me create a new a new chat
36:03 here we'll do it in thinking midi uh
36:05 thinking mini write me a poem about
36:10 nachos. Now, this may or may not kick
36:12 off much thinking, but let's see. Oh, I
36:15 just about nacho. So, there it says
36:17 thinking. It thought for a couple of
36:19 seconds and then it wrote a poem.
36:21 So, that's not a really good example,
36:23 but that's kind of mode two. And then
36:24 mode three is this agent mode. And what
36:26 happens in agent mode is it takes that
36:29 thinking model and it basically gives it
36:32 its own virtual computer. And so it
36:36 literally creates like a screen where
36:38 where it can do a bunch of things with
36:41 tools. So it can think, it can come up
36:43 with a plan. So you tell it to go do
36:46 some research. It comes up with a plan
36:48 and then it starts to execute that plan
36:50 and it has access to tools. And so one
36:52 of the tools it has access to is it can
36:54 surf the web. It can actually use its
36:56 vision to like look at websites and
36:59 click on them. The other thing it can do
37:01 is it can use deep research to go look
37:03 at a bunch of websites and get all the
37:04 data from them. The other thing it can
37:06 do is it can write and execute Python
37:09 code and do math and create, you know,
37:11 its own tools to accomplish whatever
37:14 goals you gave it. So, the way you you
37:17 kick this thing off, I'll put it into
37:20 full thinking mode. We'll go new chat.
37:24 And then you click plus here. And and
37:26 you've got agent mode. And notice I've
37:28 got 36 left that will reset on August
37:31 23rd. So tomorrow my things reset. So I
37:34 should probably play with them tonight.
37:36 Um so we're going to put it into agent
37:38 mode.
37:41 And then
37:43 you can turn on different sources if you
37:44 want to, but I'm just going to have web
37:46 search on is fine. And then I'm going to
37:49 say um
37:53 go research
38:02 Cost cutting
38:07 activity
38:12 in major
38:15 pharma companies.
38:19 We'll just say in in in the pharma
38:23 industry,
38:30 analyze
38:34 what you find.
38:43 Um,
38:46 present
38:49 this as
38:52 key trends
38:54 that a company
38:58 like Storyvine
39:01 could
39:04 exploit
39:08 and make me a presentation
39:14 I can share with my co-founder
39:20 for new go to market
39:24 strategy.
39:27 All right. And so now I kick that off.
39:30 And so it says setting up my desktop.
39:32 I'm going to make this a little smaller
39:34 so you can see.
39:36 Absolutely, Kyle. I'll research the most
39:38 recent cost cutting activities. Blah
39:40 blah blah. Basically restates what I
39:42 did. But now, okay, so notice this blue
39:44 thing. I'm starting a text string based
39:46 search. And so it's now off it's now off
39:48 using deep research
39:52 to um
39:56 to to find activity.
39:59 And the this little blue this blue bar
40:03 around it, this is the virtual computer
40:06 that it's spun up. It's spun up what's
40:07 called a container where inside that
40:11 blue box, it's like you've got a whole
40:13 separate computer, but chat GPT is the
40:15 one out surfing the web. You don't have
40:17 to do it. It's just off doing it.
40:20 All right. Um, find activity means what?
40:24 I don't know. I mean, oh, because I
40:26 asked it to go find activity. That was
40:29 part of my prompt.
40:31 Analyze. Let's see. Go research cost
40:35 cutting activity. So, so, so what this
40:38 thing does, I I should be able to
40:45 There we go.
40:47 So, along the bottom of this window is
40:52 is a little um slider. And what this is
40:57 is this is actually tracking the history
40:59 in this window. So, you can actually go
41:02 in. So the first thing it does, I'm
41:05 going to research cost cutting
41:06 activities and then you can kind of
41:08 slide this thing forward and you you see
41:10 what it does.
41:12 So it's it's basically recording all of
41:14 its activity. Now some of these things
41:17 like this may take 10 or 15 minutes and
41:21 I don't have to tell it what to do. like
41:23 it's out finding this stuff and and one
41:26 of the things that it will often do is
41:27 it'll create like a plan of here's the
41:29 things I'm going to do and then as it
41:32 finds things sometimes it'll update that
41:34 plan and realize oh I should change this
41:36 to go in a different order and then I
41:38 should do this and based on what I found
41:40 I'm going to do this
41:42 but what this is doing is I can now go
41:44 just spin up a separate chat GPT window
41:48 and this thing will just keep going.
41:54 Kyle, can you test LinkedIn? Give me the
41:57 best marketing
41:59 contacts based on a job title.
42:03 Um, I can let me let me try that as a
42:07 new
42:08 as a new window. Um,
42:13 I I I'll have to log into my own chat
42:16 GPT, which means I'm gonna have to find
42:17 my own
42:19 passwords.
42:22 My own passwords. So, I might need to
42:24 Brandon, just wait. Can you
42:29 make you a dinner reservation? I'm so
42:31 sick of the dinner reservation thing.
42:33 Let me let me see if I can get um
42:41 uh LinkedIn up. I'm going to have to log
42:45 into it. But go ahead and share my
42:48 screen, Brandon, and then the minute it
42:50 it it kicks me over to log in then then
42:53 just turn it off.
43:01 There you go. Okay. Um, so I've put it
43:04 in agent mode and then what did you say?
43:10 Test LinkedIn. Give me the best
43:12 marketing context based on job title.
43:14 Okay. um help me identify
43:22 um
43:24 pharma
43:25 marketers
43:34 with major launch brands
43:42 that uh Wait, help me identify pharma
43:46 marketers
43:49 uh
43:52 with titles
43:55 of decision making authority
44:04 on LinkedIn.
44:07 Let's see on my LinkedIn.
44:10 So that should kick us over to log in on
44:13 my LinkedIn
44:21 who work for
44:27 major launch brands
44:33 that
44:35 recently launched
44:38 or will launch.
44:40 in 20
44:43 26.
44:45 Put the names,
44:49 links,
44:51 and contact info
44:54 into a spreadsheet.
44:58 Okay.
45:00 So, I'm going to fire that off
45:05 and then we can flip back over to this
45:07 other one.
45:09 And this is still going. So like I don't
45:13 know what it's doing. It's still doing
45:14 research which one of the things that is
45:17 a absolutely mind-blowing if if you
45:19 haven't played with there there's other
45:22 there's other tools that you can play
45:24 with that have this similar
45:25 functionality. One is called Manis
45:27 manis.im m- anus.
45:30 There's another one called genspark. Gen
45:33 s p a r k. Both of those are agentic
45:37 kind of research tools. GenSpark in
45:39 particular is really good. It's fast.
45:42 It's welldesigned and it has a lot of
45:45 features. Um the chat GPT one is janky
45:50 and slow right now. They just launched
45:51 it and it's kind of crap. Um so if you
45:54 want to see a good one, go to GenSpark.
45:56 U Manis is also really quite good. It's
45:58 just not as pretty.
46:00 Um,
46:02 but these these things are going to get
46:04 very very capable in 2026.
46:07 These agents are going to be able to go
46:09 do sort of remarkable amounts of stuff.
46:11 They'll do it very very quickly because
46:14 even chat GPT can spin up 10 separate
46:17 agents on its own. It may decide, hey,
46:21 this is going too slow. I want to I want
46:23 to go do things in parallel. So that's
46:26 going to these things are going to get
46:28 quite amazing.
46:30 All right, let's see. Understood. I'll
46:33 conduct research. Okay, let's see. I'm
46:35 waiting while LinkedIn
46:39 Oh, okay. So, this is interesting
46:41 because I'm I'm logged into LinkedIn
46:44 here.
46:45 Um, it looks like like I'm logged in to
46:48 this LinkedIn right here. So, it's got
46:51 access to my LinkedIn, which is cool.
46:56 That's the other thing. There's a
46:57 there's a Perplexity now has a browser
47:03 and Perplexity
47:05 um it's called Perplexity Comet. And
47:08 what's cool about Perplexity Comet is it
47:11 can do this kind of agentic work as
47:13 well. And any tab that you have you've
47:16 logged into, it can access any tab. It
47:19 can basically go outside of its its tab
47:22 which which this is as well because I've
47:24 given it permissions to be able to do
47:25 that.
47:28 Yeah. God. Yeah. Yeah. They make the
47:30 data slippery. This is the other thing
47:32 to know about these agentic things is
47:36 privacy and security are
47:39 assume that they're garbage because
47:42 they're garbage. But assume that like
47:45 don't don't put anything in these things
47:47 that that is not stuff that if it went
47:49 out to everyone you'd be okay with. If
47:51 if it's anything less than that, you
47:53 should be careful.
