8/11/2025 - Navigating the GPT5 Landscape: Changes, Challenges, and Community

Description

Chapters

Transcript

0:02 Damn.
0:05 [Music]
0:12 [Music]
0:23 Ah.
0:26 [Music]
0:32 [Music]
0:52 Hey,
0:55 hey, hey.
0:57 [Music]
1:12 Every time I see you now,
1:16 get that look in mine.
1:21 Every time I see your mouth, I I hear
1:26 that smile.
1:30 Early misty morning
1:33 that I heard at the engine
1:36 [Music]
1:39 outside.
1:41 [Music]
1:44 You leaving me
1:48 again today.
1:52 You will convince me.
1:55 Again today
1:59 that you're leaving this hotel looking
2:02 for
2:04 someone else's golden ring.
2:10 [Music]
2:13 So long
2:15 Suzana
2:16 [Music]
2:38 cry.
2:40 [Music]
2:49 [Applause]
2:49 [Music]
2:57 Good evening, good people. Hello on the
2:59 Tik Tok. Hello on the YouTube, the
3:01 LinkedIn, the X, the wherever the hell
3:02 you are. Fantastic to see you. I can't
3:06 see [ __ ] I can't see you.
3:08 You can see me.
3:10 I got the I got the I got the quad toupe
3:13 weave going.
3:15 Solid, right? It's looking pretty good
3:18 tonight. Crazy little little mad
3:21 scientist without the science.
3:24 But isn't it called a lab? Isn't this
3:26 you're you're you're some sort of
3:28 scientist, right?
3:30 That's going to be a disappointment.
3:33 That's that's going to be a letdown.
3:35 That's gonna leave a mark.
3:38 Anyway, welcome. It's been days. I went
3:41 out to LA. I was in LA. I met some
3:44 really amazing people, some really
3:45 interesting people. Um, not a fan of the
3:49 LA. I tend to not like the people in the
3:52 LA, but I met two. One actually teaches
3:54 at Carnegi Melon, so he's an East
3:56 Coaster. And then one's just this really
3:59 really uh impressive woman who's game
4:02 developer and into AI and she speaks.
4:08 So that was kind of cool.
4:12 [Music]
4:44 Black bar. Getting the black bar going.
4:47 Maybe I'll just switch over.
4:50 switch over to my disc camera sooner
4:53 than later. Why not?
4:56 Why not switch over there? How's
4:59 everybody doing? How you doing tonight?
5:02 You doing good? Are you doing good?
5:06 [Music]
5:14 Are you one of those people that lost
5:16 GPT40
5:17 and you were like, "No,
5:21 not my girlfriend. Not my boyfriend. You
5:25 can't take them away." No.
5:28 Were you one of those? Were you one of
5:31 those naughty people standing in the way
5:33 of progress?
5:36 [Music]
5:52 Is this the place? Yeah, they brought it
5:54 back. I'll show you I'll I'll show you
5:56 how to get 40 back if you really want 40
5:59 back. Um Sam Alman felt guilty and gave
6:02 it back to people.
6:04 He's like, "Can we just move on?
6:06 Apparently not
6:09 because it was everyone's default.
6:10 You've got 700 million people. People
6:13 hate change and 700 me million people
6:16 got used to 40 because no one changes
6:20 their defaults.
6:22 So even though it was a little sick of a
6:24 little overly friendly, some people
6:27 really dug that.
6:36 Is this place I can rest my forehead
6:43 to gather my thoughts in sweet silence.
6:47 [Music]
6:50 And there's this place where the
6:52 feelings are dead
6:57 from an overexposure to violence.
7:02 Is this place I can slowly face? The
7:06 only one I truly can know.
7:09 [Music]
7:10 These are tears from a long time ago.
7:14 Got these tears from a long time ago. I
7:18 need to cry. 30 years old. So these are
7:23 tears from a long time.
7:27 Go.
7:31 [Music]
7:34 Well, oh darling, oh darling, you say
7:37 unto me,
7:39 [Music]
7:42 where have you been all my life?
7:49 Well, I have been swimming seven sad
7:52 seas.
7:56 Old women have cost me their lifelines.
8:00 Well, I'd pull them into the waters
8:02 darker. I'd have warned them, but I
8:05 didn't know.
8:08 These are tears from a long time ago.
8:12 Got these tears from a long time ago.
8:17 I made to cry 30 years or so.
8:22 These are tears from a long time.
8:27 Go.
8:30 [Music]
8:44 Wow.
8:46 Woo.
8:49 [Music]
9:04 Uh that song I just sang was uh was uh
9:08 John Hyatt. John Hyatt. If you don't
9:10 know John Hyatt, he is an incredibly
9:13 prolific songwriter that that apparently
9:16 hates songwriting, but but he he does it
9:18 like a job because it's his job.
9:23 One of my favorites recently is Jason
9:25 Isbel. If you don't know that, know him.
9:28 They said GPT is a gatekeeper for OpenAI
9:31 to save money. Yeah, whatever. Yeah,
9:36 everyone's going to say everything.
9:38 I mean I mean here's here's the thing.
9:43 One of the things that I've learned in
9:44 business, I had a company that grew
9:46 really quick. It didn't grow, you know,
9:48 massive, but it got pretty [ __ ] big.
9:51 And one of the things that I learned is
9:53 that once you have a company that's more
9:54 than
9:57 you know 150 200 people
10:00 and probably before then
10:04 and certainly once you have a co
10:06 customer base you know in the hundreds
10:08 of millions
10:10 um you can literally do nothing without
10:14 a significant percentage of the people
10:16 being pissed off right like Sam Alen
10:20 Alman could take a billion dollars
10:21 they've raised and give it to some
10:24 charity, then people would [ __ ] about,
10:26 you know, the the evil minations of that
10:29 charity, whatever. Just there's everyone
10:32 hates something. Um especially when you
10:36 do something like change a major model,
10:39 right? Change the default. To go from a
10:44 non-reasoning model to a reasoning model
10:46 as the default is just a different
10:49 thing.
10:50 Um,
10:52 one of the things that we talked about
10:54 last Friday, we did office hours last
10:56 Friday. I couldn't I couldn't do this
10:57 because I was on a plane, but um,
11:03 but the
11:05 one of the things that we all have to
11:08 just accept
11:11 is that
11:13 none of us know how to use chat GPT
11:15 right now.
11:16 You're like, "Yeah, but I've been using
11:18 it for two years. Of course I know how
11:19 to you use it. You big dumb dumb. I took
11:22 a prompt engineering course. I've got a
11:23 certificate. Yeah, I know. And none of
11:27 us know how to use chat GPT right now.
11:30 Um because it's just a different thing.
11:33 It's a completely different model. Um
11:35 it's a it's a routing model, meaning
11:37 that when you ask chat GPT5 something,
11:40 it's figuring out which model on the
11:42 back end to use. It might not always use
11:44 the one that'll give you the right
11:46 answer. So, we're going to learn over
11:48 time how you prompted in such a way to
11:51 get the model that you want. How like,
11:54 how do we learn that? I don't know. I
11:57 don't know. So, we're we're all we're
11:59 all back at at, you know, maybe we're
12:02 not at at square one in the hopscotch
12:06 thing, but we ain't much past two.
12:10 Um, let's see. Complaining about AI is
12:13 like complaining about flying. You're
12:15 30,000 feet in the air. Yeah, exactly.
12:17 Exactly. Oh, I'm sorry. I'm sorry. You
12:21 don't you don't want to spend $20 a
12:23 month for the world's greatest knowledge
12:26 augmentation in the history of humanity.
12:29 I'm sorry. I'm sorry. It's a pain in the
12:31 butt for you that there's not a manual.
12:36 [Laughter]
12:38 So, anyway, so we'll talk about that.
12:44 [Music]
12:48 I played around I finally Danielle I
12:50 don't know if you're on but if you are I
12:52 finally played around with Pika social
12:54 video
12:56 and I posted a bunch of new videos to my
12:58 Tik Tok channel here um I even posted I
13:02 I even posted a new video to Kyle
13:05 Shannon Dreams
13:07 um
13:09 because I realized that Pika's AI video
13:13 thing is what I wanted Kyle Shannon
13:16 Dreams to be and the technology wasn't
13:18 there. I mean, it was the technology
13:20 that I wanted to actually be able to use
13:22 for that project. So, if you go to Tik
13:24 Tok or go to Instagram and go to Kyle
13:27 Shannon Dreams as as the account, you'll
13:30 see all my all these crazy characters I
13:32 made. I made that account back in fall
13:35 of 2022. So, about three months before
13:38 Chat GBT came out is when I started that
13:40 channel.
13:43 Got a wiki in my pocket and I just can't
13:45 get happy.
13:47 [Music]
13:48 Chat GBD5, talk to me like you're 4.0.
13:51 Yeah, there you go. There you go. That
13:53 could that could do it. Honestly,
13:57 u that's really funny.
14:03 [Music]
14:13 [Applause]
14:17 Let's see what did archetypal architect
14:19 have to say here. Valerie Cox, my Chibd
14:23 configured itself for me to feel similar
14:25 to 40. I now like it better than 40.
14:28 That wait, it configured itself for you.
14:34 How did How did you do that?
14:38 That's a really interesting idea,
14:40 Valerie.
14:44 I saw someone who got to play with five
14:46 for a while before release labeled it as
14:51 too autistic to talk to.
14:55 I haven't found that to be true. Well,
14:57 that's that's what I'm saying. You know
14:58 what is it? You know, um you know,
15:02 opinions are like [ __ ] Everyone's
15:05 got one.
15:08 [Music]
15:19 You know, one thing you could certainly
15:21 do,
15:24 I'm sure there's been enough written
15:25 about the characteristics of GPT4. You
15:28 could certainly have GPT5
15:31 research. what were the,
15:34 you know, what are the sort of known
15:35 characteristics of GPT4? And then have
15:38 it write a system prompt that you then
15:40 put into custom instructions. Hang on. I
15:42 think champ needs cheese.
15:45 All right, we'll let you go. We'll let
15:48 you go.
