5/9/2025 - Exploring GenSpark AI Sheets and GPT-4

Description

Chapters

Transcript

0:01 You ready,
0:02 Champ? Champy, are you
0:05 ready? Are you ready?
0:09 [Music]
0:22 [Music]
0:29 You and I get all
0:32 alone. Sunday morning here at
0:37 home. Sky's blue and the coffee is
0:40 strong and it's true.
0:43 [Music]
0:45 Then I open my eyes to a dream realized
0:50 in front of
0:52 me and I haven't got a clue what in the
0:56 world is happening to
1:00 me. Think I think I'm
1:03 happy like first day summer vacation.
1:07 Happy got to get a little rest and
1:09 relaxation.
1:12 Happy like fire on Sunday morning
1:14 singing true.
1:19 [Music]
1:44 [Music]
1:57 Oh, is this a place I can rest my
2:04 forehead, gather my thoughts in Swiss
2:07 silence.
2:12 And there's this place where the
2:14 feelings
2:19 on all culture through violence.
2:24 Friday night date night. It just seems
2:26 like last Friday we had one of these. Is
2:29 Is it like every Friday? What are we
2:31 doing? Every Friday. Friday night date
2:34 night. All right.
2:37 [Music]
2:42 How do I get rid of that
2:45 sticker?
2:49 Uh, there we go. Peel it
2:53 off. It's happening. Danielle Silver Fox
2:58 shaking. Silverf Fox shared the live
3:02 because Silver Fox a mench is a mench.
3:05 She's awesome. She's a live
3:08 sharer. When was the last time you
3:10 shared a live? Are you just like a
3:12 lurker? You just come in here to lurk,
3:15 judge me
3:17 silently or out loud, but not in the
3:22 comments. That's what I would
3:29 do. YouTube comment. I see no YouTube
3:32 comments.
3:37 Ah, there I do. Champion is a vibe
3:39 vocalist for sure.
3:42 [Music]
4:00 [Music]
4:02 Oh yeah. Champy. Champy.
4:05 Champy. What are we going to talk about
4:08 [Music]
4:16 tonight? Walt Grace desperately hating
4:19 his old place. Dreamed to discover a new
4:22 space. buried himself
4:25 alive inside his basement. Turn on the
4:29 side of his facement. He's working away
4:32 on
4:33 displacement. What it would take to
4:37 survive. Cuz when you're done with this
4:43 world, you know, the next is up to
4:47 you.
4:49 [Music]
4:51 Um, you got to spank your guitar. You
4:55 got to spank it. You got to spank
5:07 it. Wrong channel, Kyle. Ah, damn it.
5:10 That was the twerking channel. Sometimes
5:12 I forget. Am I on the twerking channel
5:14 or am I on the the the brainy one? the
5:18 where we're talking about the
5:21 intelligence, artificial intelligence,
5:24 not the real intelligence.
5:26 Clearly, that's not happening here.
5:31 [Music]
6:13 [Applause]
6:14 [Music]
6:17 How about new Google
6:19 AI? Which which one, Silver Fox? The new
6:23 image one or the new 2.5 something?
6:27 What's the They got They got the one the
6:30 one that's supposed to be really good at
6:32 coding. Let's see. Gemini
6:36 2.5. I think it's the the 56 preview.
6:41 I don't know. We'll check them out.
6:42 We'll look at that
6:45 [Music]
6:47 stuff. More importantly, are we sure
6:49 we're in the right one? Oh, you mean for
6:52 Tik Tok for the for this channel? Um, I
6:56 I was thinking about having this channel
6:58 just be a crossover. I just merge the
7:00 two and I just show up here one night in
7:02 a manini and start doing some twerking
7:05 and just see how it goes, right? I don't
7:08 think they'll ask me about my
7:09 qualifications
7:11 [Music]
7:16 anymore. Source camp. Nice guitar
7:19 playing. It's lulling me to sleep.
7:24 [Music]
7:50 Thanks for letting me in today. Well, of
7:51 course that's what office hours is all
7:54 about.
7:56 [Applause]
7:57 [Music]
8:19 Hoo! Doo doo doo doo.
8:24 [Music]
8:49 Hey, good.
8:51 [Music]
8:54 [Applause]
8:55 [Music]
9:04 Did I get accepted into the new notebook
9:07 LM beta app? No, largely because I
9:11 didn't fill out the
9:15 form. Good reminder, Danielle. That's
9:17 solid. Solid of you. Let me see if I can
9:20 go find it. Here it is.
9:24 We're going to go fill that out right
9:25 now. No need to share this. I'm just
9:27 filling out a
9:29 form. Listen, I'll be with you people
9:32 when I get with you people. Do you
9:33 understand what I'm saying? Do you think
9:35 this is easy? You think it's easy being
9:39 [Music]
9:44 me? All
9:46 right, here we go. Let's see. I can
9:49 actually see now. That's nice. Let me do
9:51 my black bar for the for the good Tik
9:54 Tok people. Sorry. Which AGI do you
9:57 think is best? Well, many would argue
10:00 we're not at AGI yet. Um, but hang on.
10:02 Let me fill out this form. Kyle, please
10:05 provide the email address you'd like us
10:08 and yeah, I already did that. Kyle.
10:10 Yeah. Yeah, there we go. What country
10:13 are you in? United States. USA.
10:19 Um, I'll be testing iOS. I agree. What
10:23 will be
10:24 next? There we
10:26 go. All right. I just did that.
10:31 [Music]
10:52 Producer Brandon. Producer Brandon, give
10:54 me a little shade in the background
10:56 here. I like it. I like it. You gotta
10:58 You gotta keep You got to keep the hosts
11:00 on their toes because they'll just go
11:02 rogue. They'll just go out and do [ __ ]
11:06 Um, the question about which AGI do I
11:10 like best? Which which AI do I like
11:13 best? Yeah. I
11:15 mean, you could probably argue that
11:19 03, Manis, and Gen
11:22 Spark are what Sam Alman call was it Sam
11:27 Alman? Someone from open AI said spark
11:29 you know sparks of AGI or you
11:33 know hints of AGI whatever whatever it
11:36 is. You could argue that that the new
11:38 reasoning engine from OpenAI Manis and
11:42 Gen
11:43 Spark start to live in the neighborhood
11:45 of AGI.
11:50 But the the definition of AGI that that
11:54 I am going to go with
11:57 is you know a system that's smart enough
12:00 to do you know most economically viable
12:03 jobs to do them to do the jobs right not
12:08 to have a human sit there and babysit it
12:11 so that it does 80% of a job and then
12:14 kind of halfass does the other 20% and
12:16 humans like, "Well, this sucks because
12:19 we that's what we have with computers
12:21 now, the this sucks thing." So, I would
12:24 argue we're not at AGI yet, but you
12:26 know, we're we're starting to get hints
12:27 of it's going to be pretty [ __ ] cool.
12:30 But, good question.
12:34 [Music]
12:50 [Applause]
12:50 [Music]
12:54 She came on here like slow moving cold
13:02 front. Well, be was warmer than a look
13:06 in her
13:11 eyes. She sat on the stool and he said,
13:15 "What do you
13:17 [Music]
13:19 want?" She said, "Give me a love that
13:23 don't freeze up
13:26 [Music]
13:29 inside." Said, "I have melted some horse
13:33 in my time, dear.
13:36 [Music]
13:38 But to sit next to you, well, I shiver
13:42 and
13:46 shake. And if I knew love, well, I don't
13:50 think I'd be
13:55 here asking myself if I've got what it
14:03 takes. I do melt your icy blue heart.
14:09 [Music]
14:14 Should
14:16 pass turn what's been frozen
14:19 for
14:23 you into a river of
14:27 [Music]
14:32 tears. Oh, that Champy loves his
14:35 singing. when you get a song just right,
14:37 he just jumps
14:39 in.
14:41 Um, so what we're talking about on the
14:44 live today, we're talking a little bit
14:46 about
14:47 um, we were talking a little bit about
14:51 uh, Gen Spark came out with a new thing
14:53 called Sheets. Gen Spark Sheets, which
14:55 we'll go play with. If you haven't heard
14:56 of Gen Spark or played with it yet, it's
14:59 pretty remarkable. Um, and they're
15:01 they're rolling out features pretty damn
15:03 quick the past couple of weeks. Um,
15:06 they're really good. We were also
15:08 talking about I'm still semiobsessed
15:11 with this email from the CEO of Fiverr
15:14 to his staff.
15:17 um that basically talks about like what
15:19 he basically says in it, you've got a
15:21 few
15:22 months to get your
15:25 ass aggressively curious about
15:29 AI or you're just going to have your ass
15:31 handed to
15:33 you. And uh he he said it plainly and
15:36 clearly and you know people are arguing
15:39 arguing whether he did it right or not,
15:42 but like it was at least clear. It had
15:45 some urgency. like it's it's the first
15:47 it's the
15:48 first, you know, communication of
15:52 note. Um, you know, other than people
15:56 like Sam Alman who are building the
15:58 [ __ ] but he's even understated, but it
16:01 was the first thing of note that had
16:02 some urgency to it. So, I actually liked
16:04 it. Did you see the image I made for
16:06 you? It's in Look what I made. I did
16:10 not. Let's go look.
16:14 There's one from Vicki. Oh, that's
16:17 nice. Photo realistic unicorn.
16:29 Nice. Nice. Vicky launched a lovable
16:33 app, the visual prompt generator. Very
16:36 cool.
16:38 Nice. Steo, the new
16:43 pope. If anyone's offended, take it
16:54 down. Hey, it looks like Champy. They
16:57 got the eyes right.
17:02 [Music]
17:32 Put it here.
17:35 [Music]
18:01 Ah.
18:04 Um, nah, I was not I was not that blown
18:08 away with the with the There's a new
18:09 image gen model from Gemini, but I
18:12 wasn't that that impressed with it. They
18:14 said it was supposed to be better, but I
18:16 thought the the text was a little eh it
18:19 was a little janky. So, not worth
18:21 playing with because it's too hard to
18:23 find. Just the the Gemini [ __ ] the
18:25 Google stuff is just ridiculous to find.