47:56 Oh, this is this is really cool what it
47:58 just said there.
48:02 I don't know quite how to go back, but
48:04 it basically found this woman Jessica
48:07 and it said she's a second degree
48:09 connection for you. So, it's it's
48:11 identifying people that either, you
48:14 know, I'm not connected with, it's it's
48:16 at least aware of who I'm connected
48:17 with. It's pretty amazing.
48:22 All right.
48:24 Um, okay. Tik Tok is about to die soon,
48:27 isn't it? Uh, maybe. I don't know. Um,
48:30 the White House just started a Tik Tok
48:32 account like last week and they left
48:36 their comments on
48:40 and and and democracy is is taking
48:43 advantage of that particular little
48:46 fact.
48:48 Um, I don't know. Trump said he liked
48:51 Tik Tok, so maybe maybe they maybe it
48:53 sticks around for a while. I don't know.
48:55 Um it it does look like they're changing
48:57 their um community guidelines sometime
49:00 this month and any of the political Tik
49:03 Tok channels look like they're just
49:04 going to get decimated. So whatever. I
49:07 think we'll be fine. But I don't know.
49:09 Like the reason I dual cast on YouTube
49:12 and on Tik Tok is for precisely this. If
49:15 it goes away, it goes away. Some people
49:17 like Tik Tok, some people like YouTube.
49:19 So, but I've got more followers on Tik
49:21 Tok, so we'll just keep doing it like
49:23 this for a while.
49:25 Um, something happened in this Tik Tok
49:27 streaming media since last time I
49:29 watched.
49:31 Some people's moved. I don't know what
49:33 that means.
49:38 block.
49:49 All right. Um, let's go look at the
49:51 other the other agent. Is it still doing
49:54 its thing? Still doing its thing.
49:57 How video training is revolutionizing
49:59 pharma 4.0.
50:18 All right. Uh, let's see. We can go Oh,
50:22 you know what might be fun? Let's go.
50:23 Let's go watch some videos of what's his
50:25 name?
50:27 Um,
50:29 Tik Tok. Fernando.
50:48 Um,
50:53 all right. Let's see. He does a rock and
50:55 roll song. I turned my guitar into an
50:57 orchestra. Testing Sununo with my most
50:59 complicated song. Tested Sunso with an
51:02 anti-AI protest song.
51:11 Let's do the orchestra one. Hopefully he
51:13 talks a little louder here.
51:33 >> Oh, wait.
51:34 >> That said he was completely
51:35 quadriplegic.
51:37 Hang on.
51:38 >> All right, another test for Sunno.
51:40 Waltz for Doug. I was born and raised in
51:44 Miami, Florida. I went to Miami Beach
51:45 Senior High School. We had a teacher
51:46 named Doug Burus who taught classical
51:49 guitar ensemble recording studio class
51:52 and a thing called rock ensemble. The
51:54 thing that's crazy about Doug is that he
51:56 was completely quadriplegic
51:58 because of multiple scerosis. He could
52:00 only move his head. And
52:02 >> let me give you context of this if you
52:04 weren't here last night. So, so this
52:06 guy's a working songwriter and music
52:09 producer in LA and two nights ago, two
52:12 days ago, he posted a a video about, you
52:16 know, my job, my job is in jeopardy. He
52:19 basically discovered Sununo's ability to
52:21 take one of his songs and completely
52:23 produce it. Um, and so rather than
52:26 hiding from that, he's been kind of
52:28 leaning into it and learning what it
52:29 makes possible and testing it and doing
52:31 it publicly. One of the things that I
52:34 talk about all the time is play with
52:35 these tools and learn out loud. Share
52:38 what you're learning. He's gotten a lot
52:40 of hate for this, but he's also like
52:43 he's establishing himself as a
52:45 thoughtful musician who may not like
52:47 this, but he's also like these tools are
52:50 remarkable and they're not going away,
52:52 so why would I not learn them? Which,
52:54 you know, that's this is what he's doing
52:56 is very much in the spirit of this
52:58 channel.
53:00 you can hate this [ __ ] just don't
53:02 ignore it. And that's kind of where he
53:04 is. He's like he he's he says, "I'm a
53:07 I'm a combination of offended and
53:09 impressed." Right? And that's that's
53:12 where a lot of this stuff is
53:14 >> us by telling us what to do. Um he
53:17 taught thousands of students for 40
53:19 years and he was amazing. And when my
53:21 dad passed away, he became my second
53:23 father. I wrote this song for him. It's
53:25 called Walts for Doug. It's a classical
53:27 guitar composition. I would like Suno to
53:29 turn it into an orchestral thing. So,
53:31 here we go.
53:32 >> Can you hear this on YouTube? Oh, yeah.
53:34 You YouTube is coming through this
53:35 thing.
53:38 [Music]
54:02 Champy's singing.
54:05 [Music]
54:37 All right, let's see what it does.
54:40 Hit instrumental
54:43 string
54:45 quartet. Who's he playing the guitar
54:48 champion? Did you like that
54:49 >> classical?
54:53 No guitar.
54:59 One, Mississippi. Two, Mississippi.
55:02 Three, Mississippi. Four Mississippi.
55:05 Five Mississippi. Six Mississippi. Seven
55:08 Mississippi. 8 Mississippi. 9
55:11 Mississippi. 10 Mississippi. 11
55:14 Mississippi. 11 seconds.
55:17 [Music]
55:22 [ __ ]
55:28 Today is
55:31 [ __ ]
55:36 All right, another test.
55:38 >> All right, for my next test of Suno,
55:40 let's see.
55:40 >> Hey, why does it keep changing?
55:44 >> Hi, I'm Fernando.
55:46 All right. Another
55:50 Mississippi five Mississippi 6
55:53 Mississippi 7 Mississippi 8 Mississippi
55:57 9 Mississippi 10 Mississippi 11
56:00 Mississippi 11 seconds
56:04 [Music]
56:09 [ __ ]
56:10 [Laughter]
56:16 I bow down to you soon.
56:21 [Music]
56:22 >> That's such a that's such an honest dual
56:26 reaction. So I talk about in the stages
56:28 of AI adoption, stage three is this
56:31 stage of wonder and fear combined.
56:34 So his initial response is [ __ ]
56:38 because he doesn't think Sununo can do
56:40 the classical music thing, right? He's
56:42 tested it on rock and so he's like but
56:43 I'm sure it can't do classical and then
56:45 it does [ __ ]
56:48 and then he's like I bow down to you.
56:50 Like he doesn't know what they're doing.
56:51 He doesn't know how they're doing it but
56:52 he's like holy [ __ ] Here's this new
56:54 thing right? I'm both offended and
56:56 impressed.
56:58 And that's if you're not feeling that
57:01 weird
57:03 pull,
57:05 you're probably not deep enough. You're
57:07 probably haven't, you know, sort of
57:10 peeled back the onion enough about your
57:12 own personal work and the things you
57:14 care about. Like he clearly, this is his
57:16 life, right? And so this is fairly
57:20 existential stuff. You know, someone
57:22 made the comment here, I don't know if
57:23 it was Side Hustle, Mimi, someone made
57:24 the comment here about, you know, why
57:26 are people resisting this? so much.
57:28 There's already so much technology out
57:29 there that does a lot of this stuff in
57:31 music.
57:33 It there there's something different
57:36 about I mean this is for him to produce
57:39 that song traditionally, he would have
57:42 to go find musicians or he would have
57:44 to, you know, orchestrate it himself and
57:46 use MIDI tools and things like that.
57:49 This is 11 seconds after him playing
57:52 some crappy notes on the guitar. Um, you
57:55 know, it it turned into this. So, it's
57:57 just the fact that it's generating
57:59 music. It's not just processing our
58:02 inputs. It's actually generating new
58:04 stuff is is what's freaking people out
58:06 because he knows how long it would take
58:08 to produce that track traditionally.
58:11 [Music]
58:15 >> I wonder what Doug would have thought of
58:16 this.
58:18 [Music]
58:27 That's pretty good.
58:30 [Music]
58:37 My god. I wrote this when I was 17.
58:42 [Music]
58:44 I wrote this when I was 17. So he wrote
58:46 that little thing when he was 17. This
58:48 is one of the things when when you watch
58:50 Timberland who's a a music producer,
58:53 he's got a video of his first experience
58:55 with this and he has a similar sort of
58:59 reaction. He's like, "Wait a minute."
