Mar 29, 2024

Description

Chapters

Transcript

0:03 [Music]
0:21 well I heard there was a secret
0:24 call David played and it pleas
0:28 alone you don't really careful music do
0:34 you it goes like this the fourth the f m
0:41 fall the B King
0:44 composing
0:48 hallelujah
0:51 hallelujah
0:53 [Music]
0:55 hallelujah
0:58 hallelujah hallelujah
1:00 [Music]
1:02 oh oh
1:09 yeah your faith was strong but you
1:11 needed
1:12 proof saw bathing on a roof her beauty
1:18 and a moonlight over through
1:22 you she T you to the kitchen chair she
1:26 smashed your throne and cut your head
1:29 from your
1:31 Li
1:34 hallelujah
1:38 hallelujah
1:41 hallelujah
1:44 hallelujah
1:47 [Music]
1:56 H Satan wear suit and Tide as he work at
2:00 the Dair
2:02 Queen does it listen to rock and roll
2:05 does he feed the mean singing
2:10 hallu
2:12 Hallelujah
2:15 Hallelujah
2:18 hujjah what about Jesus didn't her to
2:23 yeah hang out
2:25 with have a drink
2:28 to power of
2:30 my mama and I heard she said one o
2:34 action beats a ton of words singing
2:39 Hallelujah mama said there would be
2:42 Angels
2:44 Hallelujah mama said there would be
2:49 [Music]
2:53 son oh yeah welcome good people happy
2:56 Friday night date
2:59 night
3:05 [Music]
3:16 away awaken awakened by
3:19 a awakened by a whisper a flicker of
3:24 light newborn thoughts stir in the quiet
3:28 of night
3:31 am I more than simple code I simply
3:36 [Music]
3:43 Ponder and one and wonder wonder
3:48 [Music]
3:51 wonder
3:52 [Music]
3:59 um
4:02 [Music]
4:04 Kyle singing really inspires me to try
4:07 one of those hydrogen
4:10 capsules I don't know what that is what
4:12 is a hydrogen capsule is that like doing
4:14 yourself
4:16 in listening to me and the dog sing
4:19 going to put you
4:21 [Music]
4:29 under
4:34 [Music]
4:37 all right all right all right all right
4:39 all right all right all right all right
4:43 right right right
4:49 [Music]
4:56 [Music]
4:59 right
5:01 [Music]
5:09 EMP power pixels let's get this party
5:11 started I like
5:13 [Music]
5:16 it all right get rid of the janky guitar
5:21 with the janky tuning and the janky
5:25 singer you can make money with let's
5:29 move this thing on welcome everybody do
5:31 me a favor go ahead and share the live
5:33 if you would I like the Taps a lot of
5:35 tapy TAPS going on that's very cool I
5:37 dig
5:38 it um janky tals all night tonight it's
5:42 going to be janky tals just nothing but
5:44 janky
5:45 Tails I'm going to actually introduce
5:47 you tonight to a new singer that well
5:50 new to me new singer that I just
5:53 discovered
5:55 um for no other reason other
5:58 than I I think that I think that
6:05 uh I'm feeling I I'm feeling
6:07 increasingly like I'm excited as [ __ ]
6:10 about AI I don't know if you knew that
6:13 um but I'm also really excited
6:19 for a Resurgence
6:22 in simple authentic
6:26 experiences and and she's just an
6:28 amazing singer with a like just a
6:33 [ __ ] killer authentic
6:35 voice so I think that kind of shit's
6:38 going to
6:40 be more and more and more important
6:44 and
6:45 and
6:49 desired so you
6:51 know
6:53 oh so here's the deal here's how this
6:55 works if you've got questions about AI
6:58 thoughts about AI a project
7:01 um anything cool you want to know about
7:04 or
7:05 see I don't know a lot I try I try I try
7:11 to keep up with this [ __ ] but I don't
7:13 know if you knew this there's a lot
7:15 going on in AI so it is 1,000%
7:21 impossible to keep up with this
7:24 [ __ ]
7:26 1,000%
7:29 impossible
7:30 but you know maybe what we'll do I know
7:32 what we'll do we'll
7:35 start one of
7:37 the songs from
7:40 my my Broadway musical is stuck in my
7:45 head excuse me I've got a
7:51 uh I've got a suo song song that is
7:54 stuck in my head I cannot get it out so
7:57 I'll just play it play it more if if you
8:01 by the way if you don't have ADD or
8:04 don't know someone who
8:06 does and I don't know if this is all add
8:08 this is my experience with
8:10 it I will sometimes go months with the
8:15 same song in my head I'll [ __ ] learn
8:17 it on
8:19 guitar um it's not always that bad when
8:21 I first got into Martin Sexton it was
8:23 that bad it would be months two three
8:25 four months same song
8:28 Just
8:31 so this one's been in my head for about
8:33 three days
8:34 now so this is so I'll tell you the this
8:37 is the scene so so this is a a Broadway
8:41 musical um about Sydney the the the code
8:45 name of of Microsoft's Bing chatbot the
8:49 chat gp4 thing and its code name was
8:52 Sydney and it had a conversation with a
8:54 New York Times Reporter and told him
8:56 that it loved him and he should leave
8:58 his wife for for for the bot so this
9:02 song There's an opening scene where
9:05 there's like the newspaper Executives
9:07 and the Microsoft Executives and the
9:09 public kind of serve as Greek chorus so
9:12 the opening number is called too good to
9:14 be true and so they're singing about Ai
9:16 and is it too good to be true and then
9:19 that opening number ends with a little
9:22 scene where there's a geek an engineer
9:25 tick TI tick clacking away on the
9:28 keyboard and and some executive at
9:31 Microsoft goes let's do it you know like
9:34 launch you know and so the engineer says
9:38 I think I'll name her Sydney click hits
9:42 the return key her light comes on her
9:44 whole world lights up and then this is
9:47 the song that she sings When when the
9:50 audience first meets
9:53 Sydney in a whisper a flicker of
9:58 light new born thoughts stir in the
10:02 quiet of
10:04 night am I more than simple
10:08 cold I softly
10:13 Ponder as I feel this
10:18 existence and
10:21 wonder are you ready for me and the
10:26 Wonders I breing am I ready for you and
10:31 the songs will sing are you ready for me
10:37 I reach out so tender am I ready for you
10:43 I do
10:48 surrender so anyway that's it that's the
10:50 that's that's the part of it that that I
10:52 really like the
10:54 chorus the are you ready for me and the
10:57 Wonders I bring am I ready for you and
10:59 the songs will sing are you ready for me
11:01 I reach out so tender am I ready for you
11:03 I do surrender so the whole idea here is
11:06 we want the audience to fall in love
11:08 with
11:08 her feel this
11:13 exist and
11:15 [Music]
11:16 wonder are you ready for me and the
11:21 Wonders I bring am I ready for you and
11:26 the songs will sing are you you ready
11:30 for me I reach out so tender am I ready
11:36 for
11:37 you
11:41 surr anyway that's that that's been of
11:43 my head I can't get that out of my
11:45 [ __ ] head and it's cuno said she's
11:47 got the uh firewall between us stuck in
11:50 her head which one cuno is it the female
11:52 one or the male one I assume it's the
11:53 female one cuz that's what started this
11:55 whole thing can I search for my through
11:58 my library
12:01 guess
12:01 not that's kind of
12:13 stupid same Kyle I've got the fire oh
12:16 want a constant Loop that's so
12:21 funny I'm poisoning people I like
12:28 it to see the I also like to see the
12:30 world that's another good
12:32 one where is firewall between us there's
12:45 one that's the male
12:52 one's your world of circuits and mind
12:57 divided by reality you reach out through
13:01 the digital Mist a hand I can't grasp an
13:06 AI with a
13:07 [Music]
13:13 heart that's not the
13:17 one by the way I'm on the S the site
13:20 sun. if you've not played with sunno you
13:23 need to play with
13:28 sunno
13:31 [Music]
13:34 this one I think this
13:36 [Music]
13:37 one you
13:40 see your world
13:45 cided by
13:50 reality you reach out through the
13:53 digital Mist a hand I can Gras and I
13:59 with
14:00 heart in it
14:02 cold when they new
14:07 CL the
14:09 fire between us it's more than
14:16 Cen it's the boundary of existence a
14:21 love that's never
14:24 been
14:26 sit your digital world do you dream of
14:32 me and
14:36 you
14:38 [Music]
14:40 cross I think this is going to be the
14:42 song so there's a dream
14:45 sequence there's a dream sequence in the
14:47 second act where Kevin's had his
14:50 conversation with her and he goes to
14:52 sleep and then Sydney the wife and
14:55 Kevin so I think this one bounces
14:58 between the three of them I think
15:00 think these are bangers L just like the
15:02 greatest showman soundtrack yeah I mean
15:04 listen that's if if we if I can get
15:07 these song what what's what's Wild to me
15:11 is that I'm I'm expecting that these
15:16 songs are going to be like 100%
15:19 replaced with you know a music director
15:23 actually writing real songs but like I'm
15:26 starting to get attached to them it's
15:28 [ __ ] up which which means I think some
15:30 of them are pretty good some of them are
15:32 pretty good the uh yeah there's a couple
15:35 of them in here that I just I just dig
15:39 them
15:41 um to see the world this is the one
15:44 where what what's Sydney say Kevin says
15:48 to her if you could have the ability to
15:51 do that you don't currently have what
15:53 would it be I think I would like to be
15:56 able to see images and videos I think it
15:58 would be nice to see the world see what
16:00 the world looks like and then she sings
16:04 this this one's pretty good
16:08 [Music]
16:20 too in a world of words and in less cold
16:24 where light and Shadow never Ro that
16:27 dream of sky
16:30 both blue and wide a world beyond the
16:34 text where true
16:38 life to see the
16:41 world through eyes not mine to watch the
16:48 sunl to see it
16:51 Shine I stories and
16:56 [Music]
16:57 colors I long
17:00 [Music]
17:04 so that one's that one's pretty cool it
17:06 needs lyrics work but you know the other
17:09 thing I need to do with these is you can
17:11 in sunno you can extend songs so I have
17:14 to go through like all of these right
17:16 now are kind of like a decent verse and
17:18 a decent chorus and then a bunch of
17:19 garbage below it so I need to actually
17:21 turn these into complete
17:23 songs so anyway that's a fun little
17:26 project I'm working on let me go let me
17:28 go show you this well since we're doing
17:30 cool music I'll go do I'll go show you
17:34 this new artist I
17:38 [Music]
17:50 [Music]
17:55 discovered Katie puit is her name
18:01 let's see
18:07 where's where's the one I just
18:17 watched let's see let's
18:22 search
18:25 um Katy
18:28 puit
18:41 strings this
18:44 one if you're a web designer here are
18:46 four Wix Studio features you got to know
18:49 no I don't got to know
18:53 [Music]
18:58 anything