18:28 What are we going to do tonight?
18:31 Um, oh yeah, that's a good thing. So, if
18:34 you go to
18:37 [Music]
18:43 uh if you go to
18:47 OpenAI and search for GPT enterprise
18:50 models and
18:52 limits, go to help.openai.com
18:57 openai.com and then on YouTube I'll post
19:01 in this
19:03 URL on Tik Tok. You're on your own, but
19:07 you can go I mean you can see it. You
19:09 can screenshot it and see it up
19:12 here. Um but they put out a list of when
19:16 to use each
19:18 model. Now, let me give you my two cents
19:21 on this stupid [ __ ] thing.
19:24 We were
19:25 promised by by Mr. Altman
19:29 himself that GPT5 was going to come out
19:32 in May and we wouldn't have to deal with
19:36 all these [ __ ] models. That they were
19:37 going to have some sort of magic model
19:40 picker. Well, the only thing they've
19:43 been doing is picking their nose
19:45 apparently because GPT5 ain't coming
19:48 anytime soon. Probably the summer. And
19:51 now they've they've put out a document
19:53 that says here's when to use each
19:57 model. The whole point of having magic
20:01 AI is that we don't have to use our
20:04 brains. So this is not this does not
20:06 make me happy.
20:08 But GPT40 excels at everyday tasks.
20:12 Brainstorming, summarizing emails, and
20:14 creative content. Fully multimodal.
20:16 Supports almost all capabilities.
20:20 GPTs, data analysis, search, image
20:23 generation, canvas, advanced voice, and
20:26 inputs, documents, images, CSV files,
20:30 audio, and video.
20:32 Wait, we can upload video to chat
20:36 GPT. Hang on a second. Hang on one go
20:39 dang
20:40 second. Let's go test this. Oh, look at
20:43 my look at my
20:51 superhero. I don't know why it put
20:53 [ __ ] Washington in the background,
20:55 but you know, cuz it because I had done
20:58 one of those earlier in the chat
21:00 apparently. All right, let's see the new
21:03 midjourney. Is there a new midjourney?
21:05 We can look at the new midjourney, but
21:07 wait, I got to see if this thing can
21:08 input videos. Let me start a
21:12 new a new chatty
21:16 chat. We'll
21:20 go.mpp4. And honestly, we could grab
21:41 [Music]
21:46 All right, this is
21:47 [Music]
21:52 good. Okay, so this is a little opening
21:58 montage for my Broadway musical that
22:02 opened on Broadway over the winter. It's
22:05 not true.
22:07 Uh uh uh uh uh uh uh uh. Come
22:17 [Music]
22:19 on. Oh, I wonder if I have to do the
22:22 instead of the
22:23 MP4. Ah, file
22:29 formats. It was going just
22:34 slow. All right, this one's going now.
22:37 Okay. White ring. Yeah, I know. It
22:40 looked like it hung a little
22:43 bit. Um I I don't know. Let's see. Um
22:48 what Let's see. Uh what is
22:52 this video about? So at least it let me
22:55 upload it. What is this video
23:01 about regarding wildlife documentaries?
23:03 What is this video about? Um
23:08 what where is the
23:15 location? Oh, movie pie
23:17 editor. So it's using a Python library
23:20 to analyze it. Extract a few
23:23 representative
23:25 frames. One duration. So it's basically
23:28 every second it's grabbing a different
23:30 frame. Then it's doing ffmpeg parsing.
23:34 Would you like me to extract key frames
23:36 so we can visually identify? Yes. I
23:39 thought we I thought we established that
23:41 when I said, "What is this video
23:43 [Laughter]
23:53 about? Thank you, Winston, for the gift.
23:56 It looks like people have been sending
23:57 some gifts over here on on the Tik Tok.
24:01 Ion China seems to think there's a new
24:02 midjourney. Vicki seems to think there's
24:11 not. I used to work on Fiverr. I can't
24:14 believe they're still in
24:18 business.
24:20 Uh, definitely the 20% of not AGI. All
24:25 right. What's this doing? Did we do it?
24:28 Here's the first frame. Here's the first
24:30 frame of the video. Extract it at the 1
24:33 second mark to help identify the
24:38 location. Yes. Stop asking me if you
24:44 should do your
24:51 job. Uh so it's so it's doing code
24:55 analysis.
25:08 Just tell me
25:10 what's in the
25:15 images. So this is not if you want to
25:19 see if you want to see what
25:22 multimodal video looks like, it does not
25:24 look like
25:26 this. This is [ __ ] Ah, but it did
25:29 finally get it. Okay. Frame half a
25:32 second. A wide view of Times Square, New
25:34 York
25:36 City. The camera zoomed in a bit. The AI
25:39 entity is more
25:42 prominent. The visual style emphasizes
25:45 futurism, artifice, and power, almost
25:47 like a goddess of information.
25:52 The video is a stylized promo intro set
25:55 in Time Square showcasing a vivid
25:57 abstract female AI figure broadcast
26:00 against the massive LED
26:02 screens. Want me to create a cinematic
26:05 description? I'm going to say based on
26:09 everything you know about
26:13 me, what could this be a promo for?
26:19 Mark, let's see if it gets
26:20 Sydney.
26:23 Sydney, based on everything I know about
26:25 you, your work in AI storytelling, your
26:27 musical Sydney, your obsession with
26:29 narrative tech, and your knack for
26:31 blending futurism with emotion, the
26:33 video could be a promo for a Broadway
26:36 show, Sydney and Artificial Love
26:39 Story. Oh, and then it guessed a bunch
26:42 of other [ __ ] All right, so it did it.
26:44 But here, let's go. Let's go show you
26:47 what real multimodal looks
26:49 like. So, I'm going to go to Now, here's
26:52 the problem. I don't know which of these
26:53 models is actually multimodal anymore.
26:57 Let's assume that maybe this is. Okay.
26:59 So, we'll do new chat. So, where I am
27:01 right now is I'm in Google uh
27:06 aistudio.google.com and I'm in the
27:08 Gemini 2.5 Pro. And let's go
27:12 see. Upload a file. Yeah, I could have
27:15 put in a
27:17 uh a YouTube a YouTube video there. So,
27:21 that's
27:22 promising. Is this
27:25 it?
27:32 No. MP4.
27:47 Why can't I not find it
27:51 now? Oh, that's
27:59 [Music]
28:07 why. Okay, this is it.
28:27 [Laughter]
28:28 Steo, I have another another image of
28:30 you and Champion on the opening night of
28:32 Sydney on
28:34 Broadway. All right. What's this thing
28:36 doing? So, this is uploading the file.
28:38 It looks like there's going to be a
28:39 video preview here, which is
28:46 interesting. I tested like Gemini in
28:50 Google Studio probably nine months ago,
28:53 you know, could fully understand video.
28:57 So like this is this is doing it.
29:05 So,
29:10 okay.
29:11 Extracting. So, it's taking it a
29:14 while. 14,000
29:17 tokens. Oh, type something. I'm going to
29:19 say, okay, um, give me a shot
29:25 list of the video.
29:30 Um,
29:33 what creative work is being
29:39 featured? We'll try something else here.
29:42 And the
29:43 style of music in the
29:55 background. Expanding to view model
29:57 thoughts. Oh, expand to view model
29:59 thoughts. Yeah. So, look at what this is
30:01 doing. Um, 01 black screen. 01 to 03
30:06 Sydney and artificial love story by K
30:07 Kyle Shannon and Andrew Watts. Image,
30:09 woman's profile, color hair. Um, 03 to
30:12 04 blurry title card. 04 to05 times
30:15 square like scene, red carpet steps,
30:17 crowds, billboards. This is what a
30:20 native video model looks
30:23 like. Consolidate and refine the shot
30:25 list. So, this is it thinking. So, it's
30:28 using reasoning. Identify creative work.
30:31 Title: Sydney and artificial love story.
30:33 Authors Kyle Shannon and Andrew Watts.
30:35 Key visuals. Theme: Analyze musical
30:39 style: drums, bass, electric guitars,
30:41 clean and slightly
30:45 distorted. Okay. The video is a
30:47 promotional piece for a fictional
30:49 Broadway musical. What do you mean a
30:51 fictional Broadway musical? Gemini is an
30:54 [ __ ] It's a real Broadway musical.
30:57 It's just just we haven't it's just not
30:59 on Broadway yet. Damn
31:02 it. I hate it when the AI is smarter
31:04 than
31:05 [Laughter]
31:09 me. Title
31:12 um type Broadway musical implied by the
31:14 play bill. Theatrical
31:17 marquee. The poster features a woman's
31:20 profile with vibrant almost digital
31:22 ethereal colorful hair looking toward a
31:24 small distant silhouette of a man. and
31:26 the background is warm theme implied.
31:30 The title and artificial love story in
31:32 the digital ethereal look of the main
31:34 female character in the artwork strongly
31:36 suggest themes of artificial
31:38 intelligence, technology, and
31:40 romance. It's quite good shot
31:46 list.
31:48 Amazing. That's what a real multimodal
31:51 response looks like. That's very cool.
31:55 All right. So, that's again
31:57 in a
31:59 studio.google.com. And then the model
32:01 that I picked was Gemini 2.5 Pro
32:09 Preview0506. And I had temperature set
32:12 at one. You're like, what's temperature?
32:14 Why do I need to deal with this [ __ ]
32:16 Because this is where the stupid AI
32:18 industry is right now. They can't figure
32:20 out if they want to be for developers or
32:22 human beings. Not that developers are
32:25 not human beings. Listen, I've known
32:27 some very nice developers that seem
32:29 almost
32:32 human, but like normal people shouldn't
32:35 have to deal with this
32:37 [ __ ] Do I want to have structured
32:39 output or function calling or grounding
32:42 with Google search? What's that
32:45 mean? Can't I just stick stuff in the
32:48 chat hole and it vomits out
32:50 brilliance? Isn't that what we were
32:58 promised? I want flying cars and I want
33:02 this [ __ ] to
33:06 work. Inspired by actual events. Yes.