59:00 Like after he sort of after he sort of
59:02 processes what's happening, he goes,
59:04 "Wait a minute. That means I could take
59:07 any little SOG fragment I have and and
59:11 explore it, right? And I could explore
59:13 it quickly." Um, Tik Tok pin. Um, I
59:16 don't understand with how all the
59:18 technology exists that people are still
59:20 resisting
59:21 or still scared of AI. Um, yeah, just
59:24 because just just because it's um,
59:29 you know, it
59:31 here here's a here's a thing that I I
59:34 could be very off base with this. Um,
59:37 one of the things I hear a lot in the AI
59:40 salon when when people first start
59:42 playing with AI
59:44 is that
59:48 they feel guilty
59:51 that the thing that they can now do in 2
59:54 minutes or 5 minutes or 10 minutes that
59:57 they know should have taken them three
59:59 hours or three days now it's just done
1:00:02 that they feel guilty.
1:00:05 Um,
1:00:10 I I personally feel like
1:00:15 there's an overhang from the industrial
1:00:17 revolution that time equals money. That
1:00:21 in order for something to have value,
1:00:23 you have to put hours into it. It has to
1:00:26 have toil in it. I think in the creative
1:00:29 world, that that's a trope that is
1:00:31 particularly true. I've got to suffer
1:00:34 for my art, right? We talk about
1:00:36 suffering for your art. I think that's a
1:00:38 very industrial revolution
1:00:41 mentality.
1:00:43 And so I think the initial instinct of
1:00:48 I feel like I'm cheating. I feel guilty.
1:00:50 This is this feels like cheating is
1:00:52 because I shouldn't be able to create
1:00:55 something of this perceived value in
1:00:57 this little amount of time. And so I
1:01:00 think the AI resistance and fear part of
1:01:03 it is that gap
1:01:06 that it just shouldn't be that easy. And
1:01:07 if it's that easy, what is my value?
1:01:11 Right?
1:01:12 So it's not just that the product may or
1:01:16 may not have value. It's that well wait
1:01:18 a minute. If if my value is in the time
1:01:22 I put in to produce this stuff, what am
1:01:24 I worth now that this stuff can just be
1:01:26 done instantly? Right? That's that's the
1:01:29 kind of stuff that he's confronting
1:01:30 right now. Tik Tok question. Do you know
1:01:33 all of the content ownership issues with
1:01:35 Sunno? Can you explain? I am I am not an
1:01:39 IP attorney at all. Um I have opinions
1:01:42 on this.
1:01:44 Okay.
1:01:45 Couple of things. First of all, you need
1:01:47 to go go read the Puno terms of service.
1:01:52 Um because I think I think my my
1:01:56 co-founder just went and read them I
1:01:58 don't know three weeks ago because we
1:02:01 were wondering about the rights for for
1:02:02 using these for video projects.
1:02:05 What they basically say is
1:02:08 you can do anything you want with this
1:02:10 music
1:02:12 but if someone sues you that's on you.
1:02:16 Okay. So so they they are they are
1:02:19 severing themselves for any liability.
1:02:21 Now, the thing about the thing about
1:02:24 these tools, Sunno won't let you say,
1:02:27 "Give me a song that sounds like the
1:02:29 Eagles." It'll block that that
1:02:32 um that that generation. Um but you
1:02:36 could go into Chat GBT, for example, and
1:02:38 say, "Describe in detail this specific
1:02:42 song from the Eagles as if I were
1:02:44 explaining it to a music producer." And
1:02:46 then you could take that over there and
1:02:48 you could generate a song that sounds
1:02:49 like Hotel California, right? So if
1:02:53 you're using these tools to produce
1:02:56 content that is directly trying to sound
1:02:59 like copywritten works, you may get you
1:03:03 may get banged for that. Now,
1:03:05 that's that's likely not the the direct
1:03:08 question you're asking about, but I but
1:03:10 it's an important thing that that if you
1:03:12 push these tools into copyright
1:03:15 infringing territory, then you're
1:03:17 infringing copyright and you know, you
1:03:19 you may get sued. The the bigger
1:03:22 question is this. The first one is how
1:03:25 did they train these models? Did they
1:03:26 train these models on copywritten work?
1:03:29 almost 100% guaranteed that they trained
1:03:32 these model on cop models on copywritten
1:03:35 work. Almost 100% guaranteed.
1:03:39 Now the next question is how does it
1:03:42 produce what it produces? Is it copying
1:03:44 and pasting from those original works?
1:03:47 And the answer there is no.
1:03:49 And that's one of the things that people
1:03:51 most misunderstand about generative AI
1:03:54 is just because it can produce something
1:03:56 that sounds like something it was
1:03:58 trained on does not mean it's copying
1:04:00 and pasting from that original work. The
1:04:02 way these things work is when the models
1:04:05 are trained, the original data gets
1:04:09 turned into all these bizarre
1:04:10 mathematical fragments that live in
1:04:13 thousanddimensional mathematical space.
1:04:15 The original works actually cease to
1:04:18 exist. There's there's current law out
1:04:20 that basically says that the act of
1:04:23 training a large language model is what
1:04:26 they call a transformative act. You're
1:04:28 transforming in a significant way, in a
1:04:31 material way the original input document
1:04:34 into something completely different. And
1:04:36 therefore, if the training document was
1:04:39 obtained legally, it is considered fair
1:04:42 use. If you stole it, then it's not fair
1:04:45 use. But if you bought a copy of this
1:04:48 book and trained it into your model,
1:04:51 because it's a transform, materially
1:04:53 transformative,
1:04:54 it is um it's considered fair use.
1:04:59 So when Sunno makes one of these songs,
1:05:01 it's actually generating an original
1:05:04 song that's never existed before.
1:05:07 And so based on that, you can use these
1:05:10 things for anything. Now, can you
1:05:13 copyright them? That's a whole other
1:05:15 thing. It depends. In his case, he wrote
1:05:18 that song when he was 17. He can
1:05:20 document that he wrote that song when he
1:05:22 was 17. He's got a Tik Tok video where
1:05:24 he plays it into the tool and it takes
1:05:27 that that song that he wrote and it
1:05:30 basically just reproduces it. It it
1:05:33 produces it in a new and unique way, but
1:05:36 it's his song, right? So, he's got a
1:05:38 chain of craft that he could probably
1:05:41 take to the copyright office and say,
1:05:42 "Yeah, this is my song. I put it in
1:05:44 there." You could potentially do the
1:05:46 same with lyrics. You could potentially
1:05:47 do the same if you do 400 iterations of
1:05:50 a song before the one that you publish
1:05:52 and you track all that. Um, so this is
1:05:56 incredibly complicated, incredibly
1:05:59 nuanced territory that I don't think is
1:06:03 going to be settled for 20 or 30 years.
1:06:05 So, so my, you know, my thing is this.
1:06:08 If you're working for, say, a big pharma
1:06:11 company like I do at StoryVine, we need
1:06:13 to be much more careful about what we
1:06:15 use. I think for most people, for most
1:06:17 use cases, don't worry about it. Just go
1:06:19 make music and play. Let the lawyers
1:06:21 figure this [ __ ] out. If you're trying
1:06:23 to infringe on someone else's um
1:06:25 copywritten material, then you're going
1:06:27 to get in trouble. But if you just want
1:06:29 to use it as a regular creator, maybe
1:06:30 not. Okay. Tik Tok pin. The the pin's
1:06:33 not there because I was talking too
1:06:34 much. So, pop it back up there. Uh but
1:06:37 really good question.
1:06:39 >> Fascinating times we live in.
1:06:52 >> I still the three hours because my
1:06:55 experience shaped the prompt. Yep. That
1:06:57 so that that's that's interesting side
1:07:00 hustle Mimi. Um,
1:07:04 I think that the
1:07:07 the core business model of the
1:07:09 advertising world is time and materials,
1:07:11 right? How how ad agencies for the most
1:07:14 part make their money is by having 50
1:07:18 copywriters sit in a bullpen and write
1:07:20 copy all day and charge $165, $250 an
1:07:25 hour for it, right? That's their
1:07:26 business model. We now have a tool that
1:07:29 can just generate copy in seconds.
1:07:32 Um,
1:07:34 you can do what's called value pricing,
1:07:36 right? You can say, "Hey, hey, you know,
1:07:38 client, you need, you know, 50 Tik Tok
1:07:42 ads. Um, in a time and materials model,
1:07:45 that would have cost you whatever
1:07:48 $50,000.
1:07:50 I'll tell you what I'm going to do
1:07:52 because I've got professional
1:07:54 copywriters and editors who are going to
1:07:56 make sure that the content's still good.
1:07:58 But now, but because we're using these
1:08:00 AI tools, we can be more efficient.