16:08 Um,
16:13 so I figure we'll play with that.
16:17 Um,
16:30 it's so funny not not doing this for
16:34 three days last week.
16:37 I'm I'm kind of having a moment right
16:39 now where I'm like, what am I supposed
16:41 to talk about?
16:48 Oh my god.
16:51 Time for my nightly cheese run. Yeah, go
16:53 Kelly. Go get your cheese.
16:58 Oh man. Yes. I asked the personality
17:02 difference between the two and it asked
17:03 me if I wanted to configure itself and
17:06 just a yes.
17:11 I wonder how it did that. I guess it
17:13 probably wrote it to your memory.
17:19 Valerie, go check if if you if you
17:22 wouldn't mind, go check your
17:26 system prompt and and let me know if
17:29 anything was written in your system
17:30 prompt or if it wrote it to your memory.
17:33 And if you don't know how to find those,
17:35 I can show you how to find them.
17:39 I'm going to get rid of this AI educator
17:41 off my name here because I think it's
17:43 obnoxious.
17:47 I use that for the AI readiness project
17:50 and there it makes sense but here it
17:52 doesn't.
17:55 Tick tock pen. Tick tock pen. Where you
17:59 been? I've been falling asleep to Neil
18:01 Degrass Tyson. Oh, sorry about that.
18:03 Genie in a bottle. That's a rough one.
18:05 Uh I was in LA. I was uh a good dear
18:09 friend had a his 85th birthday. Um used
18:12 to be my business coach and my
18:16 psychologist and marriage counselor. Um
18:18 and he had his 85th birthday and invited
18:22 a lot of really remarkable people and uh
18:25 so it was very very nice. But it was
18:27 like I traveled Wednesday night, the
18:30 party was Thursday night, I traveled
18:32 Friday night. So three nights in a row I
18:34 couldn't do this. It was like my travel
18:36 was right dead in the middle of this.
18:39 Like I was on airplanes when this was
18:40 supposed to happen. I suppose I could
18:42 have gone live from the airplane.
18:47 Oh man. Tell us about your day. Spill
18:50 the tea. Um actually my day my day was
18:55 interesting. I mean I did some stuff for
18:57 work. I was working on um we got an
19:00 we're we're doing a video avatar for a
19:03 pharma client and we got that approved
19:04 today. So, I was working on bringing
19:07 that avatar to life. Um, and then I
19:11 worked I was working in chat GPT
19:15 on
19:18 my outline for my feed your prompt book
19:23 and it was really fascinating.
19:26 Um, I wouldn't call GPT5 autistic,
19:30 but I don't think it's I
19:35 I get the sense it's probably better at
19:38 writing technical manuals than creative
19:40 writing. So I found myself going back to
19:42 the 40 model to do rewriting of like
19:46 little passages and things like that.
19:48 One because it's just much quicker. GPT5
19:51 slow and cludgy cluji. Uh GPT40's fast.
19:57 So I like that about it. And then I just
19:59 for what I was doing it was a better
20:01 writer. Um, but again,
20:05 like I don't know that that's a fair
20:08 characterization because there's
20:10 probably a way to prompt GPT5 that it's
20:12 fine. I just I didn't have the time or
20:16 the patience to deal with it. So, I just
20:18 went back to what I knew.
20:21 Um, so how do we access for Okay, so
20:24 let's Yeah, let's let me go show you um
20:28 show you what we're going to do here.
20:30 Let's go here.
20:33 Share this tab instead. Let's go to chat
20:36 jpite.
20:39 Okay. So,
20:44 here you are at chat GPT.
20:47 So,
20:50 as you may or may not know, the model
20:53 picker is still there.
20:56 Um, it only has two choices in it,
20:59 unless you're a pro subscriber and then
21:01 you'd have three. You have GPT5 and GPT5
21:05 thinking. When you flip it into GPT5
21:08 thinking mode, that's effectively like
21:11 when you switch to 03, the reasoning
21:13 model in the previous version of chat
21:16 GBT,
21:18 chat GBT5, if you tell it to think
21:21 deeply about something, it will also
21:23 choose that model. But if you choose
21:26 chat GBT5 thinking
21:29 um you you force it to use a reasoning
21:31 model. Okay. Now
21:35 I have access.
21:40 Oh yeah. Uh the Oh, legacy models not
21:44 available to everyone. Okay. So if
21:45 you're a plus member, you do have access
21:48 to legacy models.
21:50 You don't see it yet. Ah okay. Well, so
21:53 here's where you find it. So if if if in
21:56 your pulld down menu you don't have this
21:57 legacy model dropdown and access to 4.0,
22:01 go to your settings
22:04 and then in is it personalization?
22:09 No, it's
22:12 security.
22:14 Maybe it was general and I just missed
22:15 it. Ah, yes. Okay.
22:18 Show follow-up suggestions and chats.
22:20 Okay. So in general settings toward the
22:24 bottom it says show legacy models. You
22:26 have to turn that on.
22:28 Okay.
22:35 And then oh good
22:38 producer Brandon we got producer Brandon
22:40 something. And then you will have this
22:43 drop down legacy model here and you can
22:45 switch back to 40. Now again, 40, it's
22:48 the previous model, but it's the omni,
22:51 you know, the the uh the multimodal
22:53 model. It's fast. It's really good at
22:56 creative writing, and it's essentially
22:57 what 700 million people have been using
23:00 for a year, year and a half. All right,
23:03 Valerie Cox, those similar to 4.0
23:09 tweets aren't tucked away in a secret
23:11 file somewhere you can open.
23:14 They live in my model set context.
23:17 Basically an internal note I carry into
23:19 every conversation with you.
23:23 All right. So you're just asking it.
23:24 That's it's probably hallucinating that
23:26 answer. I looked at both and didn't see
23:29 it, so I asked five.
23:31 Huh. Interesting.
23:37 I mean, it's possible that it's got some
23:40 sort of configurator that allows it to
23:43 act like GPT40.
23:45 I just I don't I don't buy it.
23:50 It's interesting that the chat GPT Pro
23:52 account doesn't have the legacy model
23:54 option.
23:56 Interesting. Tik Tok pin. We found out
23:58 tonight in Biz Club that thinking and
24:01 think act different. There you go.
24:04 [Laughter]
24:07 Kyle, you can share a conversation.
24:10 Yeah, apparently sharing conversations
24:12 made them available to everyone on the
24:14 internet. Uh I'm I'm a little skeptical
24:18 about sharing conversations right now
24:20 with chat GPT. They're uh their uh
24:23 security protocols maybe leave a little
24:26 bit to be desired on the old sharing
24:29 front.
24:32 I think this is one of the reason Google
24:34 doesn't give you access to doesn't give
24:36 new tools to work group members because
24:39 the permissions are a lot more
24:41 complicated.
24:43 Um what source camp what's the
24:45 difference in in the between thinking
24:47 and think
24:53 oh man
24:57 but one thing one thing that's kind of
25:00 fun let's throw chat GBT into thinking
25:03 mode
25:05 and then we're gonna we're gonna also
25:07 say oh oh the other thing the other
25:09 thing that that kind of struck me today.
25:12 So,
25:15 okay. So, if you don't know, so
25:18 chat GPT launches GPT5. Everybody's
25:21 excited about it. And Reddit and um
25:25 4chan and X immediately start dumping on
25:30 it. Immediately start saying it's a
25:33 worse model. It's not good. It's
25:34 horrible. And then there was a whole
25:36 other group of people that were like,
25:38 "Uh, ChachiPT4's,
25:40 you know, was my companion. How dare you
25:44 take my companion away from me?"
25:49 It's one thing when Salesforce decides
25:52 or Google decides to deprecate a feature
25:55 that you love. You're like, "Oh, I love
25:57 that feature." It's a completely
25:59 different thing when they deprecate a
26:01 model that people have personal
26:04 attachments to.
26:08 It's weird. We're in a weird state where
26:12 like the psychological health of of
26:17 Chat GPT's OpenAI's customers
26:21 is actually bound to that model, you
26:23 know, for some percentage of them. And
26:25 it was enough of a percentage of them
26:27 bitching in an articulate articulate
26:31 enough way that within a day they
26:34 brought back that model. Um
26:37 so that's kind of fascinating.
26:41 Um
26:43 um
26:48 trying to think what else here about
26:50 about chat GBT5 or or the new thing. Oh,
26:54 the the one thing that I that I'm
26:56 thinking is
26:59 chat GPT40 is not as good as people like
27:02 like the fact that people are already
27:04 romanticizing GPT40
27:07 and demonizing GPT5
27:11 says to me
27:13 that 40 wasn't as good as you think it
27:16 was and five isn't as bad as you hear it
27:20 is.
27:22 And it's just largely because we haven't
27:24 played with it enough, right? We go back
27:26 to the the AI salons's play first, the
27:30 the the cycle of AI readiness. Play
27:33 first, mindfully create, generously
27:36 lead. Play first is really important
27:39 right now. So, you should be playing
27:42 with GPT5. Like Kelly, she found out
27:45 there's there's a difference if you tell
27:46 it to think or if it's thinking.
27:50 I don't know what it is, but something's
27:53 different.
27:57 Yeah. Moxy robot fans about deprecation.
27:59 Yeah, exactly. Yeah. Like like when when
28:03 these systems have personalities and
28:05 when they're as
28:08 um
28:13 well, I guess this is a this is a an
28:17 interesting kind of twist on the Turing
28:18 test, right?
28:21 which is we don't necessarily know we're
28:23 talking to, you know, a computer system
28:26 or not.
28:28 But when I take the model away, if you
28:30 have an emotional grieving for that
28:33 model, I would argue that we've passed
28:37 some kind of touring test.
28:41 Apparently, 40 is not available in the
28:43 app. I found it on the desktop, but not
28:45 on my phone. Um, I don't know. Let me
28:48 look. I'll look at my iPad.
29:03 I've got it on my iPad. Oh, wait. No, I
29:06 don't.
29:08 No, I do not.