33:11 Okay. So, we haven't done anything yet.
33:14 Oh, wait. You know, we were doing
33:15 something. We were looking
33:17 at this was a rabbit. This rabbit h
33:20 brought this rabbit hole brought to you
33:22 by chat add. Welcome to chat add where
33:25 we might occasionally finish a
33:29 sentence. Pinkies out everyone. We're
33:35 classy.
33:37 Okay. Fantastic. Fantastic. Yeah, that's
33:42 really good. That's really good there,
33:43 people. Yeah. Let me tell you
33:46 something. All right.
33:50 GPT40. So it inputs, it can input
33:53 documents, images, CSV files, audio, and
33:56 video. So it does input video, but then
33:59 what it does is it
34:01 just Okay, ffmpeg, if you don't know
34:04 what FFmpeg is, it's a um command line
34:09 video processing library that's been
34:12 around for decades. In fact, my company,
34:16 Storyvine, a lot of the tech we use
34:19 leverages FFmpeg. So, it's been around
34:22 forever. Um, so what that says to me is
34:26 it can input the video, but then they're
34:28 just using FFmpeg to analyze it. So,
34:30 it's not it's not true um video
34:33 multimodal, but it is image multimodal.
34:37 Uh, and it's audio multimodal because
34:39 that's where the advanced voice comes
34:40 from. So, there you have that. Okay.
34:44 So that's 4.0. You use it for everyday
34:47 tasks. I mean, I just use 4.0 for almost
34:50 everything. 4.5. Ideal for creative
34:53 tasks, emotional intelligence, clear
34:55 communication, creativity, and a more
34:58 collaborative, intuitive approach to
35:01 brainstorming. Create an engaging
35:03 LinkedIn post about AI trends. Write a
35:06 product description for a new feature
35:08 launch.
35:09 Develop a company apology letter with an
35:13 empathetic tone. All right, let's go try
35:16 that. So, we're going to go to chat
35:19 jet. We'll go to 4.5. We'll go
35:23 um write
35:27 an
35:29 empathetic apology letter. This is bad.
35:32 This too soon. This is this is before
35:35 it's happened. uh
35:38 informing
35:40 uh this is bad employees
35:44 um that
35:48 um 90% of
35:53 them aren't
35:56 necessary.
35:58 Yikes. Let's just see how it does. Let's
36:01 see how [ __ ] empathetic you are
36:02 there, chat. GPT
36:06 4.5. This is
36:08 bad. Sorry, I have no qualifications.
36:11 They're fired. You're fired. Here's a
36:14 carefully worded, empathetic apology
36:16 that acknowledges the difficult news
36:18 clearly and
36:20 respectfully. Difficult but necessary
36:22 changes team. This is incredibly hard
36:24 message to share. I want to start by
36:26 sincerely expressing my deep regret and
36:28 empathy for the impact it will have on
36:30 many of you. After careful consideration
36:33 and extensive evaluation of our
36:35 organizational structure and future
36:37 direction, we've reached the painful
36:40 conclusion that approximately 90% of
36:44 your current roles will no longer be
36:46 necessary. All right. So, that seems
36:48 very corporate. So, let's we'll do a
36:50 couple of variations here. Let's do one
36:53 that's like um the uh the Fiverr CEO
36:58 talked about radical cander. Um, let's
37:01 do one from a
37:06 CEO who
37:09 practices
37:11 radical
37:17 cander. Uh, oh, did I hang it? I think I
37:20 might have hung
37:23 it. Hello. Oh. Oh. Oh. because I'm in
37:28 uh I'm in
37:30 a
37:32 canvas. So, it's rewriting it. Team, I'm
37:35 going to be direct because honesty and
37:36 clarity are essential, especially during
37:39 challenging times. After rigorous
37:41 analysis of our business and future
37:44 trajectory, we've concluded that 90% of
37:47 our current roles are no longer aligned
37:49 with our strategic goals. This is a
37:51 stark reality we must confront openly.
37:53 To be I want to be absolutely clear.
37:55 This decision is not about performance
37:57 or capability. Well, it's about the
37:59 performance and capability of the
38:01 [ __ ] technology, isn't it? You
38:03 douchebag. It's about adapting
38:06 aggressively to a changing market
38:08 dynamics. That still sounds like
38:09 horseshit. So, let's go um let's go full
38:14 quotes
38:16 Lumbberg from Office Space.
38:23 apology letter. I write an apology
38:25 letter to Ann. Ann posted her her
38:28 smoking picture on LinkedIn. So, I think
38:30 Ann's
38:32 fine. Hey, team. Yeah. So, I'm going to
38:35 need you to go on ahead and listen up.
38:38 We've run some numbers, looked at some
38:40 things, and uh it turns out about 90% of
38:43 your jobs just aren't going to work
38:45 anymore.
38:46 Okay. Now, don't get me wrong. It's not
38:49 that you're doing anything particularly
38:51 bad.
38:52 It's just we're kind of shifting gears
38:54 here, aligning with our um strategic
38:57 stuff. You know how it is. So, I'm going
39:00 to need you to go on ahead and clear out
39:02 your pretty desks
39:05 soon. All right. So, I guess 4.5's good
39:08 at at letter writing and [ __ ] like that.
39:11 All right. Let's go back here. Um okay.
39:15 04
39:16 mini fast technical tasks quick STEM
39:20 related queries programming visual
39:25 reasoning provide a quip quick summary
39:28 of a scientific article. Oh, that's a
39:30 good one. So, let's go grab a let's go
39:32 to archive a
39:35 rxiv, right? Isn't that how it
39:45 [Music]
39:46 What was the latest one? There was a new
39:49 one that just came out. That's looks
39:51 like it's going to be the new
39:52 Transformers paper. Oh, I can't remember
39:55 what it was now. Damn it. Uh
39:59 uh
40:00 uh I
40:02 forget. Physics, mathematics, computer
40:09 science. Uh anyone remember that one?
40:13 40 rocks grounding with
40:16 Bing in wait in get on the bus or get
40:19 off the bus
40:21 style. When is Sydney going to Broadway?
40:24 I Mr. It as soon as if you got 20 first
40:28 $20 million take Sydney to Broadway. I
40:31 don't know what we're [ __ ] around
40:32 waiting for. If you got $20 million
40:35 sitting around, just DM me on TikTok.
40:38 You say, "Hey, dude. I'm ready to go to
40:40 Broadway. What's your wiring
40:43 information?" It's all we need. It's all
40:47 we need. Okay.
40:51 Okay. Which one did you find
40:54 there? Brandon found something.
41:03 Oh, I see what you did. You just did a
41:04 search. Okay, let me
41:07 look. Echo agent, pain assessment,
41:10 neural networks. There's just too many
41:11 of them. Uh, let's let's do search for
41:15 uh AI
41:24 Google. Well, it doesn't [ __ ] matter,
41:26 does it?
41:29 Let's just go grab one
41:34 PDF. Wait, but this doesn't have visuals
41:37 in it.
41:41 [Music]
41:59 Um, dang it. I want something with
42:03 pretty pictures in
42:06 it. Ped pediatric asthma detection. Give
42:09 me pretty
42:13 pictures. There we go. All right. This
42:16 one has pretty pictures. Beautiful.
42:19 Oh, charts, graphs. Lovely.
42:23 Okay.
42:25 Download. Uh, whatever.
42:29 Desktop. Go back
42:31 here. Go here. Go here. We're going to
42:33 go 04 mini. We're going to
42:39 upload this thing. Um, how how did they
42:43 put it over here? Provide a quick
42:46 summary of a scientific article. Um,
42:50 provide a quick summary of this and be
42:57 sure to
43:00 explain the most important charts and
43:06 graphs and what they mean in
43:10 context of
43:13 the summary.
43:16 I'm a dum
43:23 dum. I have no
43:26 [Music]
43:28 qualifications. Make the
43:32 summary quote me
43:40 proof. Sifting sifting through archive
43:43 is an art. It's called the abstract.
43:45 Most papers come with one. Okay, good.
43:47 Great. Oh, the summary. Yeah. No, I
43:49 understand that. I But I'm just I'm just
43:52 looking to see if this actually does the
43:55 does the
43:57 stuff. Take a two second clip of a child
43:59 breathing. Turn it into 512 number sound
44:03 fingerprints via Google
44:06 here. Thought for 60 seconds. The study
44:09 builds a super simple asthma detector
44:12 for kids using short two-cond breath
44:14 recordings and audios Google's big audio
44:18 model.
44:19 Here, here's what happens. Top for
44:22 performer separation in 2D space. Bottom
44:27 line, figure nine, page six. Figure
44:30 eight, page six. Figure seven, figure
44:34 six. Figure six. Figure 1. So it it
44:37 didn't use all of the the figures, so it
44:40 picked the most important ones. All
44:42 right, that seemed to work. All right,
44:45 fine. So that's 04 mini. And then 04
44:48 mini high detailed technical tasks,
44:51 advanced coding, math, scientific
44:54 explanations. Um, thinks longer for
44:56 higher accuracy. Hey Pete, give me a uh
45:00 give me something I can give 04 mini
45:02 high to explain something about tensor
45:05 tensor
45:06 optimization and then you can tell me if
45:09 you think it's full of [ __ ] So we're
45:10 going to go 04 mini high. We're going to
45:13 go new
45:15 chat. If you're still in here,
45:41 Pete, I think we might have lost Pate.
45:43 Did we lose Pate?