1:08:03 Rather than 50 grand, we'll charge you
1:08:05 35 grand. And then maybe instead of it
1:08:08 costing,
1:08:10 you know, the agency 15 grand to get
1:08:12 that 35 grand, maybe it only cost them
1:08:15 five grand, so they can actually charge
1:08:17 less and make more money. Um
1:08:22 most agencies right now are resisting
1:08:26 that shift like the plague which which
1:08:30 says to me there's a major opportunity
1:08:33 to to have a a startup agency that can
1:08:36 undercut traditional agencies um and
1:08:39 overd deliver under undercut them on
1:08:41 price o overd deliver on output.
1:08:45 She slug of doom. I need to get my new
1:08:48 glasses. I thought the Chiron had
1:08:50 changed to Kyle Shannon, alligator.
1:08:55 Oh, AI learning lab.
1:09:01 Okay, wait. We've got we've got results.
1:09:03 Oh, wait. The one Hang on. The one
1:09:05 result is still going. The original
1:09:08 prompt we did is still going. This was
1:09:10 the one. What did I do?
1:09:13 Oh, this was the farmer research one,
1:09:15 but the the LinkedIn one is done. Okay,
1:09:19 understood. I'll conduct
1:09:22 a list on LinkedIn to identify
1:09:24 pharmaceutical marketers. Worked for 20
1:09:27 minutes,
1:09:28 right? So, here's here's what it did for
1:09:31 20 minutes,
1:09:34 right? Did all this stuff. Was on
1:09:37 LinkedIn. It was cruising around
1:09:38 LinkedIn. It wrote some code. Did it
1:09:42 switch their tabs?
1:09:43 >> Oh, did it? It did not. That's rude.
1:09:48 Anyway, okay. So, it thought for 20
1:09:49 minutes. Um, I compiled a list of pharma
1:09:53 marketing leaders with companies major
1:09:55 launch brands.
1:09:57 For example, Fierce Pharma reports that
1:09:59 Vertex's blah blah blah triple therapy
1:10:02 is one of 2025's most anticipated
1:10:04 launch.
1:10:06 All right, here's here's our
1:10:08 spreadsheet.
1:10:10 So, let's go look at it and see if it
1:10:12 gave us Here's my prediction. So, it
1:10:15 thought for 20 minutes.
1:10:18 Oops. Oh,
1:10:21 this is better than I thought it would
1:10:23 be. So, let me get pe rid of people's
1:10:26 real names here.
1:10:27 Um,
1:10:31 so if found
1:10:38 one, two, three, four, four people from
1:10:40 Vertex,
1:10:42 two people from InMed, four from
1:10:44 Astroenica,
1:10:48 six from or five from Dichi Sano,
1:10:52 some from GSK, and some from Santa Fe.
1:10:54 Um, this is better than I thought it
1:10:56 would be. I my my guess is is that this
1:10:59 would have found um like two or three
1:11:03 and been done. The fact that it found
1:11:04 What did it find here? 20.
1:11:07 It's not bad. It's not bad at all. All
1:11:11 right. Well, there you go. So, that's
1:11:14 how agent works. You just turn the thing
1:11:16 loose. It goes does [ __ ] and then,
1:11:19 you know, makes you a spreadsheet.
1:11:23 Boom. Bing bang. Boom.
1:11:27 Oh, and then this thing. Oh, no. No.
1:11:28 This thing's still going. Okay. That
1:11:32 one's doing the research. All right.
1:11:33 It's still It's still Oh, you know what
1:11:35 it how training
1:11:38 how video training is revolutionizing
1:11:40 pharma. I have a feeling this hung,
1:11:52 which is one of the things it does. But
1:11:54 maybe not. We'll just let it keep
1:11:55 running.
1:11:59 Desktop view.
1:12:03 All right.
1:12:08 The Chroma Awards in the Salon. Okay.
1:12:11 Hang on.
1:12:15 So, let me
1:12:18 pop up the uh the salon address there,
1:12:22 Brandon, if you'd be so kind.
1:12:26 All right. So, see that address right
1:12:27 there? If you're not part of the AI
1:12:29 salon.
1:12:31 So, so this live is brought to you by
1:12:34 the AI Salon. The AI Salon is a
1:12:36 community of about 3,000, more than
1:12:38 3,000 now, 3,100 um AI optimists, people
1:12:42 trying to figure this AI stuff out. Um
1:12:45 if you're not a part of a community and
1:12:48 you're trying to get your head around
1:12:50 AI, it's going to be very lonely because
1:12:52 everything's changing really fast. It's
1:12:53 very hard to keep up with things. This
1:12:56 is a really remarkable community. So, um
1:13:00 sign up for this if you go to
1:13:02 community.thesalon.ai.
1:13:04 AI. Um, that'll bring you here. Um, you
1:13:08 can learn about our cycle of AI
1:13:09 readiness, play first, mindfully create,
1:13:11 generously lead, what that all means,
1:13:13 the five stages of AI adoption, our
1:13:16 values. The second thing to do here is
1:13:19 introduce yourself. So, if you if you
1:13:21 have not, go into the section called
1:13:23 introduce yourself and introduce
1:13:25 yourself. And by the way, here's another
1:13:28 thing that I would love to do. We did we
1:13:29 did this the other night.
1:13:32 Um, if you're in a regular here, do me a
1:13:34 favor. Go in to introduce yourself in
1:13:37 the salon right now and just um, you
1:13:40 know, welcome people in and and uh, like
1:13:44 give them give them a little a little
1:13:46 hearty heart or an applause or respond
1:13:50 to them. Let's let because I know
1:13:52 there's a bunch of new people in here.
1:13:53 So, let's let Oh, that's really cool.
1:13:55 Let's let uh, some folks know that they
1:13:58 are welcome to be here. thumbs up,
1:14:02 hearty hearts, things like that. So, if
1:14:04 you if you all if you irregulars would
1:14:06 go into the introduce yourself section
1:14:08 and make people feel welcome, that would
1:14:10 be swell. Okay, now Vicki has put
1:14:13 something. Where did Vicki put her her
1:14:15 stuff? So, agent will do it, but the
1:14:19 plus account won't. No, agent, you have
1:14:21 access to agent within the plus account.
1:14:25 Hey, Kyle. Decided to pay for chat GPT,
1:14:27 but disappointed.
1:14:29 still must wait if images docs are
1:14:32 involved. No, I don't know what that
1:14:35 means, Darren. You should be able to
1:14:37 generate images. You should be able to
1:14:40 generate docs. You should be able to
1:14:41 ingest docs. Ingest images if you're
1:14:45 paying 20 bucks a month for chat GPT.
1:14:47 Now, the only thing that's possible,
1:14:48 Darren, is if if you're doing it on
1:14:51 mobile and you didn't download the
1:14:54 official chat GPT app, but you
1:14:56 downloaded an app that's one of these um
1:15:00 uh parasite apps that's not really from
1:15:03 ChatGpt. It's possible you paid for
1:15:06 something that's not an actual uh
1:15:08 subscription to ChatGpt. Challenges and
1:15:11 competitions. Okay, cool. So, so in in
1:15:15 the AI salon, you've got start your
1:15:17 adventure. So, there's a little
1:15:18 onboarding section here, um, in the
1:15:21 community section, there's announcements
1:15:23 and a showand tell area and learn out
1:15:26 louds where you can, um, teach things if
1:15:28 you want to teach things. And then under
1:15:30 latest news, there's challenges and
1:15:32 competitions.
1:15:34 And so, the Chroma Awards are coming.
1:15:37 Free trials for a bunch of tools. Very
1:15:40 cool. a film, music, and games
1:15:42 competition toi unite creators,
1:15:44 communities, and companies around the
1:15:46 world.
1:15:48 If we go to the Chroma Awards,
1:15:51 film, music, video, or games. Very cool.
1:15:57 A6Z, LTX Studio, Bolt, Cling, Open Art,
1:16:01 Fairground, and Image Art. So, the way
1:16:04 these contests tend to work, I am
1:16:07 sharing this. Oh, I'm not.
1:16:10 The way these contests tend to work is
1:16:14 they want you to create content that
1:16:17 they're going to, you know, give awards
1:16:18 for and things like that, but they cut
1:16:21 deals with these these companies. 11
1:16:24 Labs, Freepick, F, um, Dreaming AI, Cap
1:16:28 Cut, A16Z, which is they're an investor,
1:16:32 LTX Studio, Bolt, Clling, Open Art,
1:16:35 Fairground, and Image Art. So, you will
1:16:37 likely get credits for all of these
1:16:40 different companies if you apply for
1:16:44 this contest. And if you're like, "Well,
1:16:47 I don't know how to use any of those
1:16:48 tools." Let's go back to to to the the
1:16:52 AI salon for a second.