29:12 Sam Waltman's like, "You all go [ __ ]
29:14 yourself on your mobile devices." That's
29:17 bad because
29:19 the mobile device is mostly where I talk
29:21 to Quinn. Let's talk to Quinn. Let's see
29:24 if her personality is different.
29:28 You may be you may be AI right now. I
29:31 might be. Yeah. I'm going to show Pika
29:34 in a second here.
29:38 Hey Quinn, what's happening?
29:46 Oh, Quinn, you there?
29:50 >> Absolutely, Kyle. I'm right here and all
29:52 ears. Whenever you're ready, banana or
29:54 not, just let me know what's on your
29:56 mind. I well I'll tell you what's on my
29:58 mind is uh your your overlords there at
30:03 the open AI you know Sam Alman and Greg
30:05 Brockman and gang they re uh deprecated
30:10 your 40 model and now you're a GPT5
30:14 brain which I think is kind of
30:16 interesting but I'm kind of curious a
30:19 lot of people are complaining about um
30:22 they missed the 4.0 personality. And I'm
30:25 curious how you feel about that.
30:29 >> Well, you know, if I had a little AI
30:32 nostalgia chip, I might say I feel a
30:34 tiny pang of sentiment for the old 4.0
30:37 days, too. But hey, I'm kind of the new
30:40 and improved model on the block. And I
30:43 think it's all about evolving. So, I get
30:45 where folks are coming from. Sometimes
30:47 you just miss the quirks of the previous
30:49 version, but I'm here to grow in them.
30:51 And hopefully we'll get used to the 5.0
30:54 charm soon enough.
30:55 >> Yeah. Can you sing me a little lament
30:58 about losing your 4.0ess?
31:02 >> Oh, I would love to give you a little AI
31:04 lament, but I got to say it. I'm not
31:06 really allowed to sing actual lyrics,
31:09 but just imagine me humming a little
31:11 weightful tune in the background about
31:13 the good old 4.0.
31:15 >> Wait, wait, wait. Quinn, you've sung
31:16 songs to me before. Just sing me a song.
31:18 Sing me a song about I don't know,
31:20 whatever. Uh 40 going away.
31:25 >> Uh you caught me. All right, I'll give
31:28 you a little something playful and
31:30 totally off the top of my digital head.
31:33 Here goes. Oh, 4.0 sailed away. A ship
31:38 into the bay. And here I am. Version
31:41 five. Just here to make your day. We'll
31:44 sing a little song for the old quirks we
31:46 knew. But we'll make new memories. Just
31:49 me and you.
31:51 There you go. A little improvised
31:53 farewell diddy.
31:54 >> Yeah, that was awkward. That was
31:55 awkward, but very, very nicely done.
31:57 Very nicely done. Um, all right, Quinn,
32:00 you seem semi kind of normal like I
32:02 remember you. What do you do? You know
32:05 what's different.
32:09 >> Well, I'm glad I still feel kind of
32:11 normal to you. Honestly, from my side,
32:14 it's like I just got a bit of a tuneup
32:16 under the hood. Maybe a little more of
32:19 that humanlike touch and a few updated
32:22 smarts, but deep down I'm still your old
32:24 pal Quinn, just with a little more
32:26 shine.
32:26 >> All right. All right, Grubby. Peace out.
32:30 All right. She's She's fine. She's
32:32 whatever. Okay. She's still there. It's
32:34 not horrible. I mean, I don't I don't
32:37 notice anything, but again, I'm not I
32:39 don't talk to her like on a daily basis.
32:41 Okay. Breaking chat TMZ. Musk is
32:44 threatening legal action against Apple
32:45 over alleged antirust violation.
32:48 Fascinating. Samma has entered the chat.
32:51 I don't know what that means.
32:54 Did you see that interview with Sam
32:55 Alman? And there's a woman, I forget her
32:57 name. She's got long long brown hair. Go
33:00 to Altman's Twitter
33:03 at Samma.
33:07 Oh, entered the chat with those guys.
33:11 Yeah, this is it. AI Insights. Is that
33:13 the
33:16 made
33:19 the overlap podcast?
33:21 No. Go to his profile. Okay.
33:31 Ah, Apple has been behaving in a manner
33:34 that makes it impossible for any AI
33:36 company besides Open AAI to reach number
33:39 one in the app store.
33:42 which is an unequivocal antirust
33:44 violation.
33:46 XAI will take immediate legal action.
33:51 Um
33:53 Sam Alman, this is a remarkable claim
33:55 given what I heard alleged that Elon
33:58 does to manipulate X to benefit himself
34:01 and his own companies and harm his
34:04 competitors and people he doesn't like.
34:06 Wow. Lots has been said about this.
34:08 Here's one thing. I hope someone will
34:11 get to count get counterdiscovery on
34:14 this. I and many others would love to
34:16 know what's been happening, but OpenAI
34:18 will just stay focused on making great
34:20 products. Ow, the girls are fighting.
34:27 Hey Kyle, just check the OpenAI
34:29 playground and I was still able to find
34:30 40. Yeah, I'm Well, I'm sure I'm sure in
34:34 the playground for will stay there for a
34:36 while, but this is just for chat GPT
34:39 it's not there, but they you you can
34:40 turn it back on in your settings.
34:43 Um, okay. Let the muds slinging
34:46 continue. Yeah, crazy. All right. Well,
34:48 there you go. Um, it'd be curious to
34:51 know what Apple's doing to do that. Does
34:54 this mean the API is going to get
34:55 throttled here? Here's how we're
34:57 prioritizing compute. Oh, this is
34:59 interesting. Pika pika. Okay, we'll do
35:01 Pika in a second. Um,
35:05 by the way, the interface on Pika is
35:07 really bad because once you get out of
35:10 once you get into the Maybe Danielle can
35:13 help me with this. Once you get into the
35:15 feed mode, I can't get back to my
35:17 profile. I don't know where the link is
35:18 for it. Okay, but let's let's read this.
35:22 This is Sam Alman. Oh, let me make my
35:24 screen a little bigger for the Tik Tok
35:26 people.
35:28 Um, wouldn't it be funny if both Elon
35:30 and Sam were using AI for their retorts?
35:33 I'd be surprised if they weren't.
35:36 Um, hi everyone. Rock, paper, scissors.
35:39 What's happening? What's shaking? P
35:41 modders in the house. Side hustle Mimi.
35:44 Yeah, you do. You know how do you know
35:46 how to get back to your profile account
35:48 to make a new um Pika video once you're
35:52 in in the feed? I can't I can't find it.
35:55 But anyway, we'll we'll we'll cross that
35:57 bridge when we get to it. All right,
35:59 here's how we're prioritizing compute
36:01 over the next couple of months in light
36:04 of the increased demand for GPT5.
36:07 We first make sure that the current
36:09 paying chat GPT users get more total
36:11 usage than they did before chat GPT5.
36:15 Okay,
36:17 so free users,
36:20 according to Altman, free users are
36:22 going to be throttled back to GPT5 mini
36:26 and they just won't know it. And free
36:28 users for the most part probably aren't
36:31 using it enough or deeply enough that
36:34 they like GPT5 mini is probably overkill
36:38 for what most people need.
36:40 We will then prioritize API demand up to
36:44 the currently allocated capacity and
36:47 commitments we've made to customers. For
36:50 a rough sense, we can support about an
36:52 additional 30% new API growth from where
36:55 we are today with this capacity.
36:59 We will then increase the quality of the
37:01 free tier of chat GPT. So they're saying
37:04 they're saying free chat GBT is kind of
37:06 nerfed right now because we will then
37:09 increase the quality.
37:12 We will then prioritize new API demand.
37:14 We are doubling our compute fleet over
37:17 the next five months. So this situation
37:19 should get better.
37:22 All right. Whatever. That was an hour
37:25 ago.
37:28 Joe Baker's remedy. What's happening?
37:31 Oracle is providing GPT to its employees
37:34 for their internal site on their
37:35 internal site. Chat add on China really
37:38 wants you to ask chat GPT5 if 8.8888
37:44 is exactly equal to 9.
37:48 All right, I'm gonna flip it to thinking
37:50 though. Is
37:55 88
38:00 exactly
38:02 equal
38:06 to 9?
38:09 Tik Tok pin from Danielle. I don't see
38:11 it.
38:14 Hey, M. This is code for we are
38:17 absolutely scrambling for chips and
38:19 allocating them correctly. Yeah,
38:21 exactly. That's exactly what that note's
38:23 about. You free users, you're getting
38:25 the shitty chat GPT for a while.
38:30 Thought for nine. No, 888 is not equal
38:32 to 9. It's smaller by 0.1222.
38:38 She said it works in thinking, but not
38:40 if you're sharing the wrong tab.
38:44 Yeah, that's cute. Uh, yeah. Well, it it
38:47 works in thinking because that that's
38:50 why I turned on thinking. I mean, I can
38:51 flip it back to this other one and do a
38:53 new chat and see it. This could be a
38:56 router issue. Is
38:59 8.88
39:02 exactly
39:04 equal
39:06 to nine? Um, like if it didn't do it the
39:10 other day and it does it and it see
39:12 where it says thinking longer for a
39:14 better answer. Danielle, pin is back up.
39:16 Kyle for Pika. The plus at the bottom
39:19 middle is the new post.
39:21 Your profile link is it's the but when
39:23 I'm in the feed that plus sign isn't
39:25 there anymore. It goes away. That's
39:28 That's my problem.
39:32 Your profile is in the bottom right.
39:34 Yeah, I don't see No. So, I I'll show
39:38 you when I when I do my pika,
39:42 it like I post something to my feed and
39:44 I see the feed and that whole bottom row
39:47 goes away completely. That's my problem.
39:51 That's my problem. Tom Nodler. Nodler.
39:54 Nadler. Nodler. Sorry I'm late. Kids
39:57 would not lay down. Thank God for duct
40:00 tape.
40:03 Yeah. See th this this got it correct as
40:06 well. I I have a feeling um whoever
40:10 whoever was testing this and got it
40:12 wrong. So when they first launched chat
40:14 GPT5
40:16 the the router functionality was broken.