45:49 He's buffering. Okay, I'll I'll start
45:52 something and then if he comes up with
45:53 something good, I'll change it. So I'll
45:56 say
45:57 um
46:03 explain the
46:06 math that is
46:09 required to
46:13 optimize
46:16 tensor
46:20 processors for both
46:23 inference
46:26 and
46:28 training of generative AI. Wait, explain
46:32 how Ah, here we go. Ah, that you know
46:35 what is you know what's funny, Pate? I
46:37 was just about to type that. I wrote
46:38 something different, but I was just
46:40 about to write explain how to use
46:42 sharding and jacks across multiple
46:44 devices. That that was it was like I was
46:48 going to do that one, but I thought it
46:49 was too obvious. So, but I'm glad you
46:51 you put it in here. It confirmed what my
46:54 instinct was.
46:57 Explain how how to use sharding. The
46:59 only thing I know about sharding is, you
47:01 know, you got to change your underwear.
47:03 Okay. Explain
47:06 how how to use Yeah, we go there. We go
47:10 there on this channel.
47:13 sharding um in
47:17 jacks
47:20 across
47:22 multiple
47:24 devices in
47:25 a tensor parallel
47:30 manner. All right. Should I have it
47:32 explain the math? Wait, we're on the
47:34 same same wavelength. It's spooky.
47:40 It's we're really sympatico there. Just
47:43 always thinking about the same uh same
47:45 mathematics frameworks, aren't we? The
47:48 user is asking about using jacks. Okay.
47:51 Step by
47:54 step. Oh, it's doing stuff with with
47:57 Python. Build your device mesh. Okay.
48:00 So, basically what it says is one, carve
48:02 up your devices into a named mesh. Two,
48:06 declare how each array should be sliced
48:08 via partition spec. Three, put your
48:11 arrays onto a mesh with the right
48:15 sharding. Write a PJIT wrapped function
48:19 pit
48:20 function that respects those
48:24 shards. So, devices npray jacks. Devices
48:29 mesh equals mesh. Define how you want to
48:32 shard your tensors. Do do you do you
48:35 really use this language and not laugh
48:37 like a like a junior high school
48:41 kid? Shard your shard your
48:50 tensors. Define how you want to shard
48:52 your tensors. Excuse me
48:55 sir. Use partition spec to say slice
48:58 this axis across model devices. You can
49:02 ask it about how it would use that to do
49:05 something like a large matrix multiply.
49:08 Okay, so let's do that at the Okay, so
49:10 Oh, wait. There's more. Run it inside a
49:13 mesh context. Bonus older PMAP-based
49:17 approach. If you only have a 1D split
49:20 and want a quick and dirty tensor
49:22 parallelism on end devices, you can
49:24 manually slice it in PMAP. You know,
49:27 I've been manually slicing my shards for
49:30 I probably three, four years now. I cut
49:35 my finger once. It was very painful. So,
49:37 shards are acidic.
49:40 Um, okay. What did you say here? If you
49:43 tensed your shards, it would have gone
49:45 better. Okay. You can ask it about how
49:47 would you do that in something? Wait.
49:49 Okay.
49:50 How? Wait. sort of explain how um and
49:57 include
49:59 explanations of the math maths is what
50:03 we're supposed to
50:05 what oh how would
50:08 you how would you use that to do
50:11 something like okay wait how explain how
50:14 you
50:15 would do
50:19 some like a
50:22 Large
50:24 matrix
50:29 multiply. Large matrix multiply.
50:33 I usually talk about large matri matrix
50:36 multiplication when I when I teach I
50:38 substitute teach uh second graders
50:41 advanced math a lot and and that we're
50:43 often talking about large mat matrix
50:46 multiplication you know for matrices
50:48 like a n to the n b n to the n you know
50:52 what I
50:55 mean below is how you take a huge matrix
50:58 multiply x= y = xwa
51:03 X equals I think it's sum R fancy R B to
51:08 the A B * H raised to the B * H LinkedIn
51:13 question Tim Lloyd Kyle happy Friday
51:16 what do you know about Enzo AI agents
51:18 anyone um I don't know anything about
51:20 Enzo but we can go look them
51:23 up see the problem with me having it
51:26 explain the math is I can't even read
51:28 the [ __ ]
51:30 formulas this does this look good. Yes,
51:33 this is how you would do it. Pjit is
51:35 outdated
51:39 though. My dissertation was about large
51:41 matrix operations. Wow,
51:44 wild. Um, so I'm going to I'm going to
51:46 tell I'm going to say I'm going to say,
51:48 hey, listen.
51:50 I have a
51:53 um
51:57 large
52:00 associate who informed
52:05 me
52:06 that PIT wait PIT is
52:12 outdated. Get with the program.
52:18 Straighten up and fly right, you
52:27 stupid. You big dumb large language
52:30 model. Straighten up and fly right.
52:33 Okay, let's see what it says. Enzo.bot's
52:36 agentive
52:38 marketplace at 49 bucks a month. Oh,
52:42 yeah. I don't the the the there may be
52:45 commercial agent marketplaces out there
52:48 that
52:49 are so
52:51 both
52:52 Claude, Gemini or all three claude,
52:55 Gemini and Open AAI all have some
52:57 agentic things in their APIs that you
53:00 can go start making agents.
53:03 Um so they might be very real, but I I
53:06 don't I don't know. I'm kind of waiting
53:08 for one of the big three to to drop the
53:11 sort of the chat GPT equivalent for
53:13 agents,
53:15 right? Like custom GPTs were where you
53:18 don't have to know any programming. You
53:19 can just build them. Like we don't have
53:21 that for agents yet. So, I'm just kind
53:23 of waiting for that thought for 20
53:25 seconds. The core idea. Yeah, Pate. I I
53:28 think um chatbt is smacking you down
53:31 here. The core idea hasn't changed. The
53:34 heavy lifting is still split the big sum
53:37 to the h n= 1 x to the i comma n w to
53:42 the nj across d devices. I mean duh,
53:46 right? That we knew like you still got
53:49 to do that. Then you got to compute each
53:51 partial sh sum on its
53:54 shard. Honest to god, you really do need
53:57 a gen xer in the corner going, "Guys,
53:59 you got you've got to change the name
54:01 shard. You just you can't
54:03 You you you have to change it. You have
54:06 to change it. You're you're adult people
54:09 talking about sharding things. You can't
54:11 do
54:13 that. Okay. Then you have to all reduce
54:17 pum the d partials into the full result.
54:20 What has modernized is how you tell Jax
54:22 about your
54:24 shards. As of Jax 0.4.11 plus you import
54:30 pjit directly from Jax. No more
54:32 jacks.experime. experimental. Jack's
54:34 experimental was always temporary. I I
54:37 think it's in the name. It's
54:39 implied. All right. Hey, chat GBT. I
54:42 know the underlying idea is the same,
54:43 but there's easier ways to do it. Oh,
54:45 wait. That's
54:46 good. This is good. This is good.
54:49 Thanks, P. You're This is good. You're
54:51 like my uh You're like my hitman here.
54:54 Okay. Uh I
54:57 understand the underlying uh what' you
55:01 say idea is the
55:06 same but there are
55:11 easier ways to do it. Think
55:18 harder.
55:20 Let's not be ninnies.
55:31 thinking. See, you know what the problem
55:33 is with doing something like this on my
55:35 channel, like on in my account, is chat
55:38 GPT is now going to think I understand
55:40 any of this [ __ ] and it's going to
55:41 start including mathematical equations
55:44 in my screenplays. I don't like it. Math
55:47 in three lines.
55:48 Okay, here's the simplest, don't
55:51 overengineer it way to spread a big X Y
55:54 equals XW
55:59 X math equation across D devices using
56:03 PMAP. You write
56:07 that then you break it into D equal
56:09 chunks H divided by D. Then you recover
56:13 in full y by y equals sum d minus1 dals
56:18 0 y to the parentheses dig which is just
56:22 all reduce over the
56:25 partials. Did it do good? Gemini agrees
56:29 the functionality is now baked directly
56:31 into jacks.jit maths of the
56:34 humanity chunky shards. Exactly. I think
56:38 listen here's what we've learned
56:40 tonight. Sure. 04 mini high can probably
56:44 do some analysis of your sharding in the
56:46 tensor optimization game, but we still
56:49 got to change the name of sharding.
56:51 We've get you it seriously needs a
56:54 rebranding. We can't be saying it with a
56:56 straight face. It's not appropriate.
56:59 It's not
57:00 professional. All right. Um Okay. So now
57:05 that we have that out of the way,
57:12 um,
57:14 sharding is
57:17 dangerously
57:19 close to
57:23 sharding and it is
57:26 making
57:28 non
57:31 engineers giggle.
57:34 and
57:37 uncomfortable. We need a
57:39 rebranding and we need it
57:44 now. Jeff on YouTube, I know you
57:47 understand what you think I said, but
57:50 I'm not sure you realize what you heard
57:52 is not what I meant exactly.
57:56 Okay, let's let's shoot sharting off
57:59 into a into the bit bucket and give our
58:03 multi-device split a fresh ungly
58:06 identity. Here are five contenders.
58:10 Tiling, API sketch, tile mesh, tile spec
58:14 in tiles, slicing, partitioning,
58:17 spreading. Spreading is almost as bad,
58:20 and stripping or
58:22 striping. I like tiling striping. I
58:26 don't know.
58:28 Anyway, I can't see everything, but the
58:31 ideas are about all right in a kind of
58:34 hand wavy, so it's not wrong. Oh, that's
58:35 cool. All right, good. All right. So, so
58:38 that's what OpenAI claims that's good
58:40 for is fancy math. And then open OpenAI
58:44 03 complex or multi-step test, strategic
58:47 planning, detailed analysis. Hang on.
58:53 extensive coding, advanced math, science
58:56 coding, visual
58:58 reasoning, and then 01 pro complex
59:01 reasoning takes a bit
59:03 longer, enterprise limits and
59:05 capabilities. Anyway, all right. So,
59:07 there we have that. There's when to use
59:09 which models. Um, just use
59:13 40. Use 40 until until it breaks and
59:17 then try some other [ __ ] That's
59:19 basically it. We need a Mother's Day
59:21 palette cleanser.