1:16:55 Let's go to the very first page you're
1:16:57 dropped into.
1:16:59 And the very first graphic that you see
1:17:00 after our welcome video
1:17:03 is this.
1:17:06 play first. Play first. So, if if you're
1:17:10 thinking that you don't have the skills,
1:17:12 you don't have the talents to join this
1:17:15 contest, [ __ ] Go do it anyway. Um,
1:17:19 Becky Rue, who's in here, reminded me. I
1:17:23 didn't realize that I did this to her,
1:17:26 but but when Becky Rue first joined
1:17:28 joined the lives here, she she made the
1:17:31 mistake of commenting, "I'm too old to
1:17:33 do this AI stuff." And I was like,
1:17:35 "Bullshit. No, you're not." And then she
1:17:38 she went and played. So, if you feel
1:17:40 like you don't have the the uh the
1:17:42 talent or you're not creative enough to
1:17:44 go make a film or a music video or a
1:17:46 game, guess again. You do. You just got
1:17:49 to go play. So, go join this and play.
1:17:52 All right. Ann Murphy in the house. Ann
1:17:55 Murphy, how does my voice sound? Huh?
1:17:58 The new microphone. I'm all
1:17:59 professional. Like, sound good.
1:18:05 Um, so let me take you to another uh
1:18:09 another site. If you go to
1:18:12 um you can pop this up on screen too. Uh
1:18:15 Brandon, are you ready for ai.com? Are
1:18:18 you ready? Number four, aai.com.
1:18:24 It's going to take you to this
1:18:26 beautifully designed website built by
1:18:28 Vicky Baptiste
1:18:30 um who also did the instructional design
1:18:33 for the AI readiness training program.
1:18:37 So, a couple of things to talk about
1:18:38 here.
1:18:40 Last between Christmas and New Year's in
1:18:43 2024, we did this thing called AI
1:18:46 Festivus, AI for the rest of us, where
1:18:48 we brought in 34 speakers from four
1:18:50 different communities. We had thousands
1:18:52 of participants
1:18:54 um and and it was 24 hours of
1:18:56 programming o across two days. So 12
1:18:59 hours at night um on the on Friday, 12
1:19:03 hours on Saturday in between Christmas
1:19:05 and New Year's. We're going to do that
1:19:07 again this year. But what happened was
1:19:11 all of the people that came got so
1:19:13 excited about the presentations, they're
1:19:15 like, "This needs to be turned into a
1:19:17 training." Because what happened last
1:19:19 year that had that had never really
1:19:21 happened before was every single
1:19:23 presentation even though they used AI
1:19:25 tools were not training people how to
1:19:28 use specific AI tools. Every single
1:19:31 presentation was about the mindset of
1:19:34 how they think about using tools or how
1:19:37 they think about privacy or how they
1:19:38 think about doing creative work when
1:19:40 they're using AI. And so an and Vicki
1:19:44 and I got together and at the request of
1:19:47 the community put together the AI
1:19:49 readiness training program which is not
1:19:51 about tools but it's about the mindset.
1:19:53 So there are five different training
1:19:56 modules within the AI readiness training
1:19:59 project. It's a really unique program um
1:20:02 and in incredibly powerful. So if you
1:20:06 want to elevate your AI skills, check
1:20:09 this training out. It is. It's robust.
1:20:12 There's a lot here. All of the videos
1:20:14 from AI Festivus are in here. They're
1:20:18 sort of the source material that that
1:20:20 you learn from. And then we extracted
1:20:22 out of them all of these universal ideas
1:20:25 across five different sections from
1:20:27 business to creative to privacy and
1:20:30 ethical use of AI to putting it all
1:20:32 together in business. There's there's a
1:20:34 bunch of different things. Um really
1:20:36 powerful. So go check that out if you
1:20:38 haven't. Okay.
1:20:41 Um,
1:20:45 [Music]
1:20:47 all right.
1:20:49 Fantastic. Fantastic. Yeah.
1:20:53 All right. What's happening, Ann Murphy?
1:20:55 Anything going on? Anything else we
1:20:57 should talk about? Oh, the other thing
1:20:59 we should talk about is AI Festivus is
1:21:01 happening again in 2025.
1:21:05 So, December 26th is a Friday. December
1:21:09 27th is a Saturday. So, we're going to
1:21:12 do the same thing again. It's going to
1:21:13 be split over two days, 24 hours of
1:21:16 programming. The one thing that we're
1:21:18 going to do this year that we did
1:21:19 different from last year is everyone was
1:21:21 so excited last year to hang out after
1:21:23 the event. Um, we're going to have an
1:21:25 official hangout area after the 24
1:21:28 hours. Anna and I are going to go to bed
1:21:31 because we'll be hosting the damn thing.
1:21:33 So, we're going to go to bed, but you
1:21:35 all are welcome to hang out. So, um, we
1:21:39 don't have we don't have the, uh, the,
1:21:41 uh, anything up and running just yet,
1:21:44 uh, for you to sign up for it, but just
1:21:46 save the dates of December 26th,
1:21:49 December 27th, the Friday and Saturday
1:21:52 between, um, you know, Christmas,
1:21:55 Hanukkah, those kind of things, and New
1:21:57 Year's uh, is is when it's going to
1:21:59 happen. All right.
1:22:01 Oh, yes. Yes, there is another thing.
1:22:06 So, one of the things Ann and I put
1:22:07 together as a as a parallel initiative
1:22:11 to the training program is we created
1:22:14 the AI readiness project, which is a
1:22:16 podcast where we interview people and we
1:22:19 talk about how ready for you, how ready
1:22:21 for AI are you and what does that
1:22:22 actually mean? And um so
1:22:27 so if you go to aire readiness
1:22:30 project.com
1:22:33 um or you go search for the AI readiness
1:22:36 project on um Apple podcasts, Spotify
1:22:40 podcasts, I think it's on Amazon
1:22:43 podcasts. Um please go subscribe to your
1:22:46 podcast delivery vehicle of choice and
1:22:50 do whatever you do there. I think you
1:22:51 subscribe to our podcast, but leave a
1:22:53 review, listen to them. Um, we just we
1:22:57 just celebrated our 25th episode, which
1:23:01 if you've ever done a podcast, most
1:23:03 podcasts die before they hit double
1:23:06 digits. They die at nine, almost all of
1:23:10 them. So, the fact that we hit 25, I
1:23:12 don't know. I'm a little proud. I don't
1:23:14 know about you, AM, but I'm a little
1:23:15 proud.
1:23:16 Um, how do you feel about MIT saying 90
1:23:20 91% of initiatives are not meeting
1:23:22 expectations? Um, I'm surprised I'm
1:23:26 surprised that Well, I think it was 95%
1:23:28 not 91%. Um, I'm surprised that that
1:23:32 many are successful here. Here's
1:23:35 here. Okay.
1:23:43 Almost all AI initiatives that big
1:23:46 corporations are rolling out right now
1:23:49 are being rolled out as IT initiatives
1:23:52 where they like okay you got access to
1:23:55 chat GPT go
1:23:58 and people like uh
1:24:01 okay
1:24:03 what do I do with this
1:24:06 the people aren't being trained
1:24:09 um
1:24:11 these initiatives. Look,
1:24:14 the reason it makes sense for me to go
1:24:16 live five nights a week here is because
1:24:19 nobody knows exactly what to do with
1:24:21 these tools because we've never had
1:24:23 capabilities like this before. One of
1:24:26 the mistakes that big corporations are
1:24:28 making right now is they're treating
1:24:29 generative AI like they've treated
1:24:31 historical AI, right? So, okay, we use
1:24:34 Salesforce, we're going to add AI tools
1:24:36 into it, and we're going to roll that
1:24:37 out to all the people. That's a very
1:24:39 specific vertical use case where if
1:24:42 you're doing a sort of vertical use case
1:24:44 like that where you're where AI is
1:24:46 amplifying an existing process, those
1:24:49 are probably more successful because you
1:24:51 know exactly how to do the process
1:24:53 before. Now AI is making that a little
1:24:55 bit easier. But if you just take
1:24:57 generative AI as a general capability
1:25:00 and you throw it at the organization,
1:25:03 nobody knows what to do with it. Nobody
1:25:06 knows how to use it. Nobody use knows
1:25:08 why to use it. What does success look
1:25:10 like? So it is not surprising to me at
1:25:13 all that these initiatives are failing.
1:25:20 What the impact of an article like that
1:25:23 are going to have
1:25:26 this is [ __ ] insane. We're we're
1:25:28 we're just we're going to just watch a
1:25:30 series a rolling series of Kodak moments
1:25:33 where where big companies are are going
1:25:35 to start getting their asses handed to
1:25:37 them by little startups. And and what's
1:25:39 going to happen is this.