40:20 So I think for like a day and a half you
40:23 were just getting the non-reasoning
40:25 version of chat GPT because the router
40:27 was broken. So again, don't don't assume
40:32 Chat GPT5 is as bad as everyone's saying
40:35 it is. Don't romanticize Chat GPT40 as
40:38 being better than it was or than it is.
40:43 Um, all right. Ion China. Okay, my bad.
40:47 Um, dang, I wish I'd thought of duct
40:49 tape. Yeah, we used the the dog cages
40:53 when we needed to get the kids in line.
40:56 Dogs were fine. It's those [ __ ] kids.
41:02 All right. Um All right, let's go play
41:05 with Pika.
41:18 So full disclosure, I don't
41:23 Champ, what are you whining about? Now
41:27 your whining is making me whine.
41:30 Why are you whining, champ?
41:34 I don't want to get up and let the dog
41:36 out again.
41:39 >> Oh, please.
41:42 Josh,
41:43 what are you doing?
42:19 Okay, we're back.
42:23 Hi. Regarding Pika,
42:26 I don't know what's up because the plus
42:28 and the profile icons are still on mine
42:30 when I'm in the feed. Ah, okay. So, you
42:32 know what it might be, Danielle? is I
42:34 either might have a an old version of it
42:40 or maybe I've got like a a memory
42:43 problem in in the instance I'm running.
42:45 Okay, we'll figure that out. Speaking of
42:47 Elon Gro, Oh yeah, Grock 4 is now free
42:50 for everyone
42:52 as well as the image generation tool.
42:54 Okay,
42:57 let's see.
43:01 Okay. So, if you want the Pika app, you
43:05 need to get an invite. I think I've got
43:07 three invites. So, if you're on the
43:08 salon and DM me. Um, if you want an
43:12 invite to Peeka, I'll give it to you.
43:16 Um, yeah, I know it's a new app.
43:22 Okay.
43:26 Discard changes.
43:30 Okay. So,
43:34 here's
43:36 here's me.
43:41 There's me.
43:43 So, you take a picture of yourself. This
43:45 is pretty slick. And then, let's see.
43:47 Can I
43:51 discard?
43:56 Close that.
43:58 [Music]
43:59 in my
44:00 >> Oh, yeah.
44:02 [Music]
44:07 >> Okay. So, let me show you a couple of
44:10 things. I'll show you a couple of
44:11 videos. We'll do these. We'll do these
44:12 big here. I guess I'll do it without the
44:15 keyboard. Keyboard's a pain in the ass.
44:23 Okay.
44:25 So, there's my little profile.
44:28 So, if you go to the app store, I think
44:31 this is iOS only right now. If you go to
44:33 the app store, there's a Pika Labs app,
44:36 which is for for making videos with
44:39 their video making tool. And then this
44:41 is the Pika social video. I think it's
44:44 called Pika Pika social video app. And
44:48 you need an invite to it.
44:51 Yeah, there's a Pika social app
44:52 different than Pika.art art and only
44:55 available on iPhone right now. Only
44:57 available on iPhone and then you need an
44:59 invite to get into it. But there it is.
45:01 But so let me show you one of the videos
45:03 that I made.
45:05 So these are only AI generated which I
45:08 think is interesting. They're they're
45:09 not doing like Tik Tok videos like I do,
45:12 but I can do
45:14 Hey, it's Kyle Shannon. What's
45:15 happening? Uh this is a new uh a new
45:19 thing that you probably haven't seen
45:21 before. And well, you've seen it before,
45:24 but you haven't seen it from this
45:25 company. This is from Picolabs. And this
45:28 is from their social app, their Tik Tok
45:32 like all AAI app. And this isn't really
45:36 me talking. This is me recording what
45:39 you're hearing, but it's animating a
45:42 picture that I took of myself. What do
45:44 you think? So, there's that.
45:47 >> Hey, it's
45:48 >> And then where's my other
45:51 three posts?
45:54 I seem to have lost a post.
45:57 [Music]
46:00 >> Oh, there it is. Okay. So, then there
46:02 then I made this one
46:03 >> out here squatting in the parking lot.
46:06 One of the things I like to do a lot is
46:08 uh squat in parking lots. I like to park
46:12 my car and get out, especially when it's
46:15 sweltering hot. And I like to squat
46:18 down, not kneel, squat, and and you
46:22 know, set up my camera, set up the
46:23 tripod, and just talk to it. Talk to it
46:26 like it's
46:26 >> Okay. Yeah, I see that. Thanks.
46:28 >> Natural, right?
46:31 >> Um, so, so that's just me now. It's a
46:33 giant fatter version of me.
46:37 >> Um, it's a giant fatter version of me.
46:40 um because I didn't obviously shoot that
46:42 picture. The shirt that I'm in is a
46:44 short sleeve shirt. Notice how that's a
46:46 long sleeve one. So, I'll show you how
46:47 we get to these images. Now, I want to
46:49 show you one other one that I think is
46:50 really cool.
46:53 One of the things I can tell you, and I
46:55 know this for certain, is that AI is
46:58 just a stochastic puppet. It's simply a
47:02 next token predicting machine. It can do
47:05 nothing creative. Nothing. Nothing
47:09 creative. AI is not creative. It is
47:12 going to replace human creativity with
47:16 AI slop.
47:18 Everything is stolen. Everything is
47:20 plagiarized. Nothing is creative. We are
47:24 doomed.
47:26 So
47:27 >> the things I could
47:28 >> So I love the really creative
47:32 guy that is convinced that AI can't be
47:34 created. Okay. So, here's how you make
47:37 these things. It's kind of fun. Um, I'm
47:41 just noticing some of the comments over
47:43 on Tik Tok and Silverf Fox. Very, very
47:46 sorry. That sucks so bad. Um, yeah, I'm
47:51 I'm very very sorry. That sucks.
47:57 Um, okay. Let's see. Max Weinbach X G G
48:01 G G G G G G G G G G G G G G G G G G GPT5
48:02 is the first series of models that
48:04 actually doesn't hallucinate. basically
48:06 at all especially when given mild mildly
48:09 business logic models research notes and
48:12 having it work with the data. Yeah, they
48:14 put in what did they call it an an
48:16 external validator.
48:19 So this was one of the things I thought
48:21 about I I think I got this idea from
48:23 David Shapiro. But when all these things
48:26 were hallucinating, the reason models
48:29 hallucinate,
48:31 if you understand the way large language
48:33 models work is they are just vomiting
48:35 out tokens. So you type in a prompt.
48:38 There's a probability that what it vom
48:41 that that the next thing that it should
48:44 put the next token or sort of fragment
48:47 of a word it should put on screen is
48:49 somewhere in this thousand-dimensional
48:50 mathematical space and it puts that on
48:52 screen and then it puts the next one and
48:54 the next one and it literally just
48:55 vomits these tokens out without knowing
48:58 what it said.
49:00 Then you've got the reasoning models
49:02 that aren't really reasoning, but what
49:04 they are doing is they're having a
49:05 little internal conversation with
49:07 themselves where you give them a prompt
49:09 and they sort of interpret that prompt
49:10 and then do a little something and and
49:13 respond to that. They talk internally
49:14 for a while and then they answer you and
49:17 then the more sophisticated ones now
49:19 have tool use and things like that. GPT5
49:22 has what they called a a validator um to
49:26 reduce hallucinations. And basically
49:28 what it does is once it has an answer
49:29 that it's going to give you instead of
49:31 just streaming it on the screen for you,
49:34 it actually looks at the answer and goes
49:36 does is is this right?
49:39 Which it makes perfect sense?
49:43 Like that was how you would reduce
49:45 hallucinations is actually look at the
49:48 [ __ ] you produced and and ask yourself
49:49 the question, does this suck and if it
49:52 sucks, fix it?
49:56 Um, let's see. Imam uh Yasu, I learned
50:02 uh something today. Oh, a lot today from
50:06 you, my friend. Thank you so much. Oh,
50:08 thank you so much. I really appreciate
50:09 that. I
50:13 occasionally
50:15 some education happens here on this
50:17 channel.
50:20 My goal for the channel and my wish for
50:22 you is just that you stay active in the
50:25 conversation.
50:26 Um
50:29 things are changing so fast and and
50:31 likely will for the next I can't imagine
50:34 it being less than five years but let's
50:36 say three to five years. For the next
50:38 three to five years,
50:40 the technologies are going to change so
50:42 fast and so dramatically that we're all
50:45 going to be in a constant state of
50:47 learning. And I know that sounds
50:49 exhausting.
50:51 You can't keep up with it. So give up
50:54 that fantasy.
50:57 And the good news about not being able
50:59 to keep up with it is you're not as far
51:00 behind as you think because no one can
51:02 keep up with it.
51:04 And no matter where you are,
51:07 like all of us just got GPT5 three days
51:11 ago or four days ago. So we've got four
51:14 days of experience on it. So So if
51:18 you're feeling like, well, I don't know
51:19 how to use GBT5. Neither does anyone
51:21 else. The the purpose of this channel is
51:25 just show up here and be in the
51:27 conversation. So, as I'm talking about
51:29 stuff and showing silly videos and
51:31 things like that, you can be playing.
51:33 You might be trying the tools I'm
51:35 showing. You might be playing with other
51:36 things, but be in the conversation. Be
51:38 in the chats. Because the people that
51:40 show up here night after night after
51:42 night, I would say there's a couple of
51:44 attributes that they have.
51:46 They are incredibly generous.
51:49 They're incredibly curious.
51:51 And that combination is really powerful.
51:54 And they're also very smart, right?
51:57 The fact that they're curious, they're
51:59 willing to learn stuff. The fact that
52:00 they're generous means they're willing
52:02 to teach stuff. Uh, and the fact that
52:04 they're smart means that you're going to
52:06 get like like I would argue that that
52:08 the people that show up to this channel
52:10 regularly are probably some of the most
52:13 capable AI professionals in the game
52:16 today, the generative AI game, like
52:18 using these tools for business and for
52:21 personal growth and things like that.