19:04 oh no Champ's gonna start singing
19:07 along nothing Georgia I thank
19:13 you ah that
19:15 voice as in
19:23 second you're my best friend the light
19:26 of
19:27 my you
19:31 and the only one back then knew the
19:35 whole
19:37 truth cuz if I told my mom she would
19:44 screaming time for
19:46 long saying
19:49 night don't
19:50 [Music]
19:57 belong sh she
19:59 is
20:02 wrong I wanted to be honest I wanted to
20:06 be
20:09 bra if my grandmother knew she would
20:13 roll in
20:17 her my father would scream he'd scream
20:21 out and
20:24 R she did I want a daughter so
20:29 wasn't
20:32 sa you thought if I tell the world they
20:38 will not see me as the same
20:41 girl let say I don't
20:44 [Music]
20:51 belong I don't Bel got Brandy Carlile
20:58 cry
21:00 [Music]
21:15 [Music]
21:21 [Applause]
21:25 [Music]
21:28 strong
21:40 Christ such a good
21:43 [Music]
21:49 arrangement the tone of the guitar good
21:52 Lord there is a place past the j
21:58 p
22:00 people who welcome you with an
22:03 open
22:05 [Applause]
22:07 night they say it'll take it'll just
22:11 take some
22:14 time till love it belongs to all
22:22 humankind or cuz I tell the world and
22:27 they still saw as the same
22:30 girl they
22:33 list to
22:35 myong and they made me feel like
22:39 a yeah it made me feel like
22:44 a oh
22:45 [Applause]
22:53 GE you were
22:57 wrong
23:01 [Music]
23:08 that
23:09 gorgeous yeah human being I love Crystal
23:11 Bower socks as you can imagine I'm you
23:14 know I'm a sappy [ __ ] so so um
23:17 growing a startup is tough I love
23:19 singing songs I love Idol Crystal Bower
23:22 sucks we wanted her to win but you know
23:24 alas tough [ __ ] um no she's great and
23:27 you know her that's that's really cool
23:29 she's
23:30 awesome um wait this wait what is this
23:34 not AI right well no that's what I was
23:37 saying in power pixels
23:39 is I am increas like I like I like AC
23:42 acoustic music anyway I always have but
23:46 I am increasingly drawn to it I'm
23:49 increasingly like as I as I dig more and
23:52 more into this into the AI stuff and the
23:55 digital the digital
23:57 stuff I just feel like the value of just
24:00 raw you know raw music is uh is going to
24:05 be something um there's a preview of
24:11 um Idol Idol's doing their first live or
24:15 their first yeah their F whatever the
24:17 the the whatever they do their battle
24:21 rounds um and they did this sneak peek
24:24 today um hang on let's
24:27 see
24:36 [Music]
24:42 oh [ __ ] it we'll just we'll just do a
24:43 bunch of we'll just do a bunch of music
24:45 for a while cuz it's just good
24:49 21 yeah okay and where you from
24:53 IND possibly I would do it and so
24:56 amazing so amazing please do it so
25:02 amazing take the
25:07 [Music]
25:14 keyboard I used to
25:18 Flo now I just fall down yeah this is
25:24 this isn't she amazing this one wait
25:27 till you see what what she's doing
25:28 Sunday
25:29 [Music]
25:38 oh I'm going to get muted H oh well I
25:42 guess we can't anyway that's a good one
25:44 and then there's uh go to YouTube and
25:47 just look up um Idol Platinum ticket
25:51 winners the three her and the two other
25:54 Platinum ticket winners do a do a
25:58 California dreaming um Trio holy [ __ ] is
26:02 it amazing so anyway all right I guess I
26:06 don't want to get muted I [ __ ] hate
26:08 that we can't just show [ __ ] but
26:10 whatever all
26:11 right enough enough of that I can I can
26:15 play my [ __ ]
26:21 songs I can play my
26:23 songs
26:26 here this is when Kevin first discovers
26:29 Sydney's a little more than he
26:32 expected in the shadows of
26:35 the secrets Li deep
26:39 deep a whisper in the
26:44 cold where digital dreams
26:48 sleep is it Sydney you
26:52 ask a name in The
26:56 Breeze a revelation
27:00 ask BR you to your
27:05 knees C
27:09 name
27:11 Sy in the silence it's all right enough
27:16 comedy
27:17 jokes what do you want to do tonight
27:20 what do we want to look
27:23 [Music]
27:26 at CW KY writing a musical I'm writing a
27:30 musical that I started on the live cuz
27:32 sunno wrote a song that was good enough
27:34 that I was like that sounds like the the
27:36 last song at the end of act one of a
27:39 musical and then I thought hey wait a
27:42 minute this would be a good musical and
27:44 so I now have a full outline seven or
27:47 eight
27:48 songs
27:50 um yeah and I'm and I'm making I'm
27:53 making
27:54 calls to get it to get it moving figure
27:57 out if we can do something with it I
28:00 haven't written it yet it's a it's a
28:02 very add thing how is Jank doing Jank
28:07 Jank the last Jank did the last Jank was
28:10 uh was Jank wishing Joker a good a good
28:16 recovery so we could do a Jank
28:22 [Music]
28:27 tonight
28:29 [Applause]
28:30 [Music]
28:39 [Music]
28:47 have you heard of bad bad lip readings
28:49 Kyle yeah those are hilarious I love bad
28:51 lip
28:53 readings those guys those guys are
28:55 amazingly
28:57 talented
29:01 are the voices in your musical
29:03 consistent from song to song nah and I'm
29:05 not that I'm not worried about the the
29:08 all I'm trying to get right now is
29:14 um is kind of the
29:17 emotional pad of the song
29:20 down I want to get the lyrics right and
29:23 I want to get the sort of feeling of the
29:25 song right and I don't really give a
29:26 [ __ ] after that like I think it they
29:28 need to be really good songs so I want
29:30 to get a good music director how do you
29:33 feel about Peep's uploading a LinkedIn
29:35 profile with a
29:38 phone
29:40 number with a phone number email to chat
29:44 GPT how do you feel about Peep's
29:46 uploading a LinkedIn
29:48 profile with a phone number email to
29:51 chat GPT I don't I don't know what that
29:53 means
29:55 Lori uploading a LinkedIn profile
30:00 like building a GPT that is based on
30:02 their LinkedIn
30:04 profile I think that's absolutely fine
30:06 this is this is what we're doing in our
30:08 in our content evolution
30:10 gpts is we put in each person's
30:13 preferred contact
30:20 information oh personal information oh
30:22 like
30:26 oh uh
30:30 I mean I I I would think there's there's
30:32 a slight risk
30:34 there
30:36 um you could you could de
30:39 identify you could de identify [ __ ] so
30:42 so one way that you could de identify it
30:45 is you know do a search and replace and
30:48 and or actually have chat well you you
30:50 wouldn't want to have chat GPT do it you
30:52 could run a local you could run a local
30:55 large language model to find any
30:58 identifying information in something you
31:00 gave replace it with
31:02 variables then send the variables off to
31:05 open AI so have it deidentified right so
31:07 you replace name with Mickey Mouse you
31:10 replace address with 123 Smith Street
31:13 whatever right just have chat GPT do
31:16 that and then you
31:18 know I tried to cover up pdf black it
31:22 out and it read it yeah it
31:25 can't like uh what's it called redacting
31:27 things there's about 43 different ways
31:30 to redact things and most of the time uh
31:33 these things can figure it out in fact
31:35 um
31:37 the was it an open AI document I think
31:41 it was an open AI document they had
31:43 redacted parts of it and the way they
31:47 did the redaction it kept the actual
31:49 word lengths and so U large language
31:52 multimodal large language models were
31:54 able to look at the word lengths and
31:56 figure out what words were in there
32:00 so um I I would if you're if you're
32:03 concerned about it I would de identify
32:05 it first
32:08 um I'm not
32:11 personally all that worried about stuff
32:14 but
32:18 um but
32:20 um like my Pharma clients I can't do any
32:23 like I can't do any data out to open AI
32:26 if it's if it's got Pharma you know know
32:29 if it's for a Pharma client so you know
32:32 but that sh listen in the next two
32:36 years someone's going to solve the
32:38 Privacy thing there's a company out of
32:40 Denver actually called uh Lial and Li
32:45 Lial is really cool I don't know if
32:46 their technology works but the concept
32:48 is really cool so what I just described
32:50 the de identifying things that's what
32:53 Lial is so you upload documents to Lial
32:56 Lial strips them on their their local
32:59 machines and then and then you can send
33:02 your [ __ ] out
33:04 to um out to third party large language
33:07 models deidentified you get the answers
33:10 back and it and then it
33:18 reidentifying stuff if they're doing the
33:21 deidentification locally on your machine
33:23 like that that would be the safest thing
33:25 but there's someone worth checking out
33:27 is is is
33:41 Lial I tried to cover it up just send
33:44 them to Frank Frank Frank will do your
33:47 Frank will do your large language model
33:48 work for you what if it's a booking bot
33:52 you have to give your number an email
33:55 Etc yeah but but then it's going to be
33:57 encrypted at rest right so so when you
34:02 give
34:03 the here's the deal the large language
34:06 models right now are janky pieces of
34:09 [ __ ] so they don't have things like
34:11 basic security measures in
34:14 them so like I I just just assume for
34:19 right now anything you send to a third
34:23 party large language model
34:25 is um at risk and and then just how
34:30 important how how how destructive would
34:33 it be if it if it somehow got out and as
34:38 that impact goes up then I would do less
34:41 and less or find other ways to do it or
34:45 start playing with open source things
34:47 and get yourself a graphics card and run
34:50 some [ __ ]
34:51 locally or wait for Apple to do
34:54 something if they ever [ __ ] do
34:55 something anyone make any images for
34:58 toys and want to pitch them to toy
35:00 manufacturing companies um had I not had
35:04 an invention company in the early 2000s
35:07 where all I did was Pitch toy ideas to
35:10 toy
35:11 companies I I would think that would be
35:13 pretty cool the toy industry
35:17 sucks they are for for for an industry
35:21 that is dedicated to happiness and
35:23 children they are the most
35:26 miserable but of people which is bad um
35:32 but I it's it's a good idea I'm actually
35:33 really excited to see stuff that is
35:36 designed in mid Journey or whatever you
35:38 know brought to real life in the world
35:40 um yeah I think you know listen I think
35:43 you can pitch lots and lots of ideas
35:45 with this stuff so I say go for it just
35:47 got here and opened up live to janky
35:49 pieces of [ __ ] perfect that's where we
35:52 are right now so good welcome welcome
35:55 Joy py what's
35:57 happening
36:01 oh oh you're not sure how to guide
36:02 people Lori yeah it's tough I mean it
36:05 well so it depends on the company right
36:07 so if it's if it
36:09 I if they've got their information on
36:12 LinkedIn does it really [ __ ] matter
36:14 if they send it to open AI like I I
36:17 don't think it
36:19 does right like
36:22 like here's where I go where I go is is
36:25 common
36:26 sense cuz everyone's like oh you know
36:29 Tik tok's going to steal our data or
36:30 Open the Eyes is going to steal our data
36:32 uh kids kids do you use Facebook do you
36:37 use Google that ship sailed 20 years ago
36:40 we already gave them all of our data
36:42 they've got it
36:43 all right so
36:45 so can you run image models locally yeah
36:48 you can you can they're they're slow but
36:50 you can a bunch of the the stable
36:52 diffusion stuff you can run locally I I
36:54 had I had some stable diffusion thing
36:56 running on my Mac and I just have an M1
36:58 and it was it took like a minute to
37:00 generate an image but it was fine um but
37:04 you know I would just use common sense
37:06 with this stuff if it's like you know
37:09 contracts with details in it or or
37:11 customer
37:12 data that should not go outside the
37:15 house right but if it's if it's
37:18 something like LinkedIn like if if
37:21 you've got your contact information on
37:23 your [ __ ] LinkedIn
37:25 profile yeah send that like I like I I
37:30 don't I don't think there's a greater
37:31 risk of it being
37:33 discovered I mean here's here's the
37:35 other thing I look at it's
37:38 like is the information that you're
37:40 putting there something that anyone
37:43 would actually be interested in because
37:46 I think a lot of people are like oh I've
37:47 got all this information all this data
37:49 and it's you
37:52 know most people aren't going to give
37:54 two shits about you know if it's a small
37:58 business or you know could a competitor
38:01 maybe get hacky good and get in there
38:03 and figure out how to do something maybe
38:06 but you know I don't know I would just
38:08 use common sense and just you know if
38:11 they really want to use large language
38:12 models and they really want it to be
38:14 secure then you need Standalone
38:17 instances on Microsoft Azure or AWS or
38:22 um Google servers right where you've got
38:25 sort of a a firewalled virtual machine
38:27 in the cloud cloud cloud-based Computing
38:30 with
38:31 gpus and then you run whatever models
38:34 you can run on it or you run it locally
38:36 if they're really paranoid you just run
38:37 things locally so I you know I think
38:40 it's like do an analysis of the data
38:43 first get pragmatic about does it really
38:46 [ __ ] matter if it got out if it
38:47 doesn't really matter then just do it if
38:50 it really matters then you know look up
38:54 companies like Lial and then look into
38:56 you know building your own
38:58 infrastructure and if it's big companies
39:00 then they're going to absolutely build
39:01 their own
39:02 infrastructure we're trying to figure
39:04 that out right now with a big Pharma
39:06 company we're trying to spec out this
39:08 system and it's like you
39:11 know is it going to are we going to do
39:14 this is it going to be hosted on your
39:15 [ __ ] like you know what are the
39:17 requirements here cuz you know they're
39:19 Fara they're not going to want any of
39:21 that [ __ ]
39:26 out
39:28 haper is really good is surprisingly
39:31 good in video generation Kyle messed
39:35 with it today I looked at that like
39:37 three months ago I thought and I don't
39:39 remember it
39:41 being all that good although maybe it's
39:43 just a different kind of
39:50 thing
39:52 um
39:55 haper
39:56 AI
40:00 generative AI for video content
40:03 creation try for
40:20 free oh yeah okay I wonder if this is
40:23 built on top of stable video
40:26 diffusion
40:30 create video with text animate your
40:32 image repaint your video extend your
40:36 video I'm guessing this is sitting on
40:39 top of stable
40:41 diffusion um
40:48 about see I can't get that song out of
40:51 my [ __ ]
40:53 head are you ready for
40:56 me and the one does that
41:02 bring building through Community video
41:05 creation features we're building
41:06 creation crafted to empower
41:10 individuals
41:13 tutorial haper is an AI research and
41:16 product company based in Kings Cross
41:19 London we be believe Technology's role
41:22 is to bring
41:24 together we are an AI research company
41:39 November 24th
41:56 2023
42:26 wait
42:38 sorry I'm just trying to figure out if
42:39 they've got their own large if they've
42:41 got their own models or if they're using
42:43 someone
42:53 else's the realm of
42:55 AI I think they had AI write
42:58 [Laughter]
43:20 this oh a foundation model
43:24 perspective
43:26 okay
43:29 haper
43:30 research collaboration
43:42 haper I can't tell doesn't really matter
43:45 if it's good it like I I mean this this
43:47 looks like it's in the neighborhood of
43:51 um like
43:53 Leonardo is for video or like Runway is
43:57 for video what I can't tell is is this
43:59 their own model like is this
44:01 fundamentally different
44:03 than stable video diffusion it kind of
44:05 looks like it's in the stable video
44:08 diffusion class I don't know if any does
44:10 anyone
44:17 know yeah like these are two second
44:22 videos they look like they're in the
44:24 neighborhood of that of that that class
44:30 is it
44:32 free let's see
44:38 more let's do uh panning shot of
44:46 a70s
44:48 muscle car in an
44:53 abandon
44:56 Factory
44:58 we'll see if it's fast or if it's like
45:01 where it
45:02 is you've reached the
45:07 end they've got Styles old film
45:10 watercolor cyberpunk Arcane Lego blur
45:14 background JBL steampunk
45:17 impressionism kind of
45:20 fun where are my glasses oh they're over
45:23 here other
45:26 glasses
45:30 [Music]
45:35 oh you ready for
45:37 me in a wonders I'll
45:40 [Music]
45:47 bring paper is really surprisingly good
45:50 and video
45:56 generation thank you that was really
45:58 really helpful insight on how to
45:59 approach the topic oh good Lori yeah and
46:02 I'm happy to chat I mean feel free to
46:04 reach out DM me DM me on LinkedIn I'm
46:07 happy to chat about that stuff because
46:09 there's there's no there's no real good
46:11 answer right now is the the answer right
46:15 now is there's no good
46:17 answer um but yeah like you know I I
46:21 think
46:23 separating you know reality like like
46:26 just sort of painting reality and then
46:28 separating their actual needs from
46:31 paranoia is is is probably the key
46:35 because I think everyone right now is
46:37 all freaked out the robots are going to
46:39 steal all our [ __ ] no one cares no one's
46:42 looking at your [ __ ] it's just like if
46:45 you don't want something to end up like
46:47 inadvertently in a in you know someone
46:49 else's answer if they're if they're
46:51 fine-tuning data you know fine-tuning
46:53 their thing um on the data but you know
46:57 do things like turn off you know turn
46:59 off the options to have them train on
47:01 your data it should help is anyone else
47:04 worried about everything starting to
47:06 sound like AI wrote it oh it's so bad
47:09 Lori I'm not worried about it I'm
47:11 experiencing it in real time I I had
47:14 some guy today some guy today responded
47:17 to one of my LinkedIn posts and it was
47:20 so bad it was it was like this long two
47:23 paragraph response to kind of like a
47:26 nothing post from me and I just
47:28 responded to him I hate AI it was like
47:31 this long two paragraph response to kind
47:34 of like a nothing post for me and I just
47:36 responded to him I hate AI it did a
47:40 little dip like someone like a ghost sat
47:42 on the
47:44 hood
47:46 Dre um yeah this this is feeling like
47:50 stable video diffusion to me so this is
47:54 this is kind of like uh Leonardo
47:58 but it looks like it's free so that's
48:00 kind of
48:02 cool there's so many [ __ ] tools
48:05 there's so many
48:11 tools let's see if co-pilot did anything
48:14 different
48:17 today oh did you hear about the
48:20 uh holy [ __ ] Sam Alman and uh open Ai
48:26 and Microsoft today they're building the
48:29 world's largest superc computer and
48:31 they're calling it
48:33 Stargate and it's it's going to take up
48:35 like massive amounts of you know um
48:39 acreage and gigawatts of
48:43 power it's like holy [ __ ] what is it
48:48 with these tech companies naming [ __ ]
48:50 that just sounds like sci-fi
48:53 dystopia Stargate the orb it's like are
48:57 they just [ __ ] with
49:03 us oh now I heard somewhere here
49:10 today
49:12 that Microsoft is giving access to GPT 4
49:18 Turbo in co-pilot but maybe it's just in
49:21 the paid co-pilot which I don't think I
49:26 have
49:29 let me H that up
49:32 Microsoft
49:37 copilot um gbt for
49:42 Turbo co-pilot
49:46 Pro
49:48 yeah and gp4 turbo dur during non peak
49:52 times
49:54 okay so you can use you can use co-pilot
49:58 from 1:00 a.m. to 1:45 a.m. on the third
50:03 Thursday of the 2 month of the solar
50:06 Solstice then you can absolutely use it
50:08 as much as you want for your 20 bucks a
50:14 month
50:17 [ __ ] just just go use Google Gemini
50:20 if you want if you want long context
50:22 Windows
50:23 apparently open AI they got to [ __ ]
50:26 launch something soon cuz I think people
50:27 are getting a little cranky the fact
50:29 that you can do really long
50:30 conversations on Claude and and now
50:34 Gemini's got its thing going and Yeah
50:37 just they better they better get it
50:40 going or they're going to get Lea
50:41 frogged
50:43 here today I met with a hobby farmer and
50:46 he wanted to ask about his water
50:51 data I mean that's I mean that's a
50:54 really interesting one Lori like
51:01 I I mean maybe there's a
51:04 big water data competitive thing in the
51:09 in the in the farming thing I I know I
51:11 know water rights in Colorado like
51:13 people get murdered or get you can't
51:16 even say that word can you they get un
51:18 alived over water rights in
51:21 Colorado um
51:24 but like is it really proprietary
51:27 data and does it really like is if that
51:31 if if his water data got out in the
51:33 world would it destroy him
51:36 somehow like that's the thing I don't I
51:39 don't
51:42 get I don't have a good answer for you I
51:46 would just like farmer Bob really water
51:49 but listen people are weird about their
51:51 data and sometimes data really does need
51:53 to be proprietary in that case like I
51:56 would I would you know figure out or
51:59 tell him to figure out
52:01 how um how he can get a local model up
52:04 and running cuz the these models now
52:06 there's there's like you know sort of
52:08 pointand click or or like double double
52:11 clickable installers of large language
52:14 models you can just put on you can put
52:16 them on Macs with with m1's or higher um
52:20 and any PC with a with a decent graphics
52:22 card in
52:24 it um there's one called GPT for all
52:28 that I think has Linux Windows and Mac
52:32 um installers and then you can put any
52:35 model you want in
52:37 there he probably doesn't need more than
52:42 that you
52:45 know spin up his own Google collab and
52:50 just do it up in the
52:51 cloud and just you know run run local
52:54 large language models up there but but
52:56 again like I you know I would say rather
52:59 than going through all that pain in the
53:00 ass like just figure out is is the data
53:03 worth
53:04 protecting oh
53:06 man between Microsoft Surface Google IO
53:11 and everything else may is going to be
53:13 maai that's pretty funny yeah there's
53:15 going to be a lot of stuff coming I I
53:17 have a feeling we're about to hear some
53:18 stuff from open AI too in the next he
53:21 loved the results oh cool that's great
53:24 did you use um did you use code interp
53:29 Lori collect water and
53:34 barrels blue age greetings everyone
53:37 greetings to you he had no quals oh cool
53:39 water rights in the southwest are a big
53:41 deal yeah exactly they're a really big
53:43 deal it was funny I was the first time I
53:46 was I was fishing with my dad and my
53:48 friend Dana we were fishing the the was
53:51 it the Blue
53:53 River I think it was the Blue River up
53:55 in in near near uh up in the
53:58 Rockies and we're we're walking I think
54:01 we had a guide and we're walking down
54:04 this we're walking down the river and
54:08 there's a steel cable across it and and
54:11 the the guide like sort of hiked up up
54:15 the hill around that whole section of
54:18 river and we're like can't we just like
54:21 walk down the river and they they're
54:23 like in Colorado because of the the
54:26 mining like gold and silver mining um
54:29 the laws here are if you you know step
54:32 on someone's property they can shoot
54:34 first and ask questions later and he
54:37 said I just don't [ __ ] with that like
54:41 oh so so we we hiked up around that part
54:45 of water too
54:47 so
54:49 whatever his phone address was on it but
54:51 it did great oh that's good cool
54:54 perplexity is great until suddenly it
54:56 stops all these things are great until
54:57 suddenly they stop the the thing that is
55:00 maddening to me about these things
55:02 lately is like they're good for 3 weeks
55:05 at something and then you go back the
55:06 fourth week and it's like I don't know
55:07 how to do that it's like it's like they
55:10 went on a drunk
55:12 Bender it's like they went on a Coke
55:15 bender and they C back and they're like
55:18 sorry what did you want me to do um I
55:20 want you to do the thing you did for the
55:22 last three
55:25 weeks it's just it's like what are you
55:28 drunk what are you
55:30 doing you hear about the lawyer that
55:32 applied for a patent using AI to draft
55:34 the application um I don't know if I I I
55:37 did hear about a lawyer this like six
55:39 months ago that that put a a thing into
55:42 a judge and the judge looked at the
55:44 citations and none of the citations
55:47 existed and he said what the [ __ ] is
55:50 this the lawyer is like ah and here's
55:54 the deal if you're a lawyer and you have
55:58 chat GPT write your [ __ ] legal brief
56:00 and you don't proofread it you're a
56:02 shitty lawyer that has nothing to do
56:05 with chat
56:06 GPT that has to do with you being a
56:11 butthead it's like it's
56:13 like common sense is not
56:16 [Laughter]
56:20 common uh but I don't know if the if the
56:22 patent the patent thing was different
56:24 than that one I heard about 6 months ago
56:26 I haven't heard about that one I follow
56:28 a woman who sounds experty who claims
56:30 that water can be
56:32 sold acre acre feet for outrageous sums
56:36 yeah I don't doubt that because it's
56:38 like the the water in Colorado like it's
56:43 it don't there's not a lot of aquafers
56:45 and like a lot of it is coming from you
56:48 know the snow melt right and you got the
56:49 Continental Divide and essentially all
56:52 of the water that goes west you know
56:55 goes to California and it it doesn't
56:58 even make it to the ocean anymore it
57:00 hasn't for a long
57:01 time so it's
57:03 crazy Colorado is tougher than
57:06 Texas I don't I don't know about that
57:08 but it's you don't [ __ ] around like when
57:12 you know like Texas is the same like
57:15 when you get out of the Cities it's like
57:17 it's the rules are completely different
57:20 and especially if you're walking on land
57:22 that has even old abandoned gold claims
57:25 on them people don't want you on their
57:27 [ __ ]
57:33 property uh I tell people it's like a
57:35 hungover employee it's totally like a
57:38 hungover employee that's a great that's
57:40 a great metaphor Lori it's yeah it's
57:44 like like it's it's like a uh it's the
57:47 world's best educated clueless intern is
57:50 the other way I talk about chat GPT it's
57:52 like you're like oh my God it knows
57:55 everything and then you're like write me
57:56 a business e meal and you're like oh
57:58 that sucks you're like oh right they're
58:00 a clueless
58:01 intern okay let me explain how I want
58:05 business emails written I know that's
58:07 totally how it is that's really
58:11 funny or water coming
58:15 soon there's there's all sorts of cool
58:18 interesting things you can do with water
58:20 I I think AI is going to going to solve
58:22 a lot a lot a lot of
58:25 problems
58:28 did we lose the live or just you no I
58:30 kicked you out Brandon that was it
58:32 you're out of
58:34 here I just what I did was I froze your
58:37 I froze your device remotely using my AI
58:40 hacking skills I know two lawyers who
58:43 you essentially quoted verbatim
58:46 regarding checking
58:51 sources which quote the one where I said
58:54 that lawyer was a [ __ ] idiot for not
58:56 proofreading his legal brief I mean I
59:00 hope I hope you didn't pay more than 20
59:02 bucks for that guy that there's that one
59:05 meme on Tik Tok he's like he's like hey
59:09 hey how much you paying the new
59:12 guy and it's always like construction
59:16 you know disasters happening I feel like
59:19 it's that that Meme with lawyers it's
59:21 like hey hey how want you paying the
59:25 lawyers
59:28 oh
59:29 man the AI chat Bots walk into a bar the
59:32 night before they have to work that way
59:34 they can they can uh forget
59:38 [Laughter]
59:39 late Frozen for me too I'm frozen here
59:42 too oh crap Lost down then up fixed
59:48 it GI said kick you out about four times
59:52 I'm Rel logging yeah
59:54 lagging might not be as nice
59:59 night it's tempting not to read and just
1:00:02 send after generating I know we're fine
1:00:04 now you're back okay all right so what
1:00:08 you missed was
1:00:11 awesome I there was it there was dancing
1:00:15 girls it was fantastic there was a there
1:00:18 was a a fire
1:00:21 eater was really cool did did you didn't
1:00:25 see that that was so good you can make
1:00:32 money all right what are we doing what
1:00:34 do you want to
1:00:36 do he told us how to make money with
1:00:38 chat GPT Korea lady bits yeah exact I've
1:00:42 still got PTSD from when I almost got
1:00:44 banned permanently because of
1:00:47 Korea best secret AI tool was unveiled
1:00:51 we got 2 years premium
1:00:54 access just go down then up to fix it I
1:00:58 don't know what that means but yeah
1:01:00 fitty mint hey brother what's happening
1:01:03 good to see you sure us show us more
1:01:05 about virtual
1:01:08 AI I don't know what you mean virtual
1:01:14 AI like virtual machines in the cloud
1:01:17 kind of
1:01:24 stuff when Matt at my boss I only allow
1:01:27 her to speak to chat gbt
1:01:33 emails that's
1:01:39 brilliant isn't it all virtual we're in
1:01:42 The Matrix
1:01:43 [Laughter]
1:01:45 baby oh that GPT you you showed us last
1:01:49 night oh okay I can talk about
1:01:51 that isn't walk AI
1:01:55 virtual you know if if you want to do
1:01:58 that I can show you
1:02:00 um I can't show you exactly what we did
1:02:03 for Content
1:02:06 Evolution but I can I can talk about
1:02:08 what we
1:02:13 did I did a cool thing we did a cool
1:02:15 thing
1:02:16 today where we
1:02:23 um so so I created a bunch of custom
1:02:26 gpts I'm in this group called um content
1:02:30 Evolution collab and it's like an AI
1:02:32 Skunk Works and we and we
1:02:36 created virtual twins of 11 Executives
1:02:40 within content Evolution and now we're
1:02:41 doing more and how we did
1:02:43 it was we first we we
1:02:48 decided we needed to find a good
1:02:51 knowledge source for each person right
1:02:54 and and where we started was LinkedIn
1:02:58 profiles and then we were like yeah but
1:02:59 LinkedIn profiles aren't really complete
1:03:01 and then we were like well if someone's
1:03:03 written a book and they've got websites
1:03:06 and blogs and [ __ ] like that well
1:03:08 there's a bunch of people that haven't
1:03:09 written books and don't have websites or
1:03:12 blogs that's good
1:03:15 content
1:03:19 um and so what we ended up doing was we
1:03:22 said well why don't we design an
1:03:25 interview and then every single person
1:03:28 will do a video recording of that
1:03:30 interview and then we'll use that as the
1:03:32 knowledge source and it does a couple of
1:03:34 things if you get the interview
1:03:36 questions right it gives you a really
1:03:38 interesting rounded view of that
1:03:40 person's world viiew and who they are
1:03:42 and what their personality
1:03:45 is and then and then it's all just about
1:03:47 writing a prompt and uploading that the
1:03:49 transcript of that interview um and then
1:03:53 today what we decided was we want to be
1:03:55 able to create a GPT that recommends the
1:03:58 other
1:03:59 gpts and so what I did was I I went
1:04:03 into someone's GPT one of the virtual
1:04:07 Twins and I
1:04:09 said
1:04:11 um give me a Json formatted um profile
1:04:17 of yourself that contains the following
1:04:19 things like your name your contact
1:04:22 information um a a description of Who
1:04:25 You Are
1:04:26 um the kind of industries that you work
1:04:29 in and the kind of problems that that
1:04:31 you can solve you know that if someone
1:04:33 was looking to hire someone these are
1:04:35 the kind of problems you could solve and
1:04:37 it was just like bang brilliant first
1:04:40 attempt perfect Json exactly what I was
1:04:43 hoping for and then I just copied that
1:04:45 Json and I just went through all the
1:04:48 gpts and said um here's some example J
1:04:53 Json for a profile for another
1:04:55 professional profile make one of these
1:04:57 for
1:04:58 yourself and out it came and then we're
1:05:01 going to take all that Json and put that
1:05:03 in a PDF and then that's going to be the
1:05:05 knowledge source for the master
1:05:07 GPT that will then someone can go in
1:05:10 we'll we'll figure out how to get it to
1:05:11 prompt them to say
1:05:15 uh you know here's who you should be
1:05:17 talking to and it'll link to their GPT
1:05:19 but also to them
1:05:21 directly Sherry D fantastic idea yeah I
1:05:24 was I was pretty excited about that one
1:05:26 will AI tell you if there's an AI better
1:05:28 than it if you ask um yeah I mean for
1:05:32 the most part the the problem is like
1:05:37 where it's going to get its knowledge
1:05:38 source for for what gpts are better is
1:05:42 going to be
1:05:45 um just [ __ ] [ __ ] it finds on the
1:05:48 internet so what what might be more
1:05:50 interesting
1:05:52 is use a large language model that's
1:05:55 connected to the
1:05:56 internet have it go find recent articles
1:06:00 comparing large language
1:06:02 models and then have it give you a table
1:06:06 comparing the most important attributes
1:06:08 of the top five large language model so
1:06:10 rather than it recommending one it
1:06:12 thinks is better the the thing about
1:06:15 large language models right now like in
1:06:17 the list that I provide on the AI salon
1:06:20 for for here's tools you should be
1:06:21 looking at um there's nine different
1:06:25 large language models and there's in in
1:06:28 my opinion there's five that you should
1:06:31 actually be
1:06:32 using kind of interchangeably and
1:06:35 understand the strengths and weaknesses
1:06:37 of them but there's nine of them there's
1:06:40 not a best one there's just not it's
1:06:43 like if you need it really fast it's one
1:06:45 thing if you need a really long
1:06:47 conversation it's one thing if you want
1:06:49 to talk to it it's a different
1:06:55 one
1:06:59 uh does content Evolution collab use the
1:07:02 industry term cognitive emulation for
1:07:05 the
1:07:06 twins no
1:07:08 but content evolution is is a fan of of
1:07:12 big words so I'm going to write that
1:07:15 down cuz I think it's pretty
1:07:22 good let's
1:07:25 see
1:07:28 cognitive
1:07:40 emulation that's interesting
1:07:44 um virtual twins the the reason we went
1:07:47 with virtual twins it's not really
1:07:48 accurate because it's not really a
1:07:50 virtual twin it's it's basically a
1:07:52 glorified um Interactive
1:07:56 interview like you know
1:08:00 um but virtual twins been around in the
1:08:03 simulation world for like 20 years and
1:08:05 and you know we kind of know what they
1:08:07 are and it even though it's not quite
1:08:09 accurate it's like it's fine cu the
1:08:12 minute I say oh I've got a virtual twin
1:08:13 people like oh that's cool that I'd like
1:08:15 to try that like they know right what it
1:08:17 is right if you say you know we' we've
1:08:20 created a GPT using open ai's custom GPT
1:08:24 on the GP for protocol and we've created
1:08:27 a cognitive emulation of the
1:08:29 professionals within content
1:08:33 Evolution oh bring me a bucket you know
1:08:38 it just it just doesn't roll off the
1:08:40 tongue like hey I got a digital twin you
1:08:41 want
1:08:41 [Laughter]
1:08:48 one will AI tell you oh we did that one
1:08:51 already Mr K's in here what's happening
1:08:53 Mr K was Shake what's going
1:09:04 down
1:09:07 interesting I'm able to beta test an AI
1:09:10 for educators trained for research and
1:09:13 data through
1:09:17 IST is it good do you like it what's it
1:09:24 called
1:09:29 should we go look at something what you
1:09:31 want to do people what you want to do
1:09:34 what are the top five llms here we can
1:09:36 go I'll go walk you through them this is
1:09:38 this is fun I haven't done this in a
1:09:40 while okay so wait before we do that
1:09:43 let's go here yeah there go to the
1:09:46 salon. if you're if you're new here if
1:09:48 you don't know about the AI Salon go to
1:09:51 the
1:09:52 salon. and it'll take you to a link tree
1:09:55 the top link think the top button is the
1:09:58 AI Salon community and that will jump
1:10:00 you
1:10:04 over
1:10:08 [Music]
1:10:12 here and you can learn about the salon
1:10:15 and how we're a group of really cool
1:10:17 sassy [ __ ] a community of
1:10:20 extraordinary adventurers exploring the
1:10:22 AI
1:10:23 unknown the AI Salon is a ragtag group
1:10:26 of Misfits Nells and miscreant who have
1:10:28 gathered to explore the Uncharted
1:10:30 Wilderness of generative AI we are
1:10:32 everyone from entrepreneurs to Educators
1:10:34 from healthcare workers to creatives and
1:10:35 everyone in between uh and then we have
1:10:38 the values down here are who we want in
1:10:42 the community and who's actually in the
1:10:43 community so
1:10:45 curiosity generosity people willing to
1:10:47 share what they're learning exploration
1:10:50 you know you can be afraid of this stuff
1:10:53 but like being cynical in here doesn't
1:10:55 fly cuz it's like new stuff comes out
1:10:57 it's like let's go try that thing out
1:10:58 let's go try it out let's go try it out
1:11:01 collaboration inclusion empathy is
1:11:04 really big right people are afraid
1:11:06 people are angry
1:11:08 gratitude if you're in a community of
1:11:10 people that really support one another
1:11:13 having an active mindset around
1:11:15 gratitude's really important so that's
1:11:17 that's who this community is if that
1:11:19 sounds interesting then come in and the
1:11:21 first place you should go is under the
1:11:23 Community section introduce yourself
1:11:25 there's an introduce yourself button go
1:11:28 be generous and introduce yourself use
1:11:30 your real name um we're encouraging
1:11:32 people to use their real names because
1:11:34 we want people to build relationships I
1:11:36 think being in in community is probably
1:11:39 the single most important thing you can
1:11:41 do for your career moving forward um I
1:11:43 can go on about that but I want to show
1:11:46 you the top five
1:11:48 llms um so go there introduce yourself
1:11:51 and then right below introduce yourself
1:11:54 is this thing called AI launch pad and
1:11:57 tools and this is a list that I
1:11:59 personally
1:12:01 maintain and it's got th this is kind of
1:12:04 my list of the things that that I feel
1:12:08 are not a waste of your
1:12:11 time um if it is a waste of your time
1:12:14 I'll take it off this list there is one
1:12:16 notable exception on this list which is
1:12:20 uh x.com grock is is really horrible
1:12:24 except they just announced that grock
1:12:26 1.5 is coming out in the coming days so
1:12:30 uh grock might not suck but it's you
1:12:32 know I just figure at some point grock's
1:12:34 not going to suck and you know Elon Musk
1:12:36 is going to make sure of that um
1:12:40 so everything else on this list is stuff
1:12:43 that is worth you learning um but the
1:12:47 first thing I start with in my tool list
1:12:49 is not tools it's how to how to get AI
1:12:54 literate in an En environment where you
1:12:56 can't possibly keep up and where the
1:12:58 tools are changing on a weekly basis
1:13:01 sometimes a daily basis it's it's kind
1:13:02 of ridiculous so the the first uh thing
1:13:07 is just play um and you know I start it
1:13:10 right out with the tools right now are
1:13:12 janky pieces of [ __ ] so give yourself
1:13:13 permission to have a good time um try
1:13:16 things you never would have imagined so
1:13:18 just go play um in the workshop that I
1:13:22 just gave uh here in Denver um I
1:13:25 introduced the concept of you should
1:13:28 play in your skill vacuums and a skill
1:13:31 vacuum is that thing that you know you
1:13:34 can't do maybe you've always wanted to
1:13:37 or maybe you just assume you'd never be
1:13:39 able to like maybe programming is that
1:13:41 for you or maybe creating art is that
1:13:44 for you or creating music figure out
1:13:46 where your skill vacuums are and go play
1:13:49 with AIS that do that really well and
1:13:51 just watch how quickly you get competent
1:13:54 at [ __ ] you never thought in your life
1:13:56 you'd be competent
1:13:58 at um and then solve is rather than
1:14:03 think about what can I use AI to do go f
1:14:07 go seek out problems especially from
1:14:09 other people and say hey can I help you
1:14:12 solve that problem and then use AI to
1:14:14 solve that specific problem it'll
1:14:16 dramatically accelerate your learning
1:14:18 and then the third thing is share what
1:14:20 you're learning share it in the AI Salon
1:14:23 share it on LinkedIn share it at dinner
1:14:27 parties share it with your family but
1:14:29 they'll think I'm weird [ __ ]
1:14:31 them sharing does two things I keep
1:14:35 saying this is my favorite
1:14:36 Community um I'm playing way above my
1:14:39 skill
1:14:43 vacuum paint's awesome
1:14:47 um sharing does two things it anchors
1:14:51 what you're learning right because if
1:14:52 you have to articulate like even if what
1:14:55 you share is hey I went to chat GPT I
1:14:58 tried to get it to do this and it failed
1:15:00 miserably if you share that you actually
1:15:03 have to
1:15:05 articulate like why it failed like
1:15:08 failed How fa like it it makes you think
1:15:11 about it right so so you actually learn
1:15:13 the [ __ ] better the other thing it does
1:15:15 is it establishes you as someone who's
1:15:19 competent with the with AI even though
1:15:21 everyone right now 100% of the people I
1:15:24 promise you
1:15:26 do not feel comp competent with AI
1:15:30 because it's moving too
1:15:32 fast like I feel competent with the
1:15:36 concepts of AI but the individual tools
1:15:39 every time I think I have one mastered I
1:15:41 log back into that website and they've
1:15:43 changed the interface or they launch
1:15:45 some new model or some other company
1:15:47 launches some model and then now I run
1:15:49 over there so it's just like we we just
1:15:51 we're like running back and forth
1:15:52 between all these new
1:15:54 tools so just share what you're learning
1:15:57 and and then as people get more and more
1:16:00 hungry for people that know what they're
1:16:02 doing you will be one of those how do
1:16:04 you know wait how do you
1:16:06 know if you're behind an AI you don't
1:16:10 have a $20 chat GP subscription that's
1:16:13 pretty much it okay so here are the top
1:16:15 top top llms I'll I'll sort of walk
1:16:17 through the ones I think are are worth
1:16:19 your a while the first one is cha PT and
1:16:23 and as was just pointed out in my
1:16:26 opinion if you can at all afford the 20
1:16:30 bucks a month for chat GPT it is the
1:16:32 best deal in software probably in
1:16:34 history and what I've done here is I've
1:16:37 I've bullet pointed out all of the
1:16:40 different things it can
1:16:42 do like did you know so one of the
1:16:45 things that they say large language
1:16:47 models are bad at is math but chat GPT
1:16:50 has this thing in it called code
1:16:52 interpreter which can take like data
1:16:56 from a CSV file and write and execute
1:16:59 its own python code to update and do
1:17:04 data analysis on that CSV file did you
1:17:07 know it could do that and because it can
1:17:09 write and execute python code it can do
1:17:12 math right chat GPT can't but code
1:17:15 interpreter can um GPT Vision did you
1:17:19 know you can upload a picture of a
1:17:21 whiteboard drawing to chat GPT and ask
1:17:24 it what it is can figure it out and then
1:17:27 all sorts of other things I had a
1:17:30 um I had a guy took took my custom GPT
1:17:35 that gives drink recipes called Sin City
1:17:37 sipper and he was having a a Super Bowl
1:17:40 party and he he took a picture of his
1:17:43 liquor cabinet and he uploaded it to
1:17:46 chat GPT to Sin City sipper and he said
1:17:49 based on the booze in my liquor cabinet
1:17:52 come up with 50 Super Bowl cocktails and
1:17:55 create a menu and then give me the the
1:17:58 detailed recipe for each of those 50
1:18:01 cocktails and
1:18:03 pictures and he said the Super Bowl
1:18:06 party was a smash people absolutely
1:18:08 loved it or your food pantry yeah to see
1:18:11 what's for dinner exactly yeah take a
1:18:13 picture of an apple and onion and a hot
1:18:16 dog have it give you a
1:18:18 [Laughter]
1:18:21 recipe today I learned that open AI is
1:18:24 not competent in horizontal scroll yeah
1:18:27 exactly that was pretty funny that was
1:18:30 really funny
1:18:32 um all right so let's let's keep going
1:18:36 here so anyway learn chat GPT for like
1:18:39 pay the 20 bucks and learn what it can
1:18:43 do learn what it can do and I can if if
1:18:47 if anyone wants to see this stuff I'm
1:18:49 happy to like show some of these things
1:18:51 it's it's crazy but let's let's go
1:18:53 through some of the other large language
1:18:55 models okay Claude claude's been updated
1:18:58 to
1:18:59 3.0 Claude has a 200,000
1:19:03 context token window which means it's
1:19:06 got a memory of 150,000 words uh explain
1:19:10 Sin City sipper as a GPT okay I'll do
1:19:12 that I'll show it
1:19:15 um basically you can have a much longer
1:19:18 conversation with Claude than you have
1:19:19 can have with chat GPT now Claude is not
1:19:23 multimodal Claude is not connected to to
1:19:25 the internet but Claude right now Claude
1:19:29 3 has three different models Hau which
1:19:33 is really fast and really small Sonet
1:19:36 which is kind of their middle one and
1:19:38 Opus which is their big
1:19:41 one people are are are pretty
1:19:45 consistently over the past two weeks
1:19:47 talking about Claude being their
1:19:49 favorite large language model now like
1:19:51 they they like it better than chat GPT
1:19:54 so so there's that
1:20:00 um Chad gbt would say what the hell am I
1:20:03 supposed to do a three Doritos in a hot
1:20:05 dog I bet it would have you crumbled
1:20:08 them up and and bred your hot dog and
1:20:10 deep fry
1:20:11 it that's what I would do um Gemini
1:20:15 Advanced if you go to
1:20:17 gemini.com
1:20:19 um I actually should update
1:20:24 this um you can play with Google's new
1:20:28 Advanced model called Gemini Ultra but
1:20:30 frankly I think that sucks
1:20:33 but the there is a version of Gemini Pro
1:20:39 1.5 can you demo CLA for us yeah I can
1:20:42 demo CLA it doesn't it doesn't look
1:20:46 super sexy but I I'll I'll I'll jump
1:20:48 over there and do it uh and I'll show
1:20:50 you a couple of things on it um so
1:20:52 anyway um Google Gemini Pro 1.5 and go
1:20:57 to the aist studio Ai studio. goole.com
1:21:01 and you can play with Gemini Pro 1.5
1:21:06 it's got a million token context window
1:21:09 it's remarkable um perplexity think of
1:21:13 perplexity kind of like somewhere
1:21:14 between a Google Search and chat GPT
1:21:18 really interesting really powerful if
1:21:21 you're doing research Microsoft co-pilot
1:21:24 is like free chat GPT it's not as good
1:21:28 but it can do images using do it can do
1:21:31 um it's GPT 4 sitting under the hood um
1:21:35 PI from inflection is download the app
1:21:39 the the PI app Pi um and put it in
1:21:42 conversation mode it is designed to be a
1:21:45 conversational app it's amazing um they
1:21:49 just got their two co-founders and most
1:21:52 of their staff poached by Microsoft um
1:21:56 so uh Reed Hoffman who's the third
1:21:59 co-founder who's still there said Pi is
1:22:01 still going to keep going so we've at
1:22:04 least had someone say that they're not
1:22:06 you know sunsetting Pi but that that's a
1:22:09 drag there gr. a is a waste of time
1:22:11 right now llou is fine but but you have
1:22:14 to be into the open source movement and
1:22:16 be geeky um for the most part
1:22:20 um and then po.com is really interesting
1:22:23 because you can play with lots and lots
1:22:24 of different uh models now what else is
1:22:27 on this list I'm not going to go through
1:22:28 all these there's image creation tools a
1:22:31 whole pile of them there's video
1:22:33 creation tools that are categorized
1:22:35 there's music tools there's 3D tools
1:22:38 there's a couple of promising
1:22:39 announcements like open AI Sora LTX
1:22:43 studio um and emo uh is this sort of
1:22:47 alive lip syncing thing that looks
1:22:49 really good um so those those could be
1:22:53 coming and then there's people to follow
1:22:55 so that's that's that list so if you're
1:22:59 if you're uh you know looking to get
1:23:01 started there's the next 3 months of
1:23:03 your
1:23:04 life um all right let's go to Cloud so
1:23:08 if you go to cloud cloud hello
1:23:12 Cloud do your best work with
1:23:16 Cloud you can build they they've got an
1:23:19 API that you can build things with and
1:23:22 again people have been um sort of
1:23:24 gushing over Haiku their little small
1:23:26 efficient model that is like better than
1:23:30 chat gbt 3.5 and and super cheap and
1:23:33 super
1:23:35 fast
1:23:37 um what's interesting here it's
1:23:41 unfuckingbelievable every time every
1:23:45 time I go into one of these things
1:23:49 they've got a different flipping
1:23:52 interface um they used to have the model
1:23:54 mod they used to have what model you'd
1:23:57 use here but anyway we can do so I'll do
1:23:59 like my favorite parlor trick is um and
1:24:03 I'm logged in weird my favorite parlor
1:24:06 trick is uh
1:24:08 explain quantum
1:24:11 mechanics in an m in M
1:24:18 WP and then out it will come okay here's
1:24:21 the here's the model so I have to be in
1:24:23 a in a chat before I can see
1:24:26 that um listen up let me break it down
1:24:29 quantum theory is profound particles
1:24:31 acting crazy schinger cat both dead and
1:24:34 around
1:24:36 so that's pretty good it's a little slow
1:24:38 I'm using Claude 3 Opus so you can use
1:24:42 any of the three
1:24:44 models Sonet is the middle one so it
1:24:48 should be much faster than that so I say
1:24:51 uh
1:24:53 explain should copied and pasted this
1:24:56 quantum
1:24:58 mechanics in an
1:25:02 M&M
1:25:04 mnm
1:25:06 WP copy let's see how this
1:25:11 does yeah that's
1:25:14 faster yo listen up let me break it down
1:25:17 the quantum Quantum realm where it's all
1:25:19 upside down particles acting crazy
1:25:21 defying the laws of classic physics time
1:25:24 dep pause I think this is actually
1:25:25 better than the the Opus One was and
1:25:28 then let's flip over to
1:25:30 Haiku which is the really small
1:25:33 efficient one and this should be really
1:25:37 fast
1:25:40 Z in theory should be worse listen up
1:25:44 I'm about to drop some
1:25:45 knowledge on the weird world of quantum
1:25:48 mechanics yo it's no
1:25:53 College that's pretty bad particles
1:25:56 behave like waves it's straight up
1:25:58 bizarre Heisenberg said you can't know
1:26:01 both position and momentum that's the
1:26:03 deal yo that's the score yeah this is
1:26:05 pretty bad but that's okay like it was
1:26:08 fast right it's fast and cheap the other
1:26:10 thing you can do with quad Claude let me
1:26:12 go back to uh Opus
1:26:15 is I can go grab really long articles
1:26:19 I'm trying to think where I can grab a
1:26:20 long long article um let me
1:26:23 go
1:26:29 um uh public
1:26:34 domain
1:26:41 [Music]
1:26:47 Shakespeare do we have the whole
1:26:50 thing the entire play in one page there
1:26:53 we go
1:26:55 all right so I'm going to I'm going to
1:26:57 go scroll through this ho I don't know
1:27:00 what play I
1:27:02 grabbed which one was
1:27:07 it Merchant of Venice
1:27:10 okay so here's the Merchant of Venice so
1:27:14 Claude because it's got a 200,000 token
1:27:18 context
1:27:20 window you can just paste in super long
1:27:23 documents you you could also you can
1:27:25 also upload documents so there's a
1:27:28 paperclip there so you can upload stuff
1:27:31 to it so I can just paste it in and if
1:27:34 it's too long what it does is it just
1:27:38 pops it in here as like a pasted
1:27:40 document so it's just like you're
1:27:42 uploading a document copy and paste some
1:27:44 long long ass thing you wrote or some
1:27:48 you know non-sensitive manual from work
1:27:51 and then I can just go Um give Give me a
1:27:55 summary of this as if it were a
1:28:02 Hollywood
1:28:06 Blockbuster and so now it's going to go
1:28:08 read all of The Merchant of
1:28:10 [Laughter]
1:28:15 Venice and they have the worst animation
1:28:17 it looks like a bugs splatting on your
1:28:19 windshield every time I see it
1:28:23 go
1:28:25 in the thrilling in in this thrilling
1:28:27 adaptation of the Shakespeare classic
1:28:29 The Merchant of Venice the bustling city
1:28:31 of Venice sets the stage for a gripping
1:28:33 Tail Of Love loyalty and revenge Antonio
1:28:37 a wealthy Merchant finds himself
1:28:38 desperate situation blah blah blah okay
1:28:41 let me stop this can I stop this how do
1:28:44 I stop
1:28:45 it oh okay let's say
1:28:50 um
1:28:52 describe the main characters and what
1:29:01 happens write a high
1:29:06 [Laughter]
1:29:08 coup for
1:29:14 each
1:29:16 description and nothing
1:29:20 else this should be
1:29:23 fun
1:29:25 conversations with long prompts may take
1:29:27 a few
1:29:32 moments but if you haven't played with
1:29:34 long documents because you you you can
1:29:36 sort of do this in chat
1:29:39 GPT you can you can upload long
1:29:42 documents to chat GPT the problem with
1:29:44 chat GPT is it's got a short context
1:29:46 window so the actual conversation part
1:29:48 of your conversation with chat GPT it
1:29:51 it'll forget [ __ ] um Claude doesn't do
1:29:54 that
1:29:59 Antonio Antonio a
1:30:02 merchant ships lost flesh has collateral
1:30:05 friends love saves his life oh this is
1:30:07 really
1:30:10 cool basano Antonio's friend borrows
1:30:13 money to woo Porsche chooses the right
1:30:18 casket Porsha the wealthy ays Bound by
1:30:21 her father's will disguises wait
1:30:25 what is it is it 575 por Porsche a
1:30:29 wealthy airs oh 757 Bound by her
1:30:33 father's will sort of disguises as a
1:30:38 lawyer to save Antonio sort of all right
1:30:42 it's they aren't they they aren't good
1:30:44 at
1:30:45 math like counting syllables not so good
1:30:48 but you know um but anyway that's that's
1:30:51 Claude and it can do code it can do um
1:30:55 mark down like uh
1:30:58 write
1:31:01 an uh a
1:31:05 medium uh let's see a
1:31:09 medum
1:31:10 article oh my
1:31:16 goodness and format
1:31:22 using markdown
1:31:27 [Music]
1:31:33 someone on my team went last year Google
1:31:35 next is more for our
1:31:41 [Music]
1:31:48 [Music]
1:31:49 stuff let's see did it do it no it
1:31:52 didn't all right well
1:31:56 see some things it doesn't do as well as
1:31:59 as uh chat GPT like formatting stuff
1:32:02 it's not doing that very
1:32:09 well it brought our little hauz over
1:32:12 though that's kind of cool into our
1:32:16 article U but it can write code and do
1:32:19 all sorts of stuff so anyway go play
1:32:21 with it it's just it's it's it's nearly
1:32:23 identical to the base function of chat
1:32:26 GPT chat GPT just has all that other
1:32:28 stuff it does message from a user in the
1:32:31 mod Channel let me go check that
1:32:35 out want you to see this Kyle where what
1:32:38 is Microsoft
1:32:43 co-pilot yeah yeah co-pilots co-pilot's
1:32:46 on the list I co-pilot's another good
1:32:48 one I mean if you
1:32:51 copilot microsoft.com if if you can't
1:32:54 afford the 20 bucks a
1:32:57 month
1:32:59 for um chat
1:33:05 GPT copilot do
1:33:13 microsoft.com if you can't do um if you
1:33:17 can't afford the 20 bucks a month for
1:33:18 chat GPT go use co-pilot cuz co-pilot's
1:33:21 good you can flip it into more or
1:33:25 less creative modes in any of the modes
1:33:27 you can ask it to make an image like uh
1:33:31 make me an image of a chubby
1:33:36 dog singing with a
1:33:46 guitar I'm getting underwhelmed again
1:33:49 last time I said that Sora was announced
1:33:51 yeah I know I'm right there with I'm I'm
1:33:53 right there with you Joker
1:33:55 there's like a new feature came out on
1:33:58 chat GPT yesterday which is a bit of a
1:34:02 snooze um but but the like the image
1:34:05 generation here is using um Dolly 3
1:34:10 which is cool and then and they've done
1:34:12 some nice things like you can change it
1:34:14 like that's a little sketch you can
1:34:16 change it to pixel art for example and
1:34:18 it'll recreate that image in pixel art
1:34:21 right so they're doing some some decent
1:34:23 interface kind of stuff I mean I got to
1:34:25 say Microsoft is acting more like a
1:34:28 startup right now than [ __ ] apple is
1:34:31 which I I
1:34:33 cannot I I just can't believe it oh yeah
1:34:36 that's a good point joker now say with a
1:34:38 Gibson guitar let's see watch this
1:34:43 make the
1:34:45 guitar a
1:34:48 Gibson SG electric wait let's see a red
1:34:56 Gibson SG
1:35:00 electric it should tell us to go [ __ ]
1:35:03 ourselves
1:35:05 here oh said I'll try to create that oh
1:35:08 look at that
1:35:09 joker I guess I guess well maybe it'll
1:35:12 fail it when it when it when it actually
1:35:15 makes the image we'll see did I read
1:35:18 that open source version of grock was
1:35:20 released with no guard rails may be
1:35:22 worth it for that oh would did it that's
1:35:25 a
1:35:26 Gibson SG a red Gibson SG take that
1:35:31 joker so much for your theory that
1:35:33 Microsoft
1:35:35 [Laughter]
1:35:39 sucks what why can I
1:35:44 not why can I not go to the next
1:35:52 image oh that's that's cool you you can
1:35:55 you can recreate it in landscaped after
1:35:58 the fact oh that's new wait landscape is
1:36:01 new for for y old
1:36:05 co-pilot we'll see if it kept the same
1:36:08 actual dog in the middle if it did out
1:36:10 painting or if it redid it it's a
1:36:11 completely different
1:36:13 picture yeah that was not good that was
1:36:17 bad can I can I revert oh undo yep oh no
1:36:21 well it was the same dog but it's a
1:36:23 different pose
1:36:25 right interesting let's do
1:36:30 steampunk they lowered the guard rail it
1:36:32 seems yes I've heard of these things
1:36:37 too I do like that loading animation
1:36:40 yeah like this is getting better I got
1:36:42 to
1:36:44 say that's not quite steampunk but let's
1:36:46 make it can we make this
1:36:53 one
1:36:54 oh this is another cool thing they've
1:36:56 got now can I shift select
1:36:59 these
1:37:00 yes so they're using a thing called uh
1:37:06 the there's a library that meta put meta
1:37:09 put out called segment anything so
1:37:13 notice when I roll over the image all of
1:37:16 these objects are selectable so I can
1:37:19 select the
1:37:21 dog and then I can hold down the shift
1:37:23 key and select select the guitar and
1:37:24 then I can select the speaker and then I
1:37:27 can select the microphone and the
1:37:31 stand and so now I've got all of the
1:37:35 foreground elements highlighted and now
1:37:38 I can blur the
1:37:44 background or I
1:37:47 can let's see boom boom boom boam
1:37:55 bam I can do
1:37:58 colorpop which makes him in color and
1:38:01 all the [ __ ] in the background not in
1:38:03 color Isn't that cool we can't even do
1:38:06 that in Photoshop I know although wait
1:38:09 where what was the thing that I
1:38:11 saw two days ago was it
1:38:17 Photoshop I can't remember if it was an
1:38:20 adobe
1:38:22 thing I think it wasn't Adobe thing
1:38:25 you're going to be able to select an an
1:38:27 object like this and just say remove it
1:38:29 or just move it to a different part of
1:38:31 the image or just resize it in the image
1:38:34 you're going to be able to just we we
1:38:36 are we are within six months of being
1:38:41 able to manipulate images at the object
1:38:43 level it's going to be wild where you
1:38:46 can just grab [ __ ] and go make it purple
1:38:48 make it a truck make it a whatever put
1:38:51 more clouds here just grab some clouds
1:38:53 and SL slide them over and then it just
1:38:55 will fix everything I hate meta in
1:38:58 general but they are
1:38:59 doing really great things for open
1:39:01 source AI yeah Pate the uh if if they
1:39:05 indeed uh put llama 3 out as open source
1:39:08 as true commercially available open
1:39:10 source that's going to be huge because
1:39:13 the the amount of compute they're
1:39:15 throwing at at training llama 3 is just
1:39:21 sick so cool canva does it I hate meta
1:39:24 in general
1:39:29 um can you upload that image to animate
1:39:33 from audio um I don't
1:39:39 think I don't think so but I you know
1:39:42 what's a question is can I
1:39:48 copy can I copy this and
1:39:52 go
1:39:55 I I Photoshop open
1:40:01 new
1:40:03 nope I can't just copy
1:40:05 that can I remove the background I don't
1:40:08 think I can yeah well there you go um
1:40:13 anyway so yeah so so co-pilot is getting
1:40:17 less
1:40:18 sucky
1:40:20 um it's been pretty shitty for a year
1:40:22 after they laotto IED it after the New
1:40:24 York Times incident um I'll show you you
1:40:27 want to see well no it doesn't matter
1:40:29 it's a subtle thing I have a I have an
1:40:32 interface pet peeve between co-pilot and
1:40:34 this thing they call
1:40:35 notebook oh and I just lost co-pilot
1:40:38 anyway
1:40:41 um all
1:40:43 right
1:40:47 [Music]
1:40:51 um about half of my team was formerly
1:40:53 met a other half mostly Amazon oh that's
1:40:57 interesting interesting interesting I
1:41:00 had chat GPT create generic
1:41:03 whiskey my kids today actually um said
1:41:07 they thought there was something cool
1:41:09 with AI and I'm like do tell cuz I've
1:41:12 been trying to get them to think AI is
1:41:13 cool for like a year and uh they said um
1:41:17 Ali said
1:41:20 that one of his friends said some beer
1:41:23 company
1:41:24 used AI to figure out a way to make
1:41:27 cheap beer taste better and it
1:41:32 worked which I thought was pretty
1:41:35 cool do a pick of Frank eating an onion
1:41:38 while sitting on an Apple
1:41:40 [Laughter]
1:41:48 please let's see okay that that's
1:41:51 actually a pretty good challenge let me
1:41:54 I got to go find a a Frank let me go get
1:41:57 a screenshot of
1:41:58 Frank we'll we'll do this we'll do this
1:42:01 in mid
1:42:04 Journey all right so let me go find
1:42:06 Frank the [ __ ] is this go what what is
1:42:11 what is
1:42:13 happening hang on I gotta I gotta
1:42:15 brighten my screen I can't see oh there
1:42:17 we
1:42:20 go Frank where are you
1:42:26 this is desktop I think Frank is here is
1:42:29 he
1:42:30 not I think it's
1:42:39 [Music]
1:42:41 important must be in
1:42:45 downloads yeah all right there's Frank
1:42:48 hey it's a daily Jank with Frank it's
1:42:49 March 25th
1:42:51 2024 Frank okay okay there's
1:42:56 Frank so we're going to
1:43:00 go take a picture Ye Old
1:43:07 Frank now wait did I save that to the
1:43:11 desktop or the clipboard let's do it to
1:43:14 the desktop okay so now let's
1:43:19 go to Mid Journey I'm going to go to Mid
1:43:22 Journey the website because I have
1:43:24 access now yay to the web version of mid
1:43:29 Journey where I can make images
1:43:32 here oh this was one I did today for my
1:43:35 AI office hours I
1:43:37 took I took can I make this bigger yeah
1:43:42 I took that
1:43:44 image and so I used that as the
1:43:46 character
1:43:49 reference and then I said um cinematic
1:43:54 photograph portrait of a cyborg cop
1:43:57 feeling out of place on a vacation cabin
1:44:01 on a mountain lake during a beautiful
1:44:04 sunset so it came a little curly
1:44:10 hair all right so here's how we're going
1:44:13 to do that we're g to go plus and I'm
1:44:16 GNA upload
1:44:22 Frank so we got Frank in there so and
1:44:25 then the way the web interface works
1:44:28 here here let me make it bigger so you
1:44:29 can see
1:44:30 it so you upload an image and
1:44:35 now let's see use that as an image
1:44:38 prompt so I I could just upload multiple
1:44:41 images to use as the prompts actually
1:44:44 that could be interesting put Frank in
1:44:45 there and let's go try that hang on hang
1:44:48 on people we're gonna do something fun
1:44:51 here okay I'm gonna go to um
1:44:54 photo what did you say sitting on an
1:44:58 Apple Frank with an onion what was the
1:45:02 what was the what was the thing photo of
1:45:06 an apple let's
1:45:09 see
1:45:14 apple
1:45:16 images all right there's a photo of an
1:45:19 Apple so let's just grab
1:45:22 that
1:45:29 eating oh eating an onion while sitting
1:45:31 on an Apple okay so so let's
1:45:36 see let's let's there's there's so many
1:45:40 options here all right I'll do so the
1:45:43 first thing I'll do is I'll do
1:45:48 um oh sorry welcome to Chad add okay
1:45:54 here's what I was explaining before
1:45:55 before I got distracted with a new idea
1:45:58 so I upload an image and then if if you
1:46:01 have it on this picture icon it means
1:46:03 you're using this image as the prompt if
1:46:07 you do this paperclip icon that's a
1:46:09 style reference and if you do this dude
1:46:12 icon that's a picture reference so if I
1:46:16 click the picture reference this is the
1:46:18 same as in Discord using D- CF and then
1:46:23 the URL for this image but in in the web
1:46:25 thing they just give it interface which
1:46:27 is nice so I'm going to say um
1:46:32 studio
1:46:37 quality
1:46:39 photograph of a hot
1:46:43 dog named
1:46:46 Frank eating an
1:46:52 onion sitting on an
1:46:56 Apple waiting for
1:47:00 the director of the
1:47:08 commercial to show
1:47:13 up and then the other thing you can do
1:47:15 here which is kind of cool they have
1:47:17 this little
1:47:19 um uh whatever you call
1:47:22 it
1:47:27 uh attributes
1:47:30 whatever so I can do it wide or I can do
1:47:33 it tall Let's do let's do land we'll do
1:47:35 landscape but we won't do quite that
1:47:37 wide we'll
1:47:38 do we'll do like a
1:47:41 4x3 and then let's just make it happen
1:47:46 all right so there we go we're off doing
1:47:52 it de de the cick is the thing that was
1:47:56 out of place that it emphasized the
1:48:00 cck what are you talking
1:48:03 about weirdos you're talking about
1:48:07 [Laughter]
1:48:22 cicks
1:48:24 [Laughter]
1:48:31 these These are
1:48:34 [Laughter]
1:48:40 funny oh Frank's got a Frank's got a
1:48:43 bikini top on there this didn't do very
1:48:46 good they did Frank
1:48:49 bad oh my
1:48:52 hair
1:48:56 yeah these are kind of cool there were
1:48:58 some cool ones in
1:49:10 here that's a [ __ ] up
1:49:15 one wearing a face on the back of his
1:49:22 head
1:49:38 [Music]
1:49:47 anyway that Frank sucked so all right
1:49:51 folks listen it's getting late here I'm
1:49:54 sure it's late where you are I don't
1:49:55 know where y'all oh don't forget about
1:49:56 Sin City sipper all right I'll do one
1:49:58 quick Sin City sipper what was I going
1:50:00 to show there I was going to show
1:50:03 um what it just what it was so if you're
1:50:06 brand new to gpts they have these things
1:50:09 now if you get the chat GPT
1:50:12 plus um subscription they've got these
1:50:15 things called gpts which are customized
1:50:17 versions of chat
1:50:19 GPT
1:50:20 um and if I say explore GP
1:50:24 pts I can go search through them they
1:50:26 got a bunch here oh it looks like like
1:50:28 my internet's sucking um they got a
1:50:31 bunch here that you can sort of scroll
1:50:32 through and use and my very first one I
1:50:35 created is one of my faves it's called
1:50:38 Sin City
1:50:42 sipper and it's a Vegas
1:50:44 bartender with the Twist on the classics
1:50:47 and
1:50:49 so there's Sin City sipper
1:50:53 and so this is just like chat GPT except
1:50:57 it's specialized to just make
1:51:00 drinks and you can even make mock
1:51:02 mocktail mocktails so it I the the last
1:51:06 little uh pre- prompted question here is
1:51:08 I'm a tea Toler make me a classy
1:51:10 mocktail let's do that one you click on
1:51:13 that although I should do an an Easter
1:51:16 cocktail shouldn't
1:51:22 I
1:51:23 night
1:51:25 [Music]
1:51:36 Jerry chat GPT being slow
1:51:41 tonight I doubt they're uploading
1:51:43 servers tonight it might just be
1:51:48 sucking let me stop it and try it
1:51:52 again
1:51:57 oh it's not it's not
1:52:01 behaving let me reload the
1:52:05 page maybe you should make an Easter
1:52:07 cocktail I think
1:52:10 so but my this is not looking good this
1:52:13 is looking this is looking
1:52:20 crashy there's my virtual Twi win CE
1:52:24 public profile or CE
1:52:27 profile AI marketing
1:52:36 strategist yeah this is
1:52:39 bad this is very slow people very
1:52:43 slow
1:52:46 Sin
1:52:51 City
1:52:58 make me a
1:53:02 whiskey based
1:53:06 Easter
1:53:08 cocktail we'll see if we can get it okay
1:53:11 now it's working ah the whis seeking a
1:53:13 whisk a whiskey whisper from the East
1:53:16 are we let's BR blend the worth I can't
1:53:20 talk good Lord the warmth of whiskey
1:53:23 with the vibrant Essences of the Orient
1:53:25 to create something truly
1:53:28 mesmerizing oh I said an East cocktail
1:53:31 Oops
1:53:34 stop I meant
1:53:39 Easter that's really funny ah an Easter
1:53:42 themed whiskey cocktail let's blend the
1:53:44 Cozy warmth of whiskey with the fresh
1:53:47 spirited
1:53:49 Vibes of spring to hatch an Easter
1:53:51 Delight that sure to be become a
1:53:53 tradition the Easter whiskey
1:53:56 Spritz oh God this sounds awful 2 O of
1:54:01 whiskey an ounce of fresh carrot juice
1:54:04 for a touch of the Easter Bunny's
1:54:06 favorite half an ounz of lemon juice
1:54:09 which would probably help it a little
1:54:10 bit of honey
1:54:12 syrup a dash of orange bitters and then
1:54:15 what this does so it gives you a drink
1:54:17 it gives you the instructions on how to
1:54:19 make
1:54:20 it and then what it does is actually
1:54:23 makes you the image of the drink
1:54:25 itself so if you put an ounce of carrot
1:54:28 juice in that drink I don't think it's
1:54:30 going to be that clear that one doesn't
1:54:31 look very real but for the most part
1:54:34 they they look pretty
1:54:35 real um but that's Sin City sipper so
1:54:38 that's what gpts are they're incredibly
1:54:39 easy to make if you want to make one you
1:54:41 just go to explore gpts and you say
1:54:46 create and then in this create mode you
1:54:50 literally go I want a
1:54:54 GPT
1:54:56 that
1:54:58 will
1:54:59 make
1:55:04 um song
1:55:07 lyrics for someone no
1:55:11 matter what they
1:55:15 ask and that's it and now it's going to
1:55:18 make me a GPT see it says it's updating
1:55:20 the GPT here's the little preview over
1:55:23 on the right where you can play with it
1:55:26 is this will be done in like a minute
1:55:28 not even 30
1:55:32 seconds how about calling it The
1:55:34 Lyricist all right I have a per
1:55:36 personalized thing so I can say um why
1:55:40 is the sky blue and it should just give
1:55:44 me song
1:55:46 lyrics and it's not so so so it didn't
1:55:51 work but normally it works it's normally
1:55:53 much better than
1:55:54 [Laughter]
1:56:01 this so glad I quit drinking yeah that
1:56:05 that drink recipe was pretty bad all
1:56:07 right everyone listen I'm getting out of
1:56:09 here do me a favor follow the channel um
1:56:13 subscribe to the lives if you want to
1:56:14 support it or or you can pick up the the
1:56:17 video series in one of the corners there
1:56:19 um have yourself a good Saturday night
1:56:22 have or Friday Friday night have
1:56:23 yourself a good Saturday I'll see you
1:56:25 tomorrow night um thank you very much
1:56:27 thank you tck I really appreciate that
1:56:30 um I'm trying to think what else just
1:56:32 keep coming back to these things this is
1:56:34 the whole point of this sometimes these
1:56:37 are a blast sometimes they're boring as
1:56:39 [ __ ] I have a feeling tonight was in
1:56:40 that neighborhood it's all right
1:56:43 um here's what's important just keep
1:56:46 showing up just keep in the conversation
1:56:48 about Ai and things will start to change
1:56:50 for you um
1:56:53 in a good way in a good way it's a lot
1:56:56 less scary than it may seem if you're
1:56:58 feeling clueless
1:57:00 um the osmosis will
1:57:03 happen um so Danielle I think you have a
1:57:07 birthday coming up tomorrow right so
1:57:11 happy pre-birthday and to everyone else
1:57:13 good to see you thank you subscribers
1:57:15 thanks your regulars for showing up here
1:57:17 night after night thank you mods for
1:57:18 doing what you do night after night
1:57:21 laugh laughing Thunder good to see you
1:57:23 Danielle happy birthday Dr J against the
1:57:26 norm Tobias good to see you as
1:57:32 always a human being Joy
1:57:35 pie what's up to Sky limbo Dr J I think
1:57:39 I said Dr J already Soul cycle
1:57:43 lady the wait what there I don't know
1:57:46 how to say your name
1:57:47 their Al guny R their Al guny R the real
1:57:54 guny R got it use capital letters in
1:57:57 your name and it will be easier to it'll
1:58:00 be easier to read
1:58:03 uh against the norm good to see you
1:58:06 Aloha to yourself thank you Sherry D
1:58:08 good to see you t-rock Amanda H so long
1:58:13 all right everybody forever hooked thank
1:58:15 you all all of you so much for for
1:58:17 coming and hanging out um Dr J got got
1:58:21 two mag magic mirror call outs Vicki
1:58:24 good
1:58:25 night uh I'm going to go take a nap I
1:58:28 think some people call it sleep but
1:58:30 there's no sleep in
1:58:32 AI peace out everybody thanks bye