1:25:42 People are going to take that article
1:25:43 and they're going to go, "Oh, AI was
1:25:45 just a bubble. I told you it was just
1:25:48 hype. It was just like NFTTS and
1:25:50 blockchain. It was just hype. We can
1:25:52 ignore AI now."
1:25:54 and and so so focus on a AI in a lot of
1:25:58 corporations is probably going to get
1:26:00 pulled back. Now what's going to happen
1:26:02 is smart corporations that put together
1:26:05 little pods of people to actually
1:26:08 explore generative AI and figure out on
1:26:10 a department by department basis um what
1:26:14 generative AI actually makes possible
1:26:17 and do things from the from the workers
1:26:20 up rather than from it down.
1:26:24 those companies are going to start to
1:26:26 innovate and start to start to really
1:26:28 leverage the [ __ ] out of AI. So, I think
1:26:31 what you're going to see is you're going
1:26:32 to see a bifurcation starting this year,
1:26:34 a bifurcation of companies that pull
1:26:36 back on AI investment and those that
1:26:39 invest more heavily into it and actually
1:26:42 understand how to train their employees,
1:26:44 how to roll these things out in a smart
1:26:46 and ethical and uh intelligent way. and
1:26:51 and those two kind of companies are
1:26:53 going to separate really quickly because
1:26:54 the ones that lean into it and really
1:26:56 figure out how to use this stuff are
1:26:58 going to start to amplify what they do.
1:27:00 So that's that's that's where I think
1:27:02 this goes. Um so I'm not surprised at
1:27:05 all that that article came out. You you
1:27:08 know
1:27:10 media companies are going to do
1:27:11 sensational headlines like that. um most
1:27:14 people are not going to read the
1:27:16 articles and they're going to completely
1:27:17 mischaracterize
1:27:20 why
1:27:22 why the success rate is so low and and
1:27:25 it's it's it's not going to be good for
1:27:27 them. So anyway, there you go. There you
1:27:29 go. There you go. There you go. There
1:27:32 you go. There you go. Um
1:27:35 we won't consult with any org if they
1:27:38 have us work with it. That's Ann Murphy.
1:27:41 So Ann Murphy is is doing this on a
1:27:43 regular basis. She's working with
1:27:44 organizations to teach them how to use
1:27:47 AI to think about it. And yeah, if
1:27:50 if you think about okay,
1:27:54 think about AI
1:27:57 like you think about the internet and
1:27:59 the worldwide web. When when when I
1:28:02 started building websites in the mid 90s
1:28:04 for big Fortune 500 companies,
1:28:08 one of the very first things that we did
1:28:10 as an agency was we would go in and we
1:28:13 would wrestle
1:28:15 the control of web development away from
1:28:18 it.
1:28:20 Now, I'll tell you a really specific
1:28:21 example of this. CocaCola put out a in
1:28:26 this was I think 1996
1:28:29 they put out an RFP for an agency of
1:28:32 record pitch for people that for
1:28:34 agencies that knew how to build
1:28:35 websites.
1:28:38 And so we got invited to pitch it. We
1:28:42 took all of our best people. We're like
1:28:43 this is Coca-Cola. We've got to it's got
1:28:45 to be polar bears and refreshing
1:28:47 graphics and we've got to figure this
1:28:49 [ __ ] out, right?
1:28:54 So, we put together this really
1:28:55 beautiful pitch. I probably have it here
1:28:57 somewhere. We put this together this
1:28:59 really beautiful pitch. We went down to
1:29:01 Atlanta. We're sitting in a room with
1:29:03 about I don't know, seven or eight or
1:29:05 nine Coca-Cola executives and seven or
1:29:08 eight or nine of us
1:29:11 and you know, we're talking about what
1:29:13 the web is and why it is and what we've
1:29:16 done and things like that. And then we
1:29:17 get to this slide. And what the slide
1:29:20 was was two screenshots. One was a
1:29:22 screenshot of Coca-Cola.com
1:29:25 and one was a screenshot of coke.com.
1:29:28 Both of those pages had been built by
1:29:31 the IT department. One was light yellow
1:29:35 with blue hyperlinks and the hyperlinks
1:29:37 went to all the annual reports in the
1:29:40 history of the Coca-Cola Corporation.
1:29:42 And the other one was a white page with
1:29:45 black hyperlinks that went to I don't
1:29:47 know uh other documents some documents
1:29:51 that it thought were important
1:29:54 and and our headline for that slide was
1:29:57 where are the polar bears.
1:30:00 The Coca-Cola executives stopped the
1:30:03 meeting.
1:30:05 They said they said stop.
1:30:11 We'll be back. They all left.
1:30:18 They all walked out of the room. We're
1:30:20 like, "What the [ __ ] did we do? What's
1:30:22 going on? Are we getting fired? Like,
1:30:24 are they going to ask us to leave? Are
1:30:25 they getting security? What's
1:30:27 happening?" Um,
1:30:30 they had never seen their website.
1:30:34 They had never gone to it. They've never
1:30:36 looked at it. when they saw how their
1:30:40 brand was being represented to the
1:30:43 world,
1:30:45 they absolutely flipped out. They walked
1:30:47 back into the room after about 15
1:30:49 minutes and said, "Can you start
1:30:50 Monday?"
1:30:52 No one had ever shown them this before.
1:30:54 So, it's not that IT groups are bad or
1:30:57 evil. It's that the worldwide web and
1:31:01 generative AI
1:31:03 affect every single part of the
1:31:05 organization from marketing to
1:31:07 operations to sales to product all of
1:31:10 it. And you and it's it's not just like
1:31:14 an enabling you know CRM that it has to
1:31:20 install and maintain. It's a completely
1:31:22 new capability and it's got to be
1:31:24 thought about.
1:31:26 I just got restricted.
1:31:31 This is primarily
1:31:33 directs people off the platform. This is
1:31:35 in line.
1:31:46 Whatever. I just appealed it.
1:31:50 whatever
1:31:52 restriction
1:31:56 I want to hear more and talk about this
1:31:57 forever. Yeah. So, so the the thing is
1:32:00 is
1:32:02 if I if it is treat is is treating chat
1:32:06 GPT like it like like it's Salesforce.
1:32:10 It's not. It's just it's a fundamentally
1:32:12 different thing. Like nobody knows how
1:32:16 to use generative AI to run their
1:32:18 business right now. Sure, you can do
1:32:21 automations. You can use AI and
1:32:24 something like Zapier or NATO to create
1:32:26 an automation to make an existing
1:32:28 workflow more efficient. You can do
1:32:30 that. A lot of people are doing that. A
1:32:32 lot of consultants are making money
1:32:33 doing that.
1:32:36 But that's using AI to do what you
1:32:38 already know how to do. Generative AI
1:32:42 lets you do things in completely new and
1:32:44 fundamental ways that you can only
1:32:46 discover through innovation and
1:32:49 experimentation and play.
1:32:51 And that's not the purview of it. It is
1:32:54 there once we have something locked
1:32:56 down. We're going to make it solid and
1:32:58 more efficient. That's not the stage
1:33:00 we're at with with AI right now. We've
1:33:02 got another decade to figure out how we
1:33:05 use this stuff, how it impacts all the
1:33:07 different areas of a business. So
1:33:08 anyway, that's my two cents on that.
1:33:12 Um,
1:33:14 [Music]
1:33:18 the people side is so important. The
1:33:20 people side's everything. Listen, these
1:33:23 are large language models. These are not
1:33:26 large programming models. They're not
1:33:28 large engineering models. They're large
1:33:30 language models. So how we interact with
1:33:34 them is very different than how we've
1:33:35 interacted with technology in the past.
1:33:37 And so this is all about people skills
1:33:40 and connecting and speaking and
1:33:42 communicating because
1:33:45 one set of skills is how you interact
1:33:47 with people. The other set of skills is
1:33:50 how you take that language and you
1:33:51 interact with large language models and
1:33:53 get the best results out of them.
1:33:55 They're very related. Right? So there
1:34:00 you go. There you go people. You
1:34:03 understand what I'm saying?
1:34:06 All right. All right. Yeah. Um, let's go
1:34:11 see if our other agent is still stuck.
1:34:27 Still stuck. Yeah, this hung. This hung.
1:34:33 Wait, maybe it didn't.
1:34:38 What is
1:34:41 share this tab?
1:34:49 Um, here I am. This thing looks like
1:34:52 it's still stuck here.
1:34:55 Um, I'm just going to say where is my
1:35:02 presentation? So, remember before when I
1:35:04 said this thing is janky and it might
1:35:06 not work all the time.