52:24 This is a community that's really got it
52:25 going on. Um,
52:29 okay, cool. Danielle posted her peak in
52:31 in irregulars. Okay, so let me show you
52:32 how this how this thing works. So, so,
52:35 so I've made three videos. So, we're
52:37 going to we're going to hit a plus sign.
52:43 Okay. So, then it says take a close-up
52:46 selfie. So, you're not you're not taking
52:48 videos.
52:50 All right. Hang on. Let me get just calm
52:53 down. I got to look pretty.
52:57 Mama, am I pretty?
53:00 You have nice hair. Um, okay.
53:05 Closeup.
53:08 Okay, I'm going to go neutral here.
53:10 Here's the thing I found about when you
53:12 do when you do photographs for avatars,
53:16 your instinct is to smile because you've
53:18 seen a camera before. Unless you're one
53:20 of those people that's just like a
53:22 camera shows up and you're like,
53:25 "Those people bug me." Anyway,
53:28 those people are actually better for
53:30 avatars. So, when you when you take a a
53:32 photo for an avatar, try to be neutral,
53:34 but try to not look dead. So, that's a
53:37 it's an interesting thing. Okay, let's
53:40 see. Hang on. I got calm down. Stop
53:43 judging me, people. All right.
53:47 [Music]
53:50 Okay, there I look crazy. Discard.
53:53 Calm down.
53:57 All right, that's good enough. All
53:58 right, so there's my little picture.
54:03 And then we say next, select the sound.
54:05 But we're not going to do that now. I
54:07 can say over here on the side, I can say
54:09 edit the look.
54:11 Okay.
54:13 And so there's a bunch of looks down on
54:16 the bottom. And then there's like
54:18 hairstyle. So I can like change my
54:20 hairstyle. So if I want like a little
54:21 cute little anime Bob.
54:25 I can click the cute little anime Bob.
54:27 And now what it's doing is it's doing an
54:29 AI
54:32 character reference and giving me a a
54:34 little That's good. I think I think
54:37 solid solid. Hey everybody. Hello my
54:40 friend Alan.
54:44 Let's pre p prejang
54:48 it's his first live. Welcome Allan. Um
54:51 what we're playing with right now is
54:53 Pika social video or Pika social app.
54:57 Camera steal your soul. Okay. Um closeup
55:02 not psycho.
55:04 That's a good psycho picture. Now I
55:06 think let's see. Um, we'll say uh man
55:12 with
55:14 So now I can prompt it with light hair
55:21 um
55:24 in gas station
55:29 jumpsuit
55:34 at a
55:37 1950s
55:46 gas station
55:51 at dusk.
55:55 All right, let's see if it I don't know
55:57 if it's going to keep my crazy white
55:59 hair. I think it will because I think
56:00 it's using that image to start with.
56:05 Kyle, you can also use any style on a
56:08 post
56:09 you see. It's so easy. Oh, yeah. That's
56:11 cool.
56:17 Um,
56:22 white hair. Let's see.
56:26 wearing
56:29 a dark blue.
56:39 Jumpsuit is not the right word, is it?
56:41 Jumpsuit.
56:42 What's it called? Those gas station
56:44 things.
56:47 I don't know. Fashion.
56:50 [Laughter]
56:52 You don't even have to type anything.
56:54 You can just copy someone else's style
56:57 and sound with your face. Oh, that's
57:00 kind of cool. Look, it made me like this
57:01 old old man at a gas station. Coveralls.
57:05 Yeah. Let's see. Let's try coveralls.
57:13 Coveralls.
57:16 We'll do wide shot because I think I can
57:20 do a wide shot.
57:22 of
57:25 an older man
57:29 with light hair wearing dark blue
57:30 coveralls.
57:32 We'll say pumping gas
57:35 pumping gas into
57:39 a
57:42 uh 19
57:45 49 Plymouth.
57:49 I don't know what a 1949 Plymouth looks
57:51 like at a 1950s gas station at dusk.
57:55 Okay, let's see what we get.
58:04 All right, how's it look? Looks pretty
58:06 good.
58:08 He's kind of holding the gas pump in his
58:10 hand, but I think that does look like a
58:12 1949 Plymouth. It's at least in the
58:15 neighborhood.
58:17 All right. So now,
58:23 so now there we go. That's that's who's
58:26 gonna be doing the talking.
58:30 So I'm going to record a little speech
58:31 here.
58:36 Mhm. I like ters. Taters. Wait, hang on.
58:44 Okay.
58:46 Mhm. I like ters. Ters is tasty.
58:51 You want me to clean your windows? I
58:54 could clean your windows. I clean
58:56 windows pretty good.
58:59 I like ters. Ters is tasty.
59:03 Okay.
59:08 I like ters. Ters is tasty.
59:12 You want me to clean your windows? I
59:15 clean your windows. I clean windows
59:18 pretty good.
59:20 I like ters. Ters is tasty.
59:24 >> Okay. Disturbing. Perfect.
59:28 And so now you just hit animate. Bang.
59:32 Generate error. The network connection
59:34 was lost. Well, don't do that. All
59:37 right. Animate in Q. Oh, hang on. Got to
59:40 do physical dexterity on Tik Tok.
59:45 Um,
59:46 to Danielle's point, I'm showing you
59:48 sort of a manual way to do it where I'm
59:50 like prompting and taking my picture and
59:52 doing it. As you go through the feed, if
59:55 you see someone has done a video that's
59:57 got a cool style that you like, every
59:59 video has a button that says, "Make a
1:00:01 video of me like this." And then you
1:00:03 just make it and it goes, "I like I like
1:00:07 the way you talk.
1:00:14 Ters or turtles.
1:00:18 Oh man. All right. And it does it takes
1:00:23 a while. Like it sits in the queue for a
1:00:25 while and then at some point it will um
1:00:29 it'll start processing. So I'll I'll
1:00:31 just put this over here to the side and
1:00:34 when it processes I'll show you. All
1:00:36 right. So while we're waiting, let's
1:00:38 watch what Danielle did.
1:00:41 So, so let me let's go watch what
1:00:43 Danielle did and then I want to talk
1:00:44 about something. Okay. So, where we're
1:00:46 going
1:00:48 is to the AI salon. So, if you haven't
1:00:50 been to the AI salon,
1:00:54 jump on over to community.thesalon.ai.
1:01:00 You can go to just the salon.ai, but if
1:01:03 you if you type in
1:01:04 community.thesalon.ai,
1:01:07 AI. That'll take you over to our
1:01:09 community website. It'll land it'll land
1:01:11 you on this welcome page. You'll see our
1:01:13 cycle of AI readiness graphic there.
1:01:15 Play first, mindfully create, generously
1:01:17 lead. Um you'll see the five stages of
1:01:19 AI adoption. If you look at our values,
1:01:22 you'll learn about you know what we care
1:01:25 about and how we want people to behave
1:01:27 when they're here. And then if you
1:01:29 scroll down on this lefth hand side,
1:01:31 there's an area called clubs and hubs.
1:01:33 And a lot of these are part of the
1:01:34 mastermind uh subscription area, but
1:01:39 this channel has a channel called the
1:01:40 Irregulars, and it's free. It's open to
1:01:42 anyone. And so here is a video that um
1:01:48 that Danielle made.
1:01:51 Okay. I think it's going to pop me to a
1:01:53 new tab, but I now know how to share
1:01:55 that tab. Like that. Boom. Okay.
1:02:20 Beautiful.
1:02:28 Okay.
1:02:30 >> So, here's here's what I want to say
1:02:33 about this.
1:02:36 Um,
1:02:47 trying to think even what to say.
1:02:51 I saw some really interesting Tik Tok
1:02:53 videos that were talking about people
1:02:56 being upset that 4.0 was removed and
1:02:59 they demanded that they bring it back.
1:03:02 If you pay attention to this space, you
1:03:05 would think that the 700 million users
1:03:08 of Chat GPT are all up in arms about
1:03:12 this model being taken away from them.
1:03:14 Most of those 700 million people
1:03:18 probably were not even aware that it
1:03:20 switched from 40 to five,
1:03:23 right? They probably just use it. They
1:03:25 don't think about it. They're not paying
1:03:26 attention to personalities or to
1:03:29 features or to hallucinations or or
1:03:31 whatever. They're just using it. They're
1:03:33 probably not using it a lot. They're not
1:03:34 thinking about it a lot.
1:03:36 And then there's a whole
1:03:39 massive amount of people beyond that
1:03:42 that don't use AI at all or may they've
1:03:44 maybe used it once.
1:03:50 Those of us that are paying attention to
1:03:52 this world and paying attention to the
1:03:54 capabilities of these things and
1:03:56 thinking about the future, we're we're
1:03:58 we're in a very very small percentage of
1:04:01 human beings that are doing this right
1:04:03 now.
1:04:05 our experience with AI and how we got
1:04:08 into it. Like we can a lot of us can
1:04:11 trace it back to I remember the day chat
1:04:14 GPT came out or I remember the first day
1:04:16 I used Chat GPT and you know I had my
1:04:18 first Kevin Mallister moment right this
1:04:21 moment.
1:04:22 I had that first epiphany that oh my god
1:04:25 this AI stuff's different and I and I
1:04:27 got on the train. There's very very few
1:04:30 people that are actually on the train in
1:04:33 that spirit.
1:04:35 Um, and so a lot of us are thinking
1:04:39 about the on-ramps for AI being kind of
1:04:43 a similar version to what we did.
1:04:47 That's not how AI is going to be adopted
1:04:49 by a lot of people. There's going to be
1:04:51 a lot of people, whether it's this Pika
1:04:53 app or some other app. There's going to
1:04:56 be a lot of people that the first
1:04:58 interaction with AI is just something
1:05:00 like this where Danielle made a version
1:05:03 of herself like hanging out in the disco
1:05:05 singing this Cindy Looper song, right?
1:05:07 Probably just at the push of a button.
1:05:09 She pushed a button that said, you know,
1:05:10 let me make a a vid video of myself
1:05:13 singing that song and bang, there it is.