59:24 Um, yeah. Let's see. What do we want to
59:28 do? Oh, you know what I heard
59:30 today from Brent Brent
59:34 Peterson? Um, let me go back to
59:38 40 and I'm going to say,
59:41 um, I want you to come up
59:44 [Music]
59:46 with 20
59:49 novel Mother's
59:52 Day card
59:55 concepts. 10
59:57 are
59:59 gutbusters funny. Gutbuster
1:00:03 funny. and 10 are
1:00:10 tearjerkers.
1:00:12 And all of them have to be we should try
1:00:18 this in 4.5 as well. All of them have to
1:00:21 be
1:00:26 original in ways we can't currently
1:00:32 imagine. So, Sunday is Mother's Day.
1:00:35 Music, poem, art. Yeah, something like
1:00:38 that. All
1:00:41 right. My eyes are getting all dry.
1:00:43 Something must be going on. Oh, you know
1:00:45 what's going on? I think I think I think
1:00:47 there's some pollen happening here in
1:00:49 Yield, Denver. I don't I don't have a
1:00:51 lot of allergies, but my eyes are really
1:00:53 really
1:00:55 uh like someone threw pollen in
1:00:59 them. Okay. Gutbuster funny cards.
1:01:03 Your womb, my first apartment. That's
1:01:05 pretty
1:01:06 good. Front blueprint style diagram of a
1:01:09 womb with stick figure furniture inside.
1:01:13 Rent was free and the food was delivered
1:01:15 introvenously and the landlord loved me
1:01:17 unconditionally. Best landlord ever.
1:01:20 That's pretty cute. Mom, I googled you.
1:01:22 Search bar that autofills with, "Why is
1:01:25 mom always right?" Turns out it's
1:01:27 genetic. You pass down your sass,
1:01:29 sarcasm, and supreme
1:01:32 correctness. Mom math. One brain plus
1:01:35 five kids equals
1:01:38 Huh. All right. Tearjerkers. The first
1:01:40 voice I ever heard. Ultrasound wave.
1:01:43 Okay. So, here's what I want to
1:01:45 try. So, it made a bunch of things. So,
1:01:47 now I'm going to
1:01:49 say make the
1:01:52 first five
1:01:55 cards covers.
1:02:00 So, the jobs at Hallmark are still safe
1:02:02 for now. Yeah. Um, Brent Peterson was
1:02:05 saying
1:02:07 today in one of his backend systems, he
1:02:10 was running a test
1:02:13 and I think he said he was using
1:02:17 03 and it decided it needed it needed to
1:02:20 make nine images and he said it just
1:02:22 went and made nine images and didn't
1:02:24 stop. So, I'm going to see if that's
1:02:26 something that they have have added in
1:02:28 here. After it makes one image, is it
1:02:30 just going to make another
1:02:35 one? Oh, most of the terms when it when
1:02:38 it came up with new sharding terms, most
1:02:40 of the terms were already used. Yeah, I
1:02:42 was wondering about that. Tiling
1:02:43 certainly seems like something that
1:02:45 would have been
1:02:46 taken. Your womb, my first apartment.
1:02:54 Well, it's a unique
1:02:58 card and it didn't do all five. Yeah.
1:03:02 So, all right, that failed. Okay. So,
1:03:04 what you should do is you
1:03:10 should here. Here's I'll give you
1:03:12 homework right now for the weekend. And
1:03:14 you can start doing this now. And maybe,
1:03:16 in fact, maybe what you do is, you know,
1:03:18 for the rest of the little live here
1:03:20 tonight, Friday night date night, make
1:03:22 some Mother's Day cards and share them
1:03:24 in in the uh in the Irregulars channel
1:03:28 on the AI Salon. Have you used Dreamface
1:03:31 app? I used it for a baby talking video.
1:03:33 Pretty simple. No, I used I used Hedra.
1:03:37 Um, right now I'm I'm kind of avoiding
1:03:41 all of the apps because there are so
1:03:43 many apps that are being dumped into the
1:03:45 app store right now that are basically
1:03:47 just calling base models. Um, so I'm
1:03:50 just trying to avoid that for now. But I
1:03:53 mean, the nice thing about those kind of
1:03:54 apps is that they do make it really
1:03:56 simple. The downside is they're really
1:03:58 [ __ ] expensive.
1:04:06 Uh, James Duder sounds rat to
1:04:12 me. It's a website, not an app. What's
1:04:14 it called? Dreamface
1:04:18 app. Dream face
1:04:21 app. And does it cost you anything?
1:04:27 [Music]
1:04:35 [Music]
1:04:59 Freely swap faces. Voice studio
1:05:03 comes. Clone voices.
1:05:06 Convert. Advanced avatar has
1:05:10 launched. Audio. Record
1:05:15 it. AI
1:05:17 script. Um, talk
1:05:21 about Mother's Day.
1:05:33 Plus photos and
1:05:35 videos. Got
1:05:49 it. That's JPEG image.
1:05:56 Why did it not like
1:06:00 that? Oh, it did. It did like
1:06:09 it. Voice. Baby Rose.
1:06:30 generate one minute video was free than
1:06:33 it costs. Okay. Yeah, that makes sense.
1:06:35 Tab when it generates. All right, I'll
1:06:37 go do the tab now.
1:06:41 [Music]
1:06:48 I assume what this is using is, you
1:06:51 know, APIs into Clling or or or Hedra or
1:06:55 or some of those. It's probably using 11
1:06:57 Labs for the voice is my
1:06:59 guess. I mean, quite
1:07:02 frankly, nothing here yet. Generation.
1:07:05 Oh, generating completed. Here we go.
1:07:09 That was quick. Is it any
1:07:11 good? Let's see.
1:07:32 Why am I seeing something different over
1:07:33 there? Oh, can't hear it. Oh, did I
1:07:36 share the wrong screen? Oh, I see what
1:07:38 it did. It It spawned a new screen. I'm
1:07:41 like, why can I not see it the same way
1:07:43 I'm seeing it?
1:07:48 and Mother's Day. This uh they're
1:07:51 probably using um they're probably using
1:07:54 Hey Jen here because this is almost
1:07:56 exactly what Hey Jen looked like. Jeff
1:07:58 Flanigan and Mother's Day card using the
1:08:00 word
1:08:04 sharding. Yeah, it's pretty good. Serves
1:08:07 as a poignant reminder of the profound
1:08:09 influence mothers exert in our lives. It
1:08:12 is a day dedicated to honoring their
1:08:14 unwavering support, selfless sacrifices,
1:08:17 and unconditional love. Celebrating this
1:08:19 occasion fosters appreciation, and
1:08:22 gratitude, allowing us to reflect on the
1:08:24 invaluable lessons learned from these
1:08:27 remarkable.
1:08:29 All right. Yeah, it
1:08:31 works. It works. It works. It works. Um,
1:08:38 yeah. I I just I don't I mean I can't
1:08:43 keep up with the base models, much less
1:08:45 apps like these.
1:08:48 I It's
1:08:50 just It's
1:08:57 like I'm trying to think I'm trying to
1:09:00 think if there's a if there's a good
1:09:04 strategy. I'm trying to think if there's
1:09:06 a good strategy to not be
1:09:12 to be
1:09:13 decently articulate about what's
1:09:16 possible with
1:09:18 AI. But it's like one of the things that
1:09:20 I've been experiencing in the
1:09:23 past two
1:09:25 months is that a lot of people that I
1:09:28 know that are very educated about
1:09:30 AI haven't heard about major new things
1:09:34 in certain sectors of AI. So, I think
1:09:36 people are kind of narrowing into in I
1:09:39 don't know if it's
1:09:41 verticals, but like there's too much to
1:09:43 keep up with. So, I I don't I don't have
1:09:44 a good strategy for it other than
1:09:47 just [ __ ] do everything you can to to
1:09:50 keep on top of it because I there's just
1:09:52 not a way to there's just not a way to
1:09:54 do it. What did you say it might be the
1:09:56 same as?
1:09:59 Um Oh, this this looks almost identical
1:10:04 to Hey Jen. So yesterday we did Hey Jen.
1:10:07 So here's the Hey Jen video. Oh wait,
1:10:09 hang on. I got to share
1:10:18 it. Is too good. So this is Hey Jen from
1:10:21 last night. So yeah. So yesterday was my
1:10:25 birthday, May 7th. And see how the
1:10:27 balloon doesn't move and just his head.
1:10:29 A little bit of hand movement and just
1:10:31 his face. You just want to write it down
1:10:32 and put it in your calendar or whatever.
1:10:33 But yeah, yesterday I uh I uh I
1:10:37 technically, you know, like on a
1:10:40 calendar using math uh turn 60. So So
1:10:45 that's Hey, Jen. But then
1:10:51 Hedra, come
1:10:53 on. I got to switch this again. Hang on.
1:11:07 So yeah, so yesterday was my birthday,
1:11:10 May 7th. And you know, in case you're in
1:11:12 case you want to write it down and put
1:11:13 it in your calendar or whatever, but
1:11:14 yeah. Yes. I mean, that's just a very
1:11:17 different level of quality, right?
1:11:21 So yeah, so yesterday was my birthday,
1:11:23 May 7th. you know, in case you're in
1:11:25 case you want to write it down and put
1:11:26 it in your calendar or whatever. But
1:11:28 yeah, yesterday I uh I uh I technically,
1:11:32 you know, like on a calendar using math
1:11:36 uh turned 60, but you know what? So
1:11:40 anyway, all
1:11:45 right. So what I would
1:11:47 say you should do over this here weekend
1:11:51 is you
1:11:55 should go above and
1:11:57 beyond using AI to make Mother's
1:12:01 Day
1:12:03 joy. Right? Make a poem, make a song,
1:12:06 make an
1:12:07 image, make a video. Turn yourself or
1:12:11 one of the kids into babies
1:12:13 again. Turn those into videos.
1:12:16 saying the poem. I don't know. Like just
1:12:19 do like a
1:12:21 multimodal, you know, explosion of joy.