1:35:09 This is janky and it didn't work all the
1:35:11 time. This one sort of worked. The one
1:35:13 the one where someone asked me to go
1:35:14 look at my LinkedIn and find, you know,
1:35:16 target target
1:35:19 uh prospects. That worked. This one did
1:35:23 not. Okay. I'm still in the process of
1:35:25 compiling research
1:35:27 and crafting my presentation based on
1:35:29 the latest cost cutting trends. I'll
1:35:32 focus on how these trends align with
1:35:33 StoryVine's offerings once I've
1:35:35 completed. So continue.
1:35:45 [Music]
1:35:48 All right. I guess we we got it back
1:35:50 going back live again. All right.
1:35:53 Whatever.
1:35:54 Um,
1:36:00 here here's another thought I have on on
1:36:02 these agent things.
1:36:07 Even if you go play with Gen Spark or
1:36:09 Manis, which are which are more capable
1:36:12 than chat GPTs. The reason chat GPTs is
1:36:15 significant is they've got 700 million
1:36:17 weekly users. Now of those probably 37
1:36:21 are using chat GPT deep enough to
1:36:24 understand what agent is right it's like
1:36:26 not a lot of people are using this right
1:36:28 now but you know relatively speaking but
1:36:31 as this tool becomes more transparent
1:36:34 and more effective you know people are
1:36:36 going to have instant access to
1:36:37 incredibly powerful tools oh it looks
1:36:40 like it's building us our our deck now
1:36:42 the report has been created all right
1:36:44 good
1:36:46 Um,
1:36:50 don't
1:36:52 one of the biggest things I would
1:36:54 caution against right now is
1:36:58 don't dismiss any of these technologies
1:37:01 because they suck right now. You know,
1:37:04 one of the things we always say in here
1:37:05 is, you know, these tools are as bad as
1:37:07 they're ever going to be. The fact that
1:37:09 this thing can go like look at my
1:37:11 LinkedIn and find people and do that
1:37:13 without me having to be there to do it.
1:37:14 The fact that this can go do research
1:37:16 and create a presentation for me even if
1:37:19 it's janky and even if it's a crap
1:37:22 report
1:37:24 means that imagine what that's going to
1:37:26 be like a year or two from now. So by
1:37:30 playing and by exploring what these
1:37:32 tools can do today, you'll be ready when
1:37:35 these things get good. And when these
1:37:37 things get good, it'll be you will occur
1:37:40 to the rest of the world like a wizard.
1:37:44 Because very few people know what's
1:37:47 coming. Very few people know what's
1:37:48 here.
1:37:50 Even people that say, "Oh yeah, I use
1:37:52 chat GPT all the time." Do you know how
1:37:55 many people like on a weekly basis tell
1:37:57 me, "Oh yeah, I use AI all the time."
1:38:00 And then I show them some simple thing
1:38:02 within chat GPT, they're like, "Oh, I
1:38:04 didn't know it could do that."
1:38:07 It's like regular all the time. So, so
1:38:11 just keep coming back to these lives.
1:38:13 Join the AI salon. Hang out with people.
1:38:16 Build relationships with people. Okay.
1:38:21 Spin B3. Dip time. Good night. Peace
1:38:23 out. Spin B3.
1:38:26 Try three months, not two years. Yeah, I
1:38:28 know. I know. This stuff is moving too
1:38:30 fast.
1:38:33 All right. This is now This is now
1:38:35 successfully building us slides, which
1:38:38 is cool. So we should we should have
1:38:40 slides to uh to look at here shortly.
1:38:44 Worked for three minutes. Yeah. After it
1:38:47 hung for an hour.
1:38:52 All right. So there's our report.
1:38:55 It looks like it's in a canvas.
1:38:59 Report comprehensive thing.
1:39:03 Um
1:39:05 is this a canvas?
1:39:07 Maybe not. Cost cutting activity
1:39:10 industry contest shrinking margin and
1:39:12 patent cliffs force restructuring.
1:39:15 Merc's cancer drug Kituda. Did you know
1:39:18 Kituda for Merc is 80% of their revenue?
1:39:21 80% one drug. It's like it's got like 46
1:39:26 indications, 46 different ways it can be
1:39:29 used to battle cancer, but it's coming
1:39:32 off patent.
1:39:36 pipeline prioritization, outsourcing,
1:39:39 digital transformation and AI adoption,
1:39:43 generative AI and analytics for
1:39:45 efficiency, decentralized clinical
1:39:47 trials,
1:39:50 implications for StoryVine, reducing
1:39:52 marketing and comm's cost,
1:39:55 supporting decentralized clinical
1:39:57 trials,
1:39:58 scaling training and compliance
1:40:00 programs, enhancing supply chain
1:40:02 visibility, internal comms,
1:40:06 Conclusion. There's that. Where's my
1:40:08 presentation? Oh, here's my
1:40:09 presentation. Look at this.
1:40:12 Preview. Play slideshow.
1:40:16 All right. So, whoever asked about this
1:40:18 before, what's an agent do?
1:40:22 It went and did some research. Even
1:40:23 though it hung and I had to kick it, we
1:40:25 now have a deck. Now, is this deck any
1:40:27 good? I don't know. Let's go look at it.
1:40:31 Um, one thing one thing that I do know
1:40:32 about uh chat GPT's presentations is
1:40:36 they're ugly as sin. They're just really
1:40:38 bad uh visually. Cost cutting drivers
1:40:41 key trends and operations opportunities
1:40:43 for storyvine go to market strategy. Why
1:40:46 cost cutting? Almost 60% of pharma
1:40:48 executives plan to optimize operations
1:40:52 and reduce costs employing
1:40:54 restructuring, offshoring and layoffs.
1:40:56 60%
1:41:00 Um
1:41:05 6,000 jobs cut from Merc. Yep, that just
1:41:08 happened. Mona, so these are real.
1:41:12 10,000 out of Mona. Biotech.
1:41:15 5,000.
1:41:20 Generative AI and automation.
1:41:23 Content cost reduction.
1:41:26 Generative AI adoption yields up to 12%
1:41:29 savings. AI could generate 60 to 110
1:41:33 billion annually. Content creation costs
1:41:37 drop 30 to 50%. Which is that's
1:41:41 that's right where our company plays.
1:41:47 Digital marketing of engagement video
1:41:49 content retains 95% of a message. Text
1:41:53 only 10. This these are good numbers.
1:41:56 84% purchase conversions,
1:42:00 56% increase in ROI for video
1:42:03 impressions, digital training,
1:42:05 decentralized trials. There's
1:42:08 opportunities for StoryVine. Go to
1:42:10 market strategy. Target the leaders
1:42:12 digital training.
1:42:15 All right. Conclusions and next steps.
1:42:18 Pilot StoryVine with marketing and
1:42:20 training teams. Integrate into
1:42:21 decentralized trial platforms. There you
1:42:24 go.
1:42:25 All right. Not great, but fine. Fine.
1:42:30 Fine. Fine. There was some There was
1:42:32 some new and interesting stuff in there.
1:42:35 All right, everybody. I'm getting a
1:42:37 little tired. We You having fun tonight?
1:42:40 Hey. Yeah.
1:42:59 All right, let's see. Let's see what
1:43:02 we've got here.
1:43:05 [Music]
1:43:07 Any comments? Any thoughts, questions?
1:43:12 All right, everybody. Uh, I'm going to
1:43:14 take my new fancy microphone to bed with
1:43:16 me.
1:43:20 Hello, little microphone.
1:43:23 um homework for this weekend.
1:43:28 Please, we don't need to hear that.
1:43:31 Me touching the microphone. Yeah. The
1:43:33 one thing I noticed about this
1:43:34 microphone is it's got really bad
1:43:39 um if you touch it like like good
1:43:41 microphones if you touch them, you don't
1:43:43 hear anything. This one's got really bad
1:43:45 versions of that. No, you mean Oh, mic
1:43:49 in bed.
1:43:52 Yeah, you know, great demo, Kyle. Thank
1:43:54 you. Thank you,
1:43:57 Champion Irregulars. All right, let's go
1:43:59 look at Champion Irregulars.
1:44:04 Hey, did you all go um connect with
1:44:07 people, give them thumbs up, and and
1:44:09 respond to their introductions in the
1:44:12 salon? If you didn't, please do. If
1:44:14 you're an irregular, please go welcome
1:44:16 some people in. Um, okay, the
1:44:18 Irregular's channel. Let's go look at
1:44:20 Ireulas. The Irregulars.
1:44:27 Yeah, Champy's cousin.