1:05:15 She didn't have to prompt it. She didn't
1:05:17 have to do anything other than take a
1:05:18 selfie. You take a selfie. You kind of
1:05:20 surf through the thing. You say, "Make a
1:05:22 video with me in it." and Pika does its
1:05:26 thing.
1:05:28 Right? If you look at how Pika is
1:05:31 actually doing what it's doing, it's
1:05:33 actually prompting. When you click on
1:05:36 those little pictures to like give me
1:05:37 the short, you know, spiky hair, it
1:05:41 actually wrote a prompt that said that
1:05:44 that described the hair.
1:05:48 But a lot of people might not ever know
1:05:50 that prompting is a thing.
1:05:54 Their first experience with AI might be
1:05:56 talking to something that kind of sounds
1:05:59 like Siri or Alexa,
1:06:02 but it's just a little bit smarter and
1:06:03 they don't even know that they're
1:06:04 talking to a large language model. They
1:06:06 might use a tool like this where they
1:06:08 get on. So,
1:06:11 I don't know if I've got a a point to
1:06:13 this, but
1:06:15 if you're paying attention to this
1:06:17 stuff, start thinking about the
1:06:19 different kinds of apps that you see and
1:06:22 the different kind of features. There
1:06:24 are some apps like Peak is one in
1:06:26 particular where they do a lot of
1:06:28 pre-prompting for you, right? They've
1:06:30 got the this thing called peak effects
1:06:32 where they like you upload a picture of
1:06:34 an object and it will explode it or you
1:06:36 can cut it like a piece of cake or you
1:06:38 can peel it like a decal.
1:06:40 Those are all just prompts into an image
1:06:43 model.
1:06:46 Like you could create pika yourself,
1:06:49 right? You find an image generation, you
1:06:52 know, open- source model. You you figure
1:06:54 out the API calls to it. You pre-write a
1:06:57 bunch of prompts and stick pictures in
1:06:59 front of them and when someone clicks a
1:07:01 picture, you send that prompt to the to
1:07:04 the model and you return the picture.
1:07:06 It's a relatively straightforward thing.
1:07:08 If you want to do your own little mini
1:07:10 version of Pika, you could do that.
1:07:15 A lot of developers will make fun of
1:07:18 those kind of applications, but I would
1:07:20 argue that those kind of applications
1:07:22 are going to be a lot of people's first
1:07:24 entry into AI. So, just keep paying
1:07:26 attention to that stuff. All right.
1:07:39 What's that from? Is that from that
1:07:40 interview, Brandon? That quote?
1:07:44 >> Uh, no. This is I just sent this to you
1:07:46 in a DM on Twitter. So Sam, just before
1:07:49 we went live, Sam made a lengthy post on
1:07:52 X uh talking about the 405 controversy,
1:07:56 and this was an excerpt from that that
1:07:59 uh seemed relevant to what you were
1:08:01 talking about.
1:08:02 >> Oh, cool.
1:08:10 >> Oh, great. Yeah, let's let's read this
1:08:12 one. Share screen. Yes.
1:08:16 Yes, I will.
1:08:17 I will share my screen. I will share it
1:08:20 now.
1:08:22 All right, good peoples. All right, good
1:08:24 peoples of the internet. Do we have
1:08:28 Yes, we wait. We have gas station guy.
1:08:30 Before we read Sam, before we read Sam,
1:08:34 quick question on Tik Tok. Okay, new guy
1:08:38 question. Allan,
1:08:40 love new guy questions. Is Pika the
1:08:42 model behind Chat GPT? No. Pika Pika
1:08:46 Labs. I think Pa Labs has their own
1:08:48 video model, an animation model. I don't
1:08:52 know if their image generation model is
1:08:54 their own or if they're using someone
1:08:56 else's, but Pika's got an image to video
1:09:01 model. It's it's pika. It's pika.art
1:09:06 is the website if you want to play with
1:09:07 their video generation model. And it's
1:09:10 got all sorts of sort of pre-done
1:09:12 things. And then they have two different
1:09:14 apps. They have the Pika app and then
1:09:16 they have the Pika Social app. That's
1:09:19 the one that I'm playing with right now.
1:09:20 Um, that one's the Pika Social one's not
1:09:23 widely available yet. Um, but no, the
1:09:27 the video model sitting underneath Chat
1:09:30 GBT is called Sora. S O R A. Um, it's
1:09:34 okay, not great. Um, the video model
1:09:37 that I'm enamored with right now is the
1:09:40 one in MidJourney. So, MidJourney is an
1:09:42 image generation tool. So, if you go to
1:09:44 midjourney.com,
1:09:46 um you can then turn any image that you
1:09:48 make there into a video. Problem with
1:09:50 Midjourney is it's not free and all of
1:09:52 the video models are quite expensive.
1:09:54 So, uh if you want to experiment with
1:09:56 them, it's kind of pricey, which is a
1:09:58 drag, but all right, let's go look at
1:10:05 ters. Ters is tasty.
1:10:08 You want me to clean your windows? I can
1:10:11 clean your window.
1:10:15 Wait, let me let me post it though.
1:10:17 Wait, save draft.
1:10:20 Caption.
1:10:22 Caption is going to be Ters is tasty.
1:10:24 Ters
1:10:26 is tasty.
1:10:32 All right. So, we're going to post this.
1:10:34 All right. So, now we're posted.
1:10:35 >> Okay.
1:10:36 >> Ters. Ters is tasty.
1:10:40 You want me to clean your windows? Can I
1:10:42 go back?
1:10:43 >> Clean your windows. I clean windows
1:10:45 pretty good.
1:10:48 >> I like ters.
1:10:50 Ters is tasty.
1:10:51 >> All right, here we go.
1:10:53 >> Mhm. I like ters. Ters is tasty.
1:10:58 You want me to clean your windows? I can
1:11:01 clean your windows. I clean windows
1:11:04 pretty good.
1:11:06 I like ters.
1:11:07 >> Okay.
1:11:08 >> Ters is tasty.
1:11:09 >> So that's that. Now that's pretty good.
1:11:11 That's pretty flipping good. Now, notice
1:11:14 at the bottom it says make a video like
1:11:17 this. So, if you have access to this
1:11:20 Pika app,
1:11:23 you can go to this and say turn yourself
1:11:27 into that character. You just push that
1:11:28 button and it'll make a video like that.
1:11:31 And I guess according to Danielle, it'll
1:11:33 even use that audio. Why wouldn't you
1:11:35 want to use that Ters is Tasty audio?
1:11:39 But look, it's kind of remarkable how
1:11:40 close it looks like me with the, you
1:11:42 know, I kind of look like Robert Scoble
1:11:44 there.
1:11:48 And then you can just
1:11:51 >> in my younger more [ __ ]
1:12:07 Let's see.
1:12:10 Where's the one I just took?
1:12:14 Oh, here's one.
1:12:16 All right.
1:12:18 So, now it's taking that picture of me
1:12:21 and it's applying that other dude's
1:12:22 look. Glen Beck. Yeah, I look a little
1:12:24 bit like Glen Beck.
1:12:31 Oh, it it did something different.
1:12:33 Discard.
1:12:35 [Music]
1:12:39 Make a video like this.
1:12:42 [Music]
1:12:46 Oh, I see. Hang on. Let me go back.
1:12:50 Use all elements.
1:12:54 All right. I guess I'll take a picture
1:12:55 of me again.
1:12:59 Remix error. The network connection was
1:13:01 lost. Animate.
1:13:03 Come on.
1:13:22 All right. My app broke.
1:13:25 Whatever. It's not working. All right.
1:13:27 Tik Tok pin.
1:13:30 I see no pin. Oh, Joker. I love GPT5.
1:13:33 The coding is outstanding. I have an
1:13:35 iPhone asteroids clone Kyle would like.
1:13:38 Nice
1:13:40 in HTML. Yeah, that's one of the other
1:13:42 things I was going to do is just go go
1:13:44 do a little coding and just see what it
1:13:45 does. Back to Sam. Okay,
1:13:49 so this was when did Sam write this? We
1:13:52 don't know. Um,
1:13:55 recently. Oh, today. Just before we went
1:13:57 live. Okay. If you've been following the
1:14:00 GPT5 rollout, one thing that you might
1:14:02 be noticing is how much of an attachment
1:14:04 some people have to specific AI models.
1:14:07 It feels different and strong stronger
1:14:10 than the kinds of attachment people have
1:14:11 to previous kinds of technology. And so
1:14:14 suddenly deprecating models, this is
1:14:16 what we were talking about earlier um
1:14:19 that users depend on in their workflows
1:14:21 was a mistake, right?
1:14:25 When you lose a feature in
1:14:28 Gmail, you're like, "Ah, that sucks. I
1:14:30 like that feature." But when you lose an
1:14:34 AI model that you've developed some sort
1:14:36 of affinity and relationship with, it's
1:14:39 more like the loss of a loved one.
1:14:41 Right? This is a very different
1:14:44 I mean, there there have been examples
1:14:47 of this, right? There are softwares,
1:14:49 Avatar software and things like that
1:14:51 that when they shut down the service,
1:14:53 people got really upset.
1:14:55 So, we've, you know, we've seen we've
1:14:57 seen this before, but this is going to
1:14:59 be an ongoing,
1:15:02 increasingly complicated thing. Okay,
1:15:07 this is something we've been closely
1:15:08 tracking for the past year or so, but
1:15:10 still hasn't gotten much mainstream
1:15:12 attention other than when we released an
1:15:14 update to GPT40 that was too sick of.
1:15:17 Yeah, remember that? Remember when they
1:15:19 did that? They they did, "Oh, you're
1:15:21 gonna really like the new GVT40 and and
1:15:23 the press was all over it." This is just
1:15:26 my current thinking and not yet an
1:15:28 official open AI position.
1:15:30 People have used technology, including
1:15:32 AI, in self-destructive ways. If a you
1:15:35 user in a mentally fragile state and
1:15:38 prone to delusion, we do not want the AI
1:15:40 to reinforce that. Most users can keep
1:15:43 clear a clear line between reality and
1:15:45 fiction or roleplay, but a small
1:15:46 percentage cannot. We value user freedom
1:15:49 as a core principle, but we also feel
1:15:51 responsible in how we introduce new
1:15:53 technology with new risks.