1:12:25 Um, and have chat GPT help you figure
1:12:28 that [ __ ] out or whatever, whichever one
1:12:30 you like. Uh, this is, uh, Hedra. H E D
1:12:35 R A is how you spell
1:12:40 it. Yeah, this one's much better. Hedra,
1:12:43 I think, is how they pronounce it. Hedra
1:12:45 is much
1:12:53 better. All right. What else? Questions?
1:12:59 Thoughts? Thoughts? It's Friday
1:13:10 night. Nah.
1:13:20 [Music]
1:13:21 Kyle, what did you use? What did I use
1:13:24 for
1:13:25 what? To make that baby
1:13:29 video. Hedra. H E D R
1:13:33 A is the tool that I used for the
1:13:35 animation.
1:13:39 Check out my images in Look what I made.
1:13:45 Fine. Look what I
1:13:47 made. Look what I
1:13:54 made. Oh, nice. Oh, that's
1:13:58 awesome. Oh, I'm not sharing. Should I
1:14:00 share?
1:14:09 That is beautiful. That is gorgeous from
1:14:12 from one
1:14:18 Steo. Sydney on Broadway, baby. It's
1:14:22 coming. Who's got $20 million?
1:14:29 Seriously. FO has been coaching me in
1:14:31 several areas. Very cool. You said at
1:14:35 the beginning that some site came out
1:14:36 with something new and we would look at
1:14:38 it. I think what I was talking about was
1:14:41 Gen Spark. We can go look at
1:14:43 GenSpark. Let me I like this
1:14:47 image. Let me
1:14:48 go Steo. I'm going to put your Steo put
1:14:53 your uh your X
1:15:02 um put your X handle in in whatever you
1:15:07 call
1:15:08 it. Uh Kyle Sydney cartoon.
1:15:30 Thanks. Scroll
1:15:33 down. Go
1:15:36 down. Oh, okay. I I'll look at there in
1:15:39 a
1:15:40 second. Uh thanks to at Steo Steo. Okay,
1:15:44 here.
1:15:47 Steve_ore. Is that 042 or 042? Looks
1:15:50 like a zero.
1:15:54 Underscore42. No, must be an
1:16:05 O. There it
1:16:09 is for
1:16:11 this dream
1:16:14 board
1:16:24 image. Where do I save that?
1:16:27 [Music]
1:16:35 Son of a
1:16:37 [ __ ] Sorry. Let me change the format
1:16:40 of
1:16:44 this so that I can use it
1:16:48 in
1:16:53 software. So [ __ ] exhausted.
1:17:06 Um, coming to a Broadway near you.
1:17:32 [ __ ] All right, heading over to
1:17:35 Irregular's
1:17:37 channel.
1:17:41 Irregulars. I asked artificial
1:17:44 intelligence write your mother's day
1:17:45 message. What's it say inside?
1:17:50 Dear mother unit, your contributions to
1:17:52 this household. Functionality are noted.
1:17:55 Meal preparation above average.
1:17:57 Emotional support
1:17:59 immeasurable. Laundry throughput high
1:18:01 volume. Minimal complaint. Affection
1:18:03 level
1:18:04 100. So I fired the robot. That's nice.
1:18:08 That's
1:18:09 cute. Wait, did it match the Yeah.
1:18:12 format? It did. The in the inside
1:18:14 matches the outside. That's cool. We
1:18:16 made it, champ. That's awesome. All
1:18:18 right.
1:18:20 Um, let me go
1:18:24 to GenSpark.
1:18:27 Gen S P R A
1:18:31 K.AI. All
1:18:35 [Music]
1:18:38 right. In Hydra, you mean Hedra? It's
1:18:42 not
1:18:43 Hydra, Mr. He hedra or
1:18:47 hedra, not
1:18:49 hydra. What's
1:19:04 that? Both Kyle and the new pope are in
1:19:07 this their 60s. That's rude.
1:19:14 Uh, I can't find it, Mr.
1:19:20 [Music]
1:19:22 It. Oh, look what I
1:19:27 [Music]
1:19:30 made. Nice date night. A AI learning
1:19:33 lab. Little popcorn. I like the unicorn.
1:19:40 There's something from Gareth.
1:19:44 Vicki Steo, why do I look so sad? Oh,
1:19:48 two hours into the live, three tequilas
1:19:50 [Laughter]
1:19:54 down. I got no Mr. It. I got nothing.
1:19:58 You've been You've been buried, Mr.
1:20:03 It. Someone is squatchching your voice.
1:20:08 All right, let's go to Jen
1:20:12 Spark. So, I figure what we'll do is
1:20:14 we'll watch this video together because
1:20:16 we watched this on office hours
1:20:19 today
1:20:20 and it's pretty
1:20:23 something.
1:20:26 Now, here's the caveat.
1:20:29 Caveat is that this is a marketing
1:20:31 video. So, as Pate rightfully points out
1:20:35 whenever I show one of these, he's like,
1:20:37 "Yeah, how much of this is real?" So, I
1:20:39 don't know how much of this is
1:20:42 real. I already did tabs, didn't I? I
1:20:46 thought I did this. I thought I did it.
1:20:48 Did I not do it right? Hang on. Hang on.
1:20:51 I did it right. But whatever. I'll do it
1:20:54 again.
1:20:56 Okay. Thank you, Winston.
1:21:08 Hey, we are shipping GenSpark AI sheets,
1:21:10 a full logentic tool that can help you
1:21:12 from extensive data collection, analysis
1:21:14 to
1:21:15 visualization. Forget Excel, now you
1:21:18 have a personal data analyst. Here's the
1:21:20 raw data of all marketing campaigns from
1:21:22 a company over the past year. Upload
1:21:24 these four Excels to Genpar. Genspark
1:21:26 quickly transformed this chaos into a
1:21:28 clean, structured AI spreadsheet. Now,
1:21:31 just ask, which type of campaign has the
1:21:33 best performance? Genpark immediately
1:21:36 executes code and applies formulas. It
1:21:38 generates not only charts for different
1:21:40 channels, but also a radar graph
1:21:42 highlighting email campaigns as the
1:21:44 winner. Let's probe deeper. What's the
1:21:47 correlation between content variation
1:21:49 and performance?
1:21:52 Genspar quickly finds negative
1:21:53 correlation with ROI and recommends one
1:21:57 to four variations as the optimal range.
1:22:00 Finally, show me a comprehensive
1:22:02 performance report. Within moments,
1:22:05 GenSpark delivers an executive summary
1:22:07 with clear data visualizations and
1:22:10 talking points addressing all key
1:22:11 questions your boss might ask. Genpark
1:22:14 can automatically find companies,
1:22:16 people, products, or any data you need.
1:22:20 Let's ask Grenark to find healthcare
1:22:21 startups that are at series A or series
1:22:24 B stage founded after 2020 with their
1:22:27 most recent funding round occurring in
1:22:29 2024 or later. Within minutes, GenSpark
1:22:33 delivers a comprehensive spreadsheet
1:22:35 providing
1:22:37 um Lissic or or Liss Lissionic or Lis 10
1:22:42 Inc. Um what are you trying to do? I'm
1:22:45 an AI query expert. Uh I'm not trying to
1:22:48 do anything. I'm just watching a video
1:22:49 from GenSpark which just launched sheets
1:22:52 which looks pretty remarkable and gen
1:22:55 and just for context Gen Spark's super
1:22:59 agent and their slides generator is
1:23:02 pretty remarkable already. So I'm
1:23:05 assuming this is going to be pretty
1:23:06 good. I haven't played with it a ton
1:23:07 though. 9 companies, all with recent
1:23:10 funding in 2024, founded after 2020 at
1:23:13 Series A or B, complete with industries,
1:23:16 founders, and
1:23:17 investors. Oop, that music is so
1:23:20 brutally bad, isn't it? This is, you
1:23:23 know what this music is? I promise you,
1:23:27 this is not
1:23:29 um pseudoenerated music. I I guarantee
1:23:32 you what this is is this is some of that
1:23:34 loopy [ __ ] stock music that you know
1:23:37 you pay $10 for the year and get, you
1:23:39 know, 10,000 songs. This is one of those
1:23:41 10,000 shitty songs. Next, if you're a
1:23:44 recruiter, tell Jensen Spark. Identify
1:23:46 senior UX designers with hyperspecific
1:23:49 needs. Watch as it compiles profiles of
1:23:51 qualified specialists with their
1:23:53 experience in AR or VR, their
1:23:55 proficiency in Unity and Unreal Engine,
1:23:57 and experience with eyetracking
1:23:59 technology. For brands tracking their
1:24:02 social media performance, GenSpark can
1:24:04 search YouTube and aggregate data from
1:24:07 all relevant videos, summarizing views,
1:24:10 likes, and comment analysis in one
1:24:12 convenient dashboard. Hey everyone, in
1:24:14 this video I'm going to be showing you
1:24:15 an AI agent. AI sheets can do much more
1:24:18 beyond traditional
1:24:20 formula. You can use LLM as your new
1:24:23 formula, processing all the data
1:24:25 throughout your
1:24:27 sheets. If you're a student preparing
1:24:29 for your AP Biology exam, simply upload
1:24:32 your most recent homework. Genpark
1:24:35 analyzes it against key concepts in this
1:24:37 chapter, color codes your proficiency
1:24:39 levels, and creates study materials
1:24:41 targeting exactly what you need to
1:24:43 improve. That's crazy. If you manage
1:24:46 e-commerce for products such as Stanley
1:24:48 tumblers, just up Wait, pay attention to
1:24:51 this, right? Just upload pictures of
1:24:53 your Stanley tumblers. Watch what this
1:24:55 does. Upload images of your
1:24:57 products. Jensen generates professional
1:25:00 advertising visuals in the gym, home,
1:25:03 and outdoor settings, scaling your
1:25:05 product photography in just
1:25:09 minutes. If you're a healthcare software
1:25:11 sales rep, ask for potential clients.
1:25:14 GenSpark delivers a list of qualified
1:25:16 prospects in your target region,
1:25:18 complete with key contact persons,
1:25:20 current software, and sales
1:25:21 opportunities. It can even drive
1:25:23 personalized outreach emails, plus
1:25:25 one-click buttons to send them. Go to
1:25:28 genpark.ai and find GenSpark AI sheets
1:25:30 today.
1:25:32 All right, so let's go try
1:25:34 something. Let's see if this is any
1:25:37 good. Is it any [ __ ] good?
1:25:40 Genpark.ai
1:25:41 AI open AI called they want their set
1:25:45 back. Yeah. What is it about I saw that
1:25:49 from source camp. Why do they all have
1:25:51 the same stupid [ __ ]
1:25:54 set? All right. So, here's
1:25:59 sheets. Okay. So, what I'm going to do
1:26:01 is I'm going to say
1:26:06 um let's go over to chat GBT for a
1:26:09 second.
1:26:14 Based on everything you know about my
1:26:19 musical
1:26:22 Sydney, write
1:26:25 up a primer about the show, the authors,
1:26:35 um, and anything you
1:26:38 think might be important to
1:26:46 identify
1:26:49 agents and
1:26:52 producers. All right. So, we're going to
1:26:54 Oh, am I not sharing? God dang
1:27:03 it. So, I'm getting chat GPT. Since chat
1:27:06 GPT knows me, I'm going to get it to
1:27:12 tell. All right, I got all this [ __ ]
1:27:15 right. This is actually quite amazing
1:27:20 that the memory function in chat GPT is
1:27:28 actually it's pretty stunning that I can
1:27:31 just say write up [ __ ] about my musical
1:27:39 Sydney. Would you like a PDF pitch deck
1:27:42 version of this?
1:27:44 uh
1:27:47 PDF
1:27:49 pitch
1:27:51 PDF version is
1:27:55 fine just format nicely. I don't
1:28:00 need
1:28:04 slides, just a
1:28:06 doc. And I'll show you why I'm doing
1:28:08 this in a second here.
1:28:20 All
1:28:22 [Music]
1:28:26 right. All right. So now chat GPT is
1:28:30 generating a PDF document using
1:28:33 Python which if you haven't been if you
1:28:36 haven't been kind of digging in deep
1:28:37 into Chat GPT recently it's pretty slick
1:28:40 that would you like me to export this as
1:28:42 a polished PDF as well?
1:28:44 Yes. Let's have it do a polished PDF and
1:28:47 see what that looks like. Model not
1:28:50 found error.
1:28:53 Ouch. All
1:29:11 right, let's go look at
1:29:18 this. Did it get everything?
1:29:44 You know what? I think I'm going to go
1:29:47 back to the original. I think in writing
1:29:49 all that it missed a bunch of
1:29:56 [ __ ] Maybe it
1:29:58 didn't. Whatever. Okay. So, let me just
1:30:01 copy all
1:30:06 this. So, here's the write up of our
1:30:09 show with contact information. All
1:30:12 right. Then, we're going to go back to
1:30:14 GenSpark and I'm going to say each data
1:30:16 search. I'm going to
1:30:20 go here is some information on my new
1:30:28 musical Sydney.
1:30:33 I want you
1:30:37 to
1:30:38 find
1:30:40 producers and
1:30:42 agents you think would be
1:30:47 good and put them in a
1:30:51 sheet with
1:30:54 [Music]
1:30:57 customized
1:30:59 Outreach. Wait, outreach emails for
1:31:04 each.
1:31:06 Also include why you selected
1:31:10 them and
1:31:13 be sure
1:31:15 um
1:31:25 to make
1:31:29 the email
1:31:33 um
1:31:35 resonate with that core
1:31:43 reasoning and then just paste in all
1:31:45 that [ __ ] I just did and we'll [ __ ]
1:31:48 turn it loose. Let's see what it does.
1:31:50 Tik Tok comment
1:31:54 pinned. I don't see it.
1:32:01 millions of jobs gone just like that.
1:32:05 Um
1:32:08 I sure like sure I I I would say
1:32:17 I millions of tasks that human beings do
1:32:21 today are going to be automated. It
1:32:24 doesn't necessarily mean that jobs are
1:32:26 gone. What it means is there's going to
1:32:27 be new kinds of jobs. What I can tell
1:32:30 you is that if you're sitting on the
1:32:32 sidelines or if anyone is sitting on the
1:32:34 sidelines just bitching about the fact
1:32:37 that AI is here, it doesn't it's not
1:32:39 going to make it go away. The people who
1:32:42 are going to fare
1:32:44 better when job displacement comes are
1:32:48 people that know how to use AI. The
1:32:50 whole purpose of this channel is not to
1:32:53 say AI is good, not to say AI is bad,
1:32:55 not to say AI is anything other than to
1:32:57 just say AI is here. And what this
1:33:01 channel is about is let's understand
1:33:02 what it makes
1:33:04 possible. And then once you understand a
1:33:07 tool like
1:33:08 um like GenSpark, which is really quite
1:33:12 a remarkable research tool, then you can
1:33:15 choose you've got two choices. you can
1:33:18 choose to engage with it or not. And if
1:33:22 not, just know this is what you're going
1:33:24 to be competing against. And if you do
1:33:26 choose to use it, and particularly if
1:33:28 you're good at data analysis and you're
1:33:30 already good at querying,
1:33:33 um, you're going to be able to look at
1:33:34 the output of a tool like this and use
1:33:37 it in a much more sophisticated way than
1:33:40 say some bozo with an acting degree can.
1:33:43 Um, but I would I would caution you if
1:33:46 if the if your comment is, you know,
1:33:49 you're sort of casting stones from the
1:33:51 sidelines with AI, I would encourage you
1:33:54 to get your ass curious about it sooner
1:33:56 than later.
1:33:58 Um, oh, and how do I raise $20 million?
1:34:01 Exactly. All right. So, let's see what
1:34:04 it's doing here. Um, what did it do?
1:34:07 Stacy Mitch, producer. Dear Evan Hansen,
1:34:11 The
1:34:12 Crucible. Deep
1:34:14 experience.
1:34:16 Techfocused. This is cool. Deep
1:34:19 experience with tech focused theatrical
1:34:23 work. Dear Jordan, your reputation for
1:34:26 championing bold, innovative works that
1:34:29 push the theatrical boundaries while
1:34:31 maintaining
1:34:32 artistic excellence makes you an ideal
1:34:34 partner for our new musical Sydney. Much
1:34:37 like your productions of Spring
1:34:39 Awakening, an American Idiot, redefining
1:34:41 the musical theater could explore,
1:34:43 Sydney examines, this is good. This is
1:34:45 really quite
1:34:48 good. So, it's basically making
1:34:52 me It's making me a [ __ ] target list
1:34:56 of who to reach out to with with a an
1:34:58 email per
1:35:00 person. Absolutely amazing. Absolutely
1:35:03 amazing.
1:35:08 Sorry if I distract you with gifts being
1:35:11 OCD. Um, just taking my coin balance
1:35:15 down to the next lowest prime number
1:35:17 like 1361 then 1259 where I am
1:35:28 currently Winston. That is the best.
1:35:31 I it never occurred to me that you could
1:35:34 you could do some fantastic things like
1:35:37 prime number coin management in Tik Tok,
1:35:40 but yeah, of course you can. Prime
1:35:41 numbers are everywhere. That's [ __ ]
1:35:45 awesome.
1:35:46 Um, let's see.
1:35:52 Uh, well, let's just let this thing keep
1:35:57 going. This is really quite amazing.
1:36:01 I I mean, honest to
1:36:03 God, like if you look at what this is
1:36:06 doing and your immediate response is
1:36:08 there, you know, a million jobs are down
1:36:10 down the
1:36:11 drain, you know, I would encourage you
1:36:14 to look at this and say, what jobs does
1:36:16 this make
1:36:18 possible, right? or what what
1:36:22 superpowers does a capability like this
1:36:24 give to an individual that might get
1:36:26 laid off that can now maybe do something
1:36:30 at a dramatically higher level than they
1:36:33 could have otherwise
1:36:35 right
1:36:38 I you know again two ways two ways to
1:36:44 respond to AI one is ignore it and one
1:36:48 is embrace it ain't going away. The only
1:36:51 thing I know is if you ignore it, it's
1:36:53 going to happen to
1:36:56 you. And that's not going to be pleasant
1:36:58 for [ __ ]
1:37:03 anyone. This is so
1:37:06 cool. It's been populated with six
1:37:08 producers who are
1:37:12 excellent. All right.
1:37:14 Producer, theatrical
1:37:41 owner. This is amazing. Look at the like
1:37:43 it it wrote into the into this
1:37:46 personalized email. The show asked
1:37:48 timely questions about consciousness,
1:37:51 personhood, and
1:37:53 intimacy. I can't
1:37:55 it's because it's working. I
1:38:04 can't h It's fascinating. This
1:38:08 absolutely fascinating. Is it giving me
1:38:10 their contact information?
1:38:14 Let me add something in here. Um, it
1:38:17 looks like we are missing
1:38:21 contact information for the
1:38:26 producers. I
1:38:28 need
1:38:30 everything possible
1:38:34 to get the word
1:38:38 out. Yeah, I can't do it while it's
1:38:40 working. So, that's a downside. So,
1:38:43 there's So, this is very much like
1:38:45 Manis. Um, one of the nice things you
1:38:48 can do with Manis is you can talk to it
1:38:50 while it's working. This one, it doesn't
1:38:52 look like you
1:38:55 can. All right. So, while that thing's
1:38:57 thinking, that's still going,
1:39:02 right? Eight contacts so far, seven
1:39:06 producers, one literary agent. All
1:39:09 right, let me go look at uh Mr. It's got
1:39:12 something in irregulars
1:39:19 here. Oh, it's a birthday video. Nice.
1:39:23 Let's go look at that. Hang
1:39:26 on. This is going to make me
1:39:27 uncomfortable, isn't
1:39:32 it? I don't deal with
1:39:36 uh Let's see.
1:39:39 An error
1:39:40 occurred. Bad playback
1:39:46 ID.
1:39:48 Huh? Close
1:39:53 that. It's
1:39:55 broken. Posted a second video below
1:39:58 that.
1:40:02 Kyle Shannon, thank you, my friend, for
1:40:05 changing my life. You and AI have
1:40:08 shifted the course of my dreams, my
1:40:10 mind, and my heart. AI has expanded my
1:40:13 creativity and highlighted the
1:40:15 intelligence and passion I carry inside.
1:40:18 Now I only need to better my ADD
1:40:20 medicine dosage to work
1:40:22 harder. Thank you irregulars for being
1:40:25 my e friends. Our realities may be
1:40:28 different, but somehow keeping our eyes
1:40:31 on AI has guided us toward a common
1:40:33 goal. I truly hope that in the future
1:40:37 that goal will belong to everyone, not
1:40:40 just a few. During the day, I wear many
1:40:43 hats. Among the many projects I juggle,
1:40:47 infinite AI is the one closest to my
1:40:49 heart. I carve out time each day to
1:40:52 build it. But I can't do it alone. Right
1:40:55 now, I don't have the funds to invest in
1:40:57 my dreams. The only treasure I have is
1:41:00 your friendship. Can you follow me? Help
1:41:03 me. Maybe even a like, a follow, or a
1:41:07 kind word means the world. Kyle, will
1:41:11 you follow me? Thank
1:41:14 you. That's awesome. That's awesome. Are
1:41:17 you sure you made that in Hedra? H E D R
1:41:20 A or did you do it in some some other
1:41:22 tool? Because that Hedra looks like it's
1:41:25 more sophisticated than than what that
1:41:27 did. Um I'm curious where that was done.
1:41:33 That's
1:41:35 sad. Oh, yeah. Okay, let me go back to
1:41:39 share my tabs the right
1:41:41 [Music]
1:41:45 way. But thank you very much for the
1:41:48 kind
1:41:51 [Music]
1:41:56 thoughts. All right, are we done here?
1:41:58 No, we're still going. So, what do we
1:42:01 have now? We've got two literary
1:42:05 agents. Dear Beth,
1:42:24 [Music]
1:42:30 Hey, Kyle. Yes. Got one more uh
1:42:34 generated birthday gift for you. Oh, no.
1:42:37 Put it on the uh No, this isn't anything
1:42:40 bad. You'll like this one. Uh I zero
1:42:43 shot a uh an image swapper for you.
1:42:46 That's on the post-it note. Oh, this is
1:42:49 this was done in Lovable.
1:42:54 [Laughter]
1:42:56 That's great. So what what can I
1:42:59 convert?
1:43:04 So first you can share the tab so we can
1:43:06 see it. Yeah. Yeah. Yeah.
1:43:10 Um and so this is a lovable prompt a
1:43:13 Google translate style site but for
1:43:16 image file format swapping. Oh that's
1:43:18 great. And it is it can take your AVIF
1:43:21 files and convert them to JPEG. You just
1:43:23 drag it in, drop it, and then it goes.
1:43:26 Have you Have you tested it? I have. If
1:43:29 If I drop this a AVIF in, and it doesn't
1:43:32 do it. File
1:43:35 uploaded. So, oh, there it is.
1:43:39 Nice. Convert from that to ping.
1:43:42 Convert. That's awesome.
1:43:49 And I mean the whole site, the whole
1:43:51 site, the help about the feedback, the
1:43:53 footer links, the privacy policy. I
1:43:56 mean, this didn't exist 15 minutes ago.
1:43:59 Loable did at all. That's amazing. I'll
1:44:02 drop it in irregular
1:44:04 so yeah. Yeah, drop it somewhere so that
1:44:06 I can find it. I'll I should I should
1:44:08 also bookmark it.
1:44:10 I mean, this is where we are, people.
1:44:13 Like, um, Jim Ross the other day, um,
1:44:17 got substituted in to do a keynote pre
1:44:22 presentation 20 minutes before he had to
1:44:24 give give the presentation. And rather
1:44:27 than panicking and writing a deck, you
1:44:29 know, and then giving the presentation,
1:44:31 he generated the deck for the audience.
1:44:36 like his part of his presentation was
1:44:38 generating the presentation that he then
1:44:41 gave. Like this is the world we live in
1:44:44 now. Just make [ __ ] You need [ __ ] you
1:44:47 make [ __ ] You ask for it, there it is.
1:44:49 [Laughter]
1:44:52 Boom. You know, a 100red million jobs
1:44:55 destroyed. How many jobs
1:44:57 generated? How many jobs generated? And
1:44:59 what kind of jobs?
1:45:02 I'm not sure we can ever trust that AI
1:45:04 results will not be 100% accurate. I
1:45:07 guess similar to human error. Yeah. Have
1:45:08 you worked with
1:45:13 humans? AI is
1:45:16 99.99999% accurate. Like it'll it's
1:45:19 going to get there. And then humans are
1:45:21 still going to be bitching. And you
1:45:22 know, we're consistently delivering, you
1:45:24 know, in the low 70% accuracy wise.
1:45:30 That's, you know, it's going to be like
1:45:31 uh self-driving cars, too, right? Where
1:45:34 I think self-driving cars at this point
1:45:36 are already safer than humans, but every
1:45:39 time, you know, one goes off the rails
1:45:41 and kills someone, everyone's like, you
1:45:44 know, they're evil. They're bad. No.
1:45:46 Well, considering the
1:45:50 competition. Oh, man. Joker. I makes I
1:45:53 make apps on the go as I need them.
1:45:55 That's super cool. Uh, is this thing
1:45:58 still going? Yeah, it's still going.
1:45:59 What has it found? Why does it
1:46:03 keep? It keeps
1:46:06 updating the same two
1:46:16 agents. Let me stop
1:46:18 this and I'm going to say you missed the
1:46:21 contact information. Let's throw that in
1:46:24 there. Regenerate response.
1:46:28 I bet I ran out of
1:46:30 tokens. I bet I ran I shouldn't have
1:46:32 stopped it. Damn
1:46:34 it. Anyway, um go play with Gen Spark.
1:46:38 Gen Spark's really good. Um I should
1:46:41 probably just download this, shouldn't
1:46:44 I? Is that what this
1:46:46 is? Copy URL to
1:46:53 clipboard. How do you download this?
1:46:57 Visualize. Fact check
1:47:00 content. Continue
1:47:17 searching. Yeah, it's not going to it's
1:47:20 not going to let me
1:47:21 go. All right. Well, there you have it.
1:47:26 There you have it, good
1:47:29 people. You can just right click on a
1:47:32 name and convert
1:47:39 it. Oh, file
1:47:42 formats.
1:47:49 Um, I don't know how to download
1:47:53 this.
1:48:10 [Music]
1:48:11 Anyone
1:48:28 Hey view.
1:48:34 Ah,
1:48:35 download. No.
1:48:41 Well, if anyone can figure out how to
1:48:45 export Gen Spark sheets, let me know.
1:48:49 Far right
1:48:51 icon
1:49:01 where anyone with the link I can copy
1:49:04 the URL to the clipboard.
1:49:09 anyone with the
1:49:14 link. I bet I could just copy. Yeah, I
1:49:18 can probably do it like
1:49:30 this. No, it's not working. All right.
1:49:34 Well,
1:49:39 Command
1:49:46 A. Nope.
1:50:18 quit while you're behind. Oh, I got it
1:50:20 thinking
1:50:22 again. I'll fact check the content and
1:50:25 add the contact information. All right,
1:50:27 I'm going to let that do its thing. You
1:50:29 don't need to watch me try to make this
1:50:31 thing work. Anyway, I would go play with
1:50:33 this thing. It looks pretty remarkable.
1:50:34 Um, not as quite as remarkable as my
1:50:36 hair. That's
1:50:41 fantastic. All right. I skipped out on
1:50:44 the B birthday because it made me feel
1:50:46 old. Yeah. No [ __ ] Yeah, it made you
1:50:48 feel old. How do you feel about
1:50:53 me? Uh, what happens if you upload? I
1:50:57 don't know. I don't know.
1:51:02 [Music]
1:51:07 Third from the left. That did not do it.
1:51:09 And of course, conversions. Kyle. F1 may
1:51:12 help bring up help.
1:51:15 F1. Uh, how do I do F1
1:51:20 function?
1:51:21 F1. Nope.
1:51:45 All right, I'm going to get out of
1:51:48 here. Thank you all. Hope you had a good
1:51:50 Friday night, date night, and
1:51:56 uh this weekend, do something nice for
1:52:00 Mother's Day. Um go make lots of
1:52:03 content. Just make stuff. make go make
1:52:06 make. Here's what I would encourage you
1:52:08 to do is um try a new tool this weekend.
1:52:13 Try something
1:52:14 new. And I like at this point I'm I am
1:52:18 no longer able to recommend like what's
1:52:21 good or not because there's just too
1:52:23 many of them. Like
1:52:25 just have chat GPT go research image
1:52:29 generation tools and just go play or
1:52:31 video generation or
1:52:33 music and just explore and then make
1:52:36 nice stuff. Put some joy in the world.
1:52:39 All right. And when people ask you how
1:52:41 did you do that? Tell them tell them how
1:52:43 you did
1:52:44 it. All
1:52:47 right. Beauty. All right. Um
1:52:52 Da
1:52:54 da. Beautiful. Thanks everybody. Have a
1:52:56 fantastic weekend. I will see you Monday
1:52:58 night. If you're on YouTube, like and
1:53:01 subscribe. Subscribe to the YouTube
1:53:03 channel. Okay. Thank you. Join the AI
1:53:05 salon. Beautiful. All right. Peace out,
1:53:08 everyone. Hope you have fun tonight.
1:53:12 Did