1:44:30 That does look like Champy. Although
1:44:32 this is like the skinny version of
1:44:33 Champy.
1:44:35 Champy's face is squattier,
1:44:38 but very cute. Oh, wait. I'm not
1:44:40 sharing. Share this tab instead. Look
1:44:43 how cute. Yeah.
1:44:48 Nice,
1:44:51 nice, nice. What brand is it? It is a
1:44:54 Mayono. The Mayo Mayono 300
1:45:02 300 DX or 300X.
1:45:05 Yeah, Mayono 300X.
1:45:09 And it came with a little swing arm.
1:45:12 And I was going to get the 400X,
1:45:16 but
1:45:17 This is newer, so I got it. I got it.
1:45:21 Yeah. Here you go. All right, people.
1:45:26 Tik Tok question.
1:45:29 Do you have
1:45:32 a site we can check to learn about
1:45:34 setting up agents? Um,
1:45:37 I do not. What I can tell you is, well,
1:45:42 okay, setting up agents. I don't know
1:45:44 what setting up agents means. One of the
1:45:46 things that's happening right now is
1:45:48 everyone is calling everything an agent.
1:45:52 An agent might be like a chatbot that
1:45:55 you interact with like an agent like
1:45:58 that. That's that people are calling
1:46:00 those agents. People are calling
1:46:02 automations agents and then people are
1:46:04 calling agentic tools agents.
1:46:08 The only thing that are really agents
1:46:10 are the agentic tools. Um so it depends
1:46:13 what you mean by the question. None of
1:46:16 the agentic tools in my opinion are
1:46:18 really ready for prime time right now.
1:46:20 They're they're all kind of in
1:46:22 experimental mode. So what I would say
1:46:24 is the the four things I would recommend
1:46:27 playing with are Chat GBT agent mode,
1:46:32 Genpark,
1:46:34 Manis, Mus,
1:46:36 and then the Perplexity Comet browser.
1:46:39 if you can get access. I think it's in
1:46:41 it's not open beta right now, but it's
1:46:43 in like invited beta. People people who
1:46:46 have access to it might be able to get
1:46:47 you an invite code. Um,
1:46:51 those are the four that I'd play with
1:46:53 right now. There are probably a lot of
1:46:55 other ones, but those are kind of the
1:46:56 major ones that you can at least start
1:46:58 to see
1:47:00 the state-of-the-art for agentic
1:47:03 workflows for automate fully automated
1:47:06 things that are going out and using
1:47:08 tools on your behalf. Um,
1:47:12 and then I would just say be patient
1:47:14 about it. Um, if you want to do things
1:47:17 like build automations, there are lots
1:47:19 and lots of courses and workshops,
1:47:22 people offering that. Um, and then if
1:47:26 you're talking about agents where you're
1:47:27 interacting either with voice or text or
1:47:30 like a a video call, those kind of
1:47:32 agents, that's a whole different
1:47:34 category.
1:47:36 Naden has tutorial on their on their
1:47:39 site. Yeah, a lot of them probably have
1:47:41 tutorials. There's probably a lot of
1:47:42 YouTube videos. Quite frankly, you can
1:47:45 just use chat GPT. Put chat GPT in voice
1:47:48 mode, turn on the camera, and point it
1:47:51 at your screen and have it give you
1:47:53 instructions on how to set those things
1:47:54 up. That might be a way to do it as
1:47:56 well.
1:47:57 Thank you for the suggestions. Love what
1:47:59 you're doing. Very inspiring. Thank you
1:48:01 very much. I appreciate that. Very, very
1:48:03 cool. Um, cool. All right. So, it is the
1:48:06 weekend. I hope you guys have a
1:48:07 fantastic weekend. I would continue if I
1:48:10 were you. I would continue to explore,
1:48:13 start pushing all the buttons in chat
1:48:15 GPT. Let me let me actually share my
1:48:18 screen and show you what I would play
1:48:19 with over the weekend. Other than my
1:48:23 microphone.
1:48:26 Okay.
1:48:28 Um
1:48:30 sharing this tab.
1:48:35 All right. So, here we are in chat GBT.
1:48:39 First thing I would I would experiment
1:48:41 with if I were you, if you haven't done
1:48:43 it yet, if you go to your settings
1:48:46 and go to customize chat GBT, this is
1:48:49 what's called your um custom
1:48:51 instructions. You can play around with
1:48:54 all of these things, but there's one one
1:48:57 little hidden right hidden in the middle
1:48:59 of this is this little line right here.
1:49:02 What personality should chat GPT have?
1:49:05 And if you click on this pulld down
1:49:07 menu, there's five choices. Default,
1:49:10 cynic, robot, listener, and nerd.
1:49:14 Take the same prompt. So, take a prompt
1:49:17 that you've used before or take a prompt
1:49:19 where you know you're wanting to get a
1:49:21 particular kind of answer.
1:49:24 Use the same prompt
1:49:27 flipping into these different modes and
1:49:29 just look how it changes how chat GBT
1:49:33 responds. What you might find is that
1:49:36 one of these personalities changes chat
1:49:39 GPT into something you absolutely love
1:49:42 where otherwise you might not have liked
1:49:43 it or vice versa. There might be one of
1:49:46 these that you absolutely just hate it.
1:49:48 Um someone today told me they were like
1:49:52 um they were like why is chat GPT being
1:49:54 so mean to me? Oh Dr. J Claire was like
1:49:56 all of a sudden chat GPT started being
1:49:58 really tur and and sort of snotty to
1:50:02 her. She's like, "What's going on? Does
1:50:04 it not like me anymore?" She had it set
1:50:06 to robot efficient and blunt, right? And
1:50:10 she forgot to turn it off. So, go play
1:50:11 with those. That's a thing I'd
1:50:13 experiment with. Then the other thing is
1:50:16 um
1:50:18 in your in your model chooser, the
1:50:21 default mode is auto, right? And so what
1:50:24 happens in auto mode when you interact
1:50:26 with chat GPT, it is deciding which of
1:50:29 the different models to use.
1:50:32 What I would do if I were you is
1:50:35 manually switch into these different
1:50:37 modes and again take the same prompt and
1:50:42 apply it in different modes and see what
1:50:45 answers you get. See the different kinds
1:50:46 of answers you get. You can also under
1:50:48 legacy models, you can go back to GPT40,
1:50:52 which is how chat GPT used to used to
1:50:54 act,
1:50:56 but then use GPT5 instant, GPT5 thinking
1:51:00 mini if you have it, GPT thinking if you
1:51:04 have it. They're very, very different
1:51:06 models. My favorite right now is
1:51:09 thinking mini. I don't know why. It's
1:51:11 just I like the answers it gives me and
1:51:15 its creative writing ability is way
1:51:17 better than uh GPT5 instant which sucks.
1:51:22 I think this model sucks. Thinking Mini
1:51:24 is the one I like. But
1:51:27 that's what I would do is just keep
1:51:29 playing and experimenting with GPT5
1:51:31 because we're all trying to figure out
1:51:33 how this works. What model are you using
1:51:35 for your custom GPTs? I have not even
1:51:37 touched my custom GPT. Chef Clly, I will
1:51:41 probably flip them into
1:51:44 I'll probably leave most of them in
1:51:46 GPT40 mode if I can. And then I think
1:51:50 there's a point at which they
1:51:51 automatically flip over. I don't know
1:51:53 which one I'll use. I might use Thinking
1:51:55 Mini. I just I like that one better
1:51:57 right now. I feel like their GPT5
1:52:00 Instant if it it feels like it must not
1:52:03 be a
1:52:05 My guess is they have a better version
1:52:08 of that model. that's too big for them
1:52:10 right now because they don't have enough
1:52:11 GPUs. So, I have a feeling that's a
1:52:13 crappy that's a crappy model. So, that
1:52:17 thing may get better over the next
1:52:18 couple of months, but right now it
1:52:20 sucks. So, I don't know. I haven't
1:52:21 really thought about my custom GPTs.
1:52:24 I can only pick GP T5 thinking. No Mini
1:52:27 for me. Yeah, I don't I honest to God
1:52:29 don't know why. Um on on one of my calls
1:52:32 today, I had a couple of people on the
1:52:35 call had access to Mini and some did
1:52:37 not. I don't know why. Again, this is
1:52:39 probably OpenAI
1:52:43 doing stuff to manage their GPU usage is
1:52:46 my guess. Um,
1:52:49 we'll be back Monday. Tuesday, there's
1:52:52 no AI salon, so it's just a normal week
1:52:54 next week. So, have a great weekend,
1:52:56 everyone. And, uh, I will see you
1:53:00 I will see you Monday. All right. Have
1:53:02 fun this weekend. Peace out.