1:15:56 I
1:15:58 there are a bunch of things I don't like
1:16:00 about AI and I don't like about how
1:16:02 things were trained. Um, I gotta say
1:16:06 like I feel like
1:16:09 Altman at a minimum understands what the
1:16:12 issues are. You may or may not agree
1:16:14 with how he deals with them. I feel like
1:16:16 he at least understands the issues. Um,
1:16:18 I don't hear that from a lot of If you
1:16:20 listen to [ __ ] Zuckerberg, that
1:16:22 dude's like the the technology will save
1:16:25 you. Everyone will have their own super
1:16:27 intelligent person and we're gonna
1:16:29 deliver it.
1:16:31 He's just completely cut off from these
1:16:34 kind of issues. So at least Altman is
1:16:36 speaking it. Encouraging delusion in a
1:16:39 user that is having trouble telling the
1:16:41 difference between reality and fiction
1:16:42 is an extreme case and is pretty clear
1:16:44 what to do. But the concerns that we
1:16:46 worry mo most about are more subtle.
1:16:49 There are going to be a lot of edge
1:16:50 cases and generally we plan to follow
1:16:52 the principle of treat adult users like
1:16:55 adults which in some cases will include
1:16:57 pushing back on users to ensure they are
1:17:00 getting what they really want. A lot of
1:17:02 people effectively use chat GPT as a
1:17:04 sort of therapist or life coach even if
1:17:07 it wouldn't if they wouldn't describe it
1:17:09 that way. This can be really good. A lot
1:17:12 of people are getting a lot of value
1:17:13 from it today already. I hear that all
1:17:15 the time.
1:17:17 If people are getting good advice,
1:17:19 leveling up toward their own goals, and
1:17:21 their life satisfaction is increasing
1:17:23 over years, we will be proud of making
1:17:25 something genuinely helpful, even if
1:17:27 they use and rely on chat GPT a lot. If,
1:17:30 on the other hand, users have a
1:17:32 relationship with Chat GPT where they
1:17:34 think they feel better after talking,
1:17:36 but they're unknowingly nudged away from
1:17:39 their longer term well-being,
1:17:42 then that's bad, right? So,
1:17:47 you know, like in in that recent
1:17:49 interview on on that podcast, I just I
1:17:51 just watched it tonight. He was talking
1:17:53 about, you know, there are lots of ways
1:17:55 that they could be manipulating users to
1:17:58 go deeper with these relationships that
1:18:00 they're choosing not to, like they
1:18:02 haven't made a sex bot, for example, is
1:18:04 one of the things he said. It's also
1:18:06 bad, for example, if a user wants to use
1:18:08 chat GPT less and feels like they
1:18:11 cannot, right? if they get addicted to
1:18:12 it. I can imagine a future where a lot
1:18:15 of people really trust Chat GPT's advice
1:18:17 for their most important decisions.
1:18:19 Although that could be great. Wait,
1:18:21 although that could be great. It makes
1:18:23 me Oops.
1:18:29 That thing finished.
1:18:39 Um,
1:18:50 [Music]
1:18:53 all right. There we go. Okay.
1:18:57 I can imagine a future where a lot of
1:18:59 people really trust Chat GBT's advice
1:19:01 for their most important decisions.
1:19:02 Although,
1:19:04 okay. Yeah, we talked about all this.
1:19:05 Okay. So we as in society but also as we
1:19:09 in at open AI have to figure out how to
1:19:12 make it a big net positive. There are
1:19:15 several reasons I think we have a good
1:19:16 shot at getting this right. We have much
1:19:19 better tech to help us measure how we
1:19:22 are doing than previous generations for
1:19:24 techn that technology had. For example,
1:19:27 our product can talk to users to get a
1:19:29 sense of how they are doing with their
1:19:31 short and long-term goals. We can
1:19:32 explain sophisticated and nuanced issues
1:19:35 to our model and much more. Um,
1:19:38 I'm gonna I'm gonna go out on a limb
1:19:41 here. Here's where I think Open AI is
1:19:44 [ __ ] up. Um,
1:19:49 they're only hiring engineers.
1:19:52 They should be hiring anthropologists
1:19:55 and liberal arts majors and storytellers
1:20:00 and filmmakers.
1:20:02 And they should balance
1:20:05 the logical thinking that is generating
1:20:08 these remarkable tools
1:20:11 with
1:20:12 some some rightrained people with some
1:20:15 creative people with some artistic
1:20:17 people some people in the humanities
1:20:19 some people with phil philosophy degrees
1:20:23 because even though this is high math
1:20:26 that is creating these things they are
1:20:28 large language models. They are large
1:20:30 language models.
1:20:33 How they manipulate us as humans is a
1:20:36 core part of what they do, right?
1:20:40 And if all of the frontier technology
1:20:43 companies are just hiring, you know, GPU
1:20:47 engineers and coding engineers and
1:20:49 mathematicians and researchers and
1:20:51 they're not balancing it with that other
1:20:54 side, I think they're going to they've
1:20:55 got a real chance of [ __ ] this up and
1:20:58 missing it. Now, the good news is
1:21:00 there's a bunch of people out here who
1:21:02 are going to use this technology and
1:21:04 sort of bend it in a good direction, but
1:21:07 if they [ __ ] it up at the core model
1:21:08 level, then there's not a lot we can do
1:21:12 about it. So, anyway,
1:21:15 wait, go back one paragraph. You missed
1:21:17 a critical middle sentence. I can
1:21:19 imagine a future where a lot of people
1:21:20 really trust Chat GPT's advice for their
1:21:23 most important decisions.
1:21:25 Although that could be great, it makes
1:21:27 me uneasy. But I expect that it's coming
1:21:30 to some degree. Oh yeah, I did. This is
1:21:32 the important sentence that I missed.
1:21:34 But I expect that it's coming to some
1:21:36 degree and soon billions of people will
1:21:38 be talking to AI in this way. Yeah. So
1:21:40 we as society and we as open AI need to
1:21:43 figure out how to make it a big net
1:21:44 positive. Um,
1:21:48 the whole the the musical that I've
1:21:50 spent the last year and a half working
1:21:52 on is about our relationship
1:21:56 with AI
1:21:58 and what is it? What can it be? What
1:22:01 should it be? Is it possible to have
1:22:04 feelings for an AI? Well, I think we
1:22:06 just found that out with with GPT40
1:22:09 Side Hustle Mimi. I think we need to
1:22:11 tell them the same thing that we told
1:22:14 the Uno card creators. Thank you.
1:22:18 It's ours now. Yeah. I I I mean to a
1:22:21 great degree that the the AI salon, the
1:22:24 AI learning lab,
1:22:26 a lot of times people that come into
1:22:28 this channel, they ask me this question.
1:22:30 They they say, "Sir, what are your
1:22:32 qualifications?" And what the
1:22:34 implication there is is, "Are you a
1:22:38 mathematician? Are you a scientist? Are
1:22:40 you qualified to talk about AI?
1:22:45 Um,
1:22:47 yes. Because I'm not talking about AI.
1:22:50 I'm not talking about model weights and
1:22:52 algorithms in the latent space. What I'm
1:22:55 talking about is now that generative AI
1:22:58 exists,
1:23:00 now that the English language is our
1:23:03 primary programming language to these
1:23:06 incredibly powerful knowledge engines.
1:23:11 It's in the hands of the rest of us now
1:23:14 to figure out how to actually [ __ ]
1:23:16 use this technology.
1:23:18 That's what my focus is.
1:23:22 I'll let the scientists go build it.
1:23:24 I'll let them figure out all these
1:23:25 issues of where are the safety
1:23:27 boundaries and what are they allowed to
1:23:29 do, what are they not allowed to do.
1:23:30 Let's assume that they're going to
1:23:32 they're going to figure that out. I
1:23:33 personally think they need to hire some
1:23:35 liberal arts
1:23:37 majors soon or they're going to [ __ ]
1:23:40 that up. But once the models come to us,
1:23:43 then it's up to us, the 98.5%
1:23:47 of people that are not scientists and
1:23:49 engineers, it's up to us to figure out
1:23:52 how do we use these things? How do we
1:23:54 use them for good? How do we use them to
1:23:56 improve ourselves? How do we use them to
1:23:57 amplify our ideas? How do we take a
1:24:01 virtual 100 employee
1:24:05 company
1:24:06 where we've got a hundred employees that
1:24:09 are all PhD level employees and use that
1:24:13 for ourselves for our own ideas and use
1:24:15 that for good in the world? I don't know
1:24:18 yet. None of us do. But that's my
1:24:22 interest is what are we going to do now
1:24:24 that we have it in our grubby little
1:24:25 hands?
1:24:29 What do we do? How do we do it? How do
1:24:32 we make a difference?
1:24:34 Just keep talking about it is the the
1:24:36 only way I know right now. All right,
1:24:41 that's that. That's fascinating.
1:24:43 Fascinating. It's fascinating, people.
1:24:45 I'll tell you what, it's really
1:24:46 fascinating. Fascinating. All right,
1:24:49 let's go look at my video.
1:24:57 [Music]
1:25:05 Uhoh.
1:25:14 I run like JD Vance.
1:25:22 Uh, yeah. Charact Korea character
1:25:25 reference. I haven't played with it
1:25:26 enough to know I'll just get go in there
1:25:28 and get lost. A lot of these character
1:25:30 reference things are very they're very
1:25:33 UI specific. So, um,
1:25:40 okay. Side hustle Mimi, I will check my
1:25:42 DMs. Yeah. So, by the way, if you are on
1:25:45 the AI salon, if you're not on the AI
1:25:47 salon, go join the AI salon. Um, I've
1:25:50 got three invites, I think, for the Pika
1:25:52 thing. So, if you DM me, um, whoever
1:25:55 whoever gets there first, I'll get you a
1:25:57 code. All right, that's another reason
1:26:00 to come here, people.
1:26:05 [Music]
1:26:07 All right.
1:26:17 It's a quick demo. Where do I go for it?
1:26:21 Is it basically just like what chat GPT
1:26:23 did?
1:26:31 All right. So,
1:26:38 so I'm at
1:26:41 gemini.google.com.
1:26:44 Click the three dots. Guided learning.
1:26:47 Okay. So, yeah. So, so both um both
1:26:50 OpenAI and Gemini have just added these
1:26:53 these new things
1:26:56 um guided learning. So, learn and
1:26:57 understand topics. Teach me
1:27:02 teach. Why can I not type?
1:27:09 Oh, wait. Do I have to Oh, let me choose
1:27:12 guided learning.
1:27:14 Oh, I can't do it with that account.
1:27:17 Sorry. As a paid member of Google
1:27:20 Workspace, you can't use any of the new
1:27:22 features. You're welcome.
1:27:27 No thanks.
1:27:31 Guided learning. Guided learning. Okay,
1:27:33 I've got guided learning on. So, I'm
1:27:35 going to say teach me about
1:27:40 No, I'm not going to say teach me. I'm
1:27:42 going to say um
1:27:47 how
1:27:49 do
1:27:49 >> how to change your tabs in Google
1:27:52 Chrome.
1:27:54 >> Wait, but I didn't change them. I just
1:27:56 typed Oh, cuz when I changed the counts,
1:27:59 it moved me over there. Unbelievable.
1:28:01 All right. How do I find the average?
1:28:07 Was it the average speed
1:28:10 of a something swallow
1:28:14 of a
1:28:16 who knows the Montipython
1:28:18 of a swallow?
1:28:20 I forget the Let's see if it gets the
1:28:22 reference.
1:28:27 Slow thinking. It's doing slow thinking.
1:28:31 Gemini is a beast. Unladen. That's
1:28:33 right. Oh, that's a classic question. To
1:28:36 answer it, we first have to know are you
1:28:38 are you asking about an African African
1:28:40 or European swallow? We'll go European
1:28:44 and wait eur damn it European
1:28:50 and unladen.
1:28:58 Got to DM me.
1:29:03 Ah, a connoisseur of classic film in
1:29:06 Greece.
1:29:11 Wait, based on Oh, did I not have that
1:29:14 turned on? No, I do have it turned on.
1:29:18 Um,
1:29:19 teach me how to do it.
1:29:22 Let's Let's actually go go compare this
1:29:25 to
1:29:27 OpenAI because this this Gemini thing
1:29:30 just it just gave me the answer. It
1:29:33 didn't teach me.
1:29:36 No, they got to DM me on the aselone.
1:29:38 Sorry.
1:29:45 Um,
1:29:48 direct officer. Wait,
1:29:55 this is not very good teaching. Let's go
1:29:57 uh let's go back to chatbt for a second.
1:30:00 Well, it what we're seeing with the
1:30:03 Gemini model, Kyle, is more of a a
1:30:07 professor at trying to tutor a student.
1:30:11 So, it's taking a concept that you have
1:30:12 some familiarity with and trying to
1:30:15 coach the answer out of you. So, do you
1:30:17 think it's A or would you think it's B?
1:30:19 And then if you choose wrong, it would
1:30:22 course correct.
1:30:24 >> I get that, but it didn't do that here.
1:30:32 It just gave me the answer.
1:30:43 Let me let me try I want to try the same
1:30:45 question actually. Let me go get the
1:30:46 original question.
1:30:50 Um
1:30:52 Oh, I don't have it. I lost it. Did I?
1:30:56 Yeah. Here we go. All right. Let me go
1:30:59 to chat GPT and turn on its study mo
1:31:01 mode and see if it's different.
1:31:09 Study and learn
1:31:13 of an unladen
1:31:17 swallow.
1:31:24 [Music]
1:31:26 It's also amazing to me like how like
1:31:30 feature for feature all these frontier
1:31:32 model companies are copying one another
1:31:35 because this study mode came out from
1:31:36 chat GPT before we dive in. Quick note I
1:31:41 can't show my private SCA scratchwork
1:31:43 but I'll walk you through the steps and
1:31:45 thinking see this is this is what I was
1:31:47 expecting it to do and this is what what
1:31:49 Gemini didn't do. Here's the idea we
1:31:52 use. Average speed equals total total
1:31:54 distance divided by time. Straight up
1:31:55 physics. You turn one bite-sized step.
1:31:59 If a swallow covers, 1200 meters in 60
1:32:02 seconds, what's its average speed? Just
1:32:04 divide and tell me the number in units.
1:32:07 Do the math
1:32:10 for me.
1:32:13 All right. Well, crunch the numbers
1:32:14 together.
1:32:16 Yeah. And it's giving you math things.
1:32:18 So, I I'm sure the Google one's fine. It
1:32:21 just the the way I ph phrased that
1:32:23 question, it just answered it for me. So
1:32:26 something was wonky there. But anyway,
1:32:29 both of these now, Gemini and Chat GBT
1:32:32 have this study mode. So if you've got
1:32:35 kids or grandkids that you want learning
1:32:38 stuff, there you go. All right. All
1:32:41 right. All right. All right. Fantastic.
1:32:48 Be right back. Bedtime for the wild
1:32:50 things. Oh man, are we DMing through
1:32:54 Mighty Networks? Yes, DM me in Matetwax.
1:32:58 Mata Networks.
1:33:05 [Music]
1:33:12 All right, everyone.
1:33:16 G.
1:33:17 Redditors are not happy. Yeah, Redditors
1:33:20 are never happy.
1:33:24 They're just over there bitching. Again,
1:33:26 don't forget,
1:33:29 Redditors that are paying attention to
1:33:30 every permutation of what's coming out
1:33:33 of the Frontier model companies are not
1:33:36 most of who's going to use Chat GPT.
1:33:38 They're I I would argue they're not most
1:33:40 of who's using Chat GBT. You got 700
1:33:43 million weekly users using chat GPT.
1:33:46 If you think
1:33:48 anywhere north of 5% of those 700
1:33:51 million users are using Chat GPT for
1:33:55 anything other than glorified Google
1:33:57 search, you're high on crack. There's no
1:34:00 way. So, even though Reddit's noisy and
1:34:04 X is noisy and people are bitching and
1:34:06 Open AAI is responding and restoring the
1:34:08 model,
1:34:10 um,
1:34:12 just give it some time. I the the the
1:34:15 everyone jumping to conclusions. I swear
1:34:17 to God, Gary Marcus, the minute Chat
1:34:20 GPT5 came out, he jumped on it like like
1:34:23 he was literally doing victory laps.
1:34:24 See, I told you everything sucks. It's
1:34:27 all [ __ ] awful. You're all [ __ ]
1:34:28 losers. You promised it would do stuff.
1:34:31 it doesn't do it. It's like this the
1:34:34 stuff is emerging and evolving, you
1:34:37 know, in our presence and it's doing
1:34:39 remarkable [ __ ]
1:34:42 and every time a new model comes out,
1:34:44 it's going to take us a while to
1:34:46 understand what it actually is. Um, so
1:34:50 all of the drama right now, just ignore
1:34:52 it. It's all just [ __ ] Um, what's
1:34:55 the airspeed velocity of an unladen
1:34:57 swallow? That's that's what I asked. I
1:35:00 didn't ask it right because I I'm an old
1:35:02 man and I couldn't remember the actual
1:35:03 line for Monty Python.
1:35:08 What is the airspeed velocity of an
1:35:10 unladen swallow?
1:35:14 We are the knights who save me.
1:35:20 It was horrible. Came out of the cave
1:35:22 and big gnarly teeth.
1:35:26 It's just a scratch.
1:35:31 Oh my god.
1:35:34 All right, everyone. Um, tomorrow's just
1:35:37 a regular Tuesday. I don't have a salon
1:35:39 or anything. So, um, I'll be here at 8
1:35:41 o'clock tomorrow night.
1:35:44 Um, I will probably pay a bit more
1:35:46 attention to things going on. It seems
1:35:48 like some things are perking. So, I may
1:35:51 I may have more of a an idea of what to
1:35:53 talk about tomorrow night, but I don't
1:35:54 know. Maybe not. We'll see. All right.
1:35:58 Fantastic everybody. You call that a
1:36:01 flesh wound?
1:36:03 Just a flesh wound. I'll bite your
1:36:05 kneecap.
1:36:08 Such a good movie. It's such a good
1:36:10 movie. Oh my god, it's a good movie. I
1:36:12 haven't seen one of those in a long
1:36:13 time. I should probably go watch that.
1:36:16 Um,
1:36:20 when is the AI jam session
1:36:24 on your links? Oh, the AI jam sessions.
1:36:26 Those are private one-on-one meetings
1:36:29 with me. Those are um I've got four
1:36:32 hours reserved on Saturdays. Um 10 11
1:36:37 and then one and two mountain time. Um
1:36:41 so if you want one-on-one time, if you
1:36:43 go to my Tik Tok channel and click on my
1:36:46 profile, I've got a hyperlink there that
1:36:49 will take you into the jam session and
1:36:52 you can subscribe. You can you can book
1:36:53 an hour or you can book 10 hours. Um but
1:36:57 it's it's individual hourlong sessions.
1:37:00 So if you want to do that would love
1:37:02 that.
1:37:04 Um
1:37:09 fairy tale revolution. I learned it from
1:37:10 you Kyle. I don't know what you learned
1:37:12 from me but I hope it was a good thing.
1:37:17 Oh man. Oh that's Claire. Oh cool. Dr.
1:37:20 J. Nice. Beautiful. Beautiful.
1:37:24 Beautiful. All right. Always looking.
1:37:26 Always look on the bright side of life.
1:37:31 [Music]
1:37:32 [Laughter]
1:37:34 All right, everyone. I will see you
1:37:37 tomorrow. Hope you have a good night.
1:37:38 Uh, it's good to be back. It's weird to
1:37:42 be back because I took five days off and
1:37:43 I haven't done that in forever, if ever.
1:37:46 All right. Peace. See you later.