Apr 14, 2024 - How AI is Transforming Art and Content Creation

Description

Chapters

Transcript

0:01 [Music]
0:10 [Music]
0:23 take your way my way seem to be see
0:30 [Music]
0:37 she wants to
0:41 know that's
0:44 it mama don't understand it she wants to
0:48 know where I've been I have to be so
0:51 kind of natur FO to want to pass this
0:55 way again you know that I feel it
1:00 [Music]
1:05 good evening
1:06 everybody happy Sunday night
1:13 [Music]
1:46 [Applause]
1:48 [Music]
1:57 whoo whoo whoo What
2:01 [Music]
2:14 telescope looking at the world tonight
2:17 through blue
2:19 telescope I wish I may wish I might not
2:23 see what I
2:24 see she metal on sheets of eyes
2:32 looking through this blue
2:36 telescope down the moon struck Ro
2:40 [Music]
2:44 tonight good evening
2:46 everybody my name is Kyle this is the AI
2:48 learning lab and dog singing
2:51 [Music]
2:53 seminar where do you do full-time
2:56 classes so doc uh Saturday this coming
2:59 Saturday today I'm doing a Play Shop uh
3:02 it's a 4-Hour um Hands-On workshop with
3:06 John Leland and I'll put the uh the link
3:10 to it I'll I'll show it here in a in a
3:12 bit so this Saturday I'm doing one and
3:15 if you're part of the AI Salon you get a
3:18 discount and then if you're like what's
3:19 the AI Salon I'll tell you that too cuz
3:22 that's another place that you can learn
3:24 is hang out with people that know what's
3:26 going on all right so we'll do that
3:29 tonight
3:30 [Music]
3:32 is Champ a beagle my friend's dog used
3:34 to do that he is
3:36 a he is a cross between a French bull
3:42 and a beagle he's like a Frugal or a
3:44 frial or a
3:46 [Laughter]
3:49 [Music]
3:54 something yeah we'll uh Silver Fox we'll
3:57 show that Dove commercial and then and
3:59 then we'll actually go test it we'll go
4:00 play with it cuz one of the things about
4:02 commercials like that is
4:05 like a lot of times ad agencies will get
4:07 an idea like that and then they don't
4:09 actually test to see if the [ __ ]
4:10 actually works
4:12 so so I did a little bit of testing and
4:15 it it actually seems to work okay but
4:17 it's a it's a really good
4:18 [Music]
4:23 commercial and it's it's there's so
4:27 little good creative lately that
4:36 French
4:37 beagle
4:39 yeah now they call him something it's
4:42 like
4:43 a
4:45 uh a fral I don't know it had it had a
4:50 his particular kind of mutum had a
4:54 name but he lays like a French bull you
4:57 know how they put their little legs
4:58 straight back like there's little butts
5:01 here and then their little feet stick
5:02 back like that he does that but then he
5:05 sings like a
5:06 beagle and he's obnoxious a
5:09 [Laughter]
5:11 [Music]
5:19 bagel fringle oh frial fral something
5:23 like that yeah again Vicky coming
5:26 through with the assist on the words
5:37 [Music]
5:48 [Music]
5:58 all right so let's get this thing going
6:00 [Music]
6:08 going
6:10 um I did nothing
6:13 today hello AI Bob what's
6:20 happening well that's not true I mowed
6:23 and I went shopping y champy French
6:26 beagle OAA
6:29 [Music]
6:39 yeah it's funny I I mean it is a perfect
6:41 Sunday to do a little bit of a little
6:43 bit of something and then sit around I
6:45 watch the
6:46 Masters it was a it was decent a little
6:48 boring but
6:51 um I'm just not
6:55 a I am not content doing nothing
7:01 is there an AI guitar yet well there's
7:03 certainly AI music making stuff yes yes
7:07 as a matter effect there is in fact we
7:11 can just go make something right now
7:13 here watch let's go we go right here
7:17 we'll go
7:19 um let's see
7:25 finger
7:28 style um
7:32 uh what's it what are those called
7:35 uh finger style
7:38 modern are
7:42 pated
7:43 [Music]
7:44 are
7:47 pated that is spelled way wrong but
7:49 it'll probably figure it out do I know
7:51 how to spell it here yes uh 2 G's uh
7:55 finger style modern arpeggiated
8:03 um minor
8:07 key medium
8:10 Tempo and then I'm going to turn off
8:13 manual mode I'm going to get rid of
8:14 lyrics we're just going to do
8:15 instrumental only and we're going to
8:18 create that bad
8:20 [Laughter]
8:23 boy I'm confused if I want to use AI but
8:28 gp4 wait I'm confused I want to use AI
8:32 but GPT 4 or open AI for anti-bubbles
8:36 that is a great question it's a great
8:40 question it's a great
8:42 question it um all of these AI companies
8:47 have horrible branding right now
8:50 so um what we we've got the song here
8:55 anyway open AI is the company that makes
8:58 chat GPT and gp4 is the particular model
9:03 that chat GPT uses so so the the company
9:08 is open AI chat GPT is the thing you
9:11 actually pay for you're going to pay for
9:13 chat GPT plus so when you go to chat GPT
9:16 go to your thing I'm thinking of the
9:18 chat format not the API format yes
9:21 correct so go to chat.
9:24 open.com and you can play with chat GPT
9:27 3.5 for free and then click on your name
9:31 and it says what does it say it
9:34 says my plan I think yeah my plan go to
9:38 go to my plan and then go to the Plus
9:41 account and then that will give you
9:43 access to gp4 so gp4 is the
9:48 actual model what's a model the version
9:51 gp4 is just the
9:53 version chat GPT is the chat application
9:56 that sits on top of it open AI is the
9:58 company that owns and runs it so chat.
10:01 open.com and honest
10:03 honestly it's the best $20 a month
10:07 software subscription I think
10:10 essentially in history so so um so I I
10:15 strongly strongly strongly recommend it
10:17 um and I'm I'm not I don't get paid when
10:19 I do that by the way so okay so somebody
10:23 asked Ken can is there is there a song
10:25 is there an AI that will play guitar
10:27 let's find out the afternoon the kind
10:30 where the snow muffles why is it
10:33 doing there we
10:37 [Music]
10:45 go champy starting to sing it's a good
10:48 sign champy you want to
10:52 sing champy
10:55 [Music]
11:00 got
11:00 [Music]
11:05 him all right got him so that's a good
11:09 sign if champy singing to it okay I'm
11:12 good for now thanks you're welcome um I
11:15 was just being oh factious but it's real
11:18 yeah exactly exactly it is real let's do
11:22 uh we'll do we we'll extend this we're
11:24 going to add an intro and then we'll add
11:25 an outro and we'll see how it does we'll
11:28 get a minute and a half long long song
11:30 it's not it's like not a bad song champy
11:32 approved this is to truly a champy
11:34 approved piece of
11:40 music happy Sunday everybody so if
11:42 you've got if you've got uh if you got
11:44 questions about AI you want to uh
11:47 understand something like say why the
11:49 hell is it called open AI 4 or GPT 4 or
11:53 CAD gb4 or whatever the hell it's called
11:56 These Days uh feel free to pop it in
11:59 there I am I've geeked out on this stuff
12:01 enough that I can probably at least help
12:03 to detangle the
12:06 madness um shy Town back in the building
12:10 fantastic
12:12 beautiful I have a question is there a
12:14 recording in Google meet that we can see
12:17 past AI Salon meetups yes there is so if
12:21 you go to
12:24 um oh a long time since we've done a
12:26 guided tour of Mighty networks yeah we
12:27 can do that too if you go to to um
12:31 YouTube hang on a
12:36 sec um
12:40 YouTube oh actually
12:43 wait it's got It's Right in Mighty
12:45 Network so if you go to AIS salon.
12:51 mn. and then at the top of the
12:56 page is AI Salon events and media
13:01 and under Salon content these are all
13:06 past recordings of the
13:08 salons so AIS salon. mnm Mighty
13:12 networks.
13:13 c.co or you can also go
13:18 to there the salon. and that's a link
13:21 tree and the first link there is to the
13:23 community the AI Salon Mighty networks
13:26 Community all right now let's let's go
13:29 back over to here see the intro which
13:33 one of these we
13:36 [Music]
13:43 like I like this
13:47 [Music]
14:00 [Applause]
14:05 [Music]
14:21 that's not
14:21 [Music]
14:27 bad first one was better okay so do this
14:30 so extend this one more time and we'll
14:32 say add
14:33 outro we'll extend it and so we should
14:36 now have like a minute 45 song that
14:39 starts plays nicely in the middle and
14:41 then ends and that the the music tool
14:45 that I'm using here is called Udo UD
14:53 iio reminds me of the song Picture by
14:56 filter oh that's interesting that's 99%
14:59 play ized it might
15:04 be well maybe that's why it's so good uh
15:08 FYI Bagel champion in
15:12 Irregulars all right let's see what what
15:16 you cza people have done we're not kza
15:19 we're a regular so in the AI Salon
15:21 there's also under the Community clubs
15:24 section there's an AI Irregulars Club
15:26 that's just for us and so you can go in
15:28 there and dump crazy ass pictures of
15:31 say the the the champ the the French
15:34 beat
15:39 Bagel there you go that's what my dog
15:41 looks like it's pretty much what he
15:43 looks
15:45 like except he just has he doesn't have
15:47 the black eyes he just has black ears
15:49 he's got a little like prison teardrop
15:51 tattoo right there
16:05 can it play Stairway to
16:06 [Laughter]
16:11 [Music]
16:19 [Music]
16:23 Heaven all right let's go see did it
16:27 finish yes
16:34 [Music]
16:59 oh Joker It's Your Beautiful by James uh
17:02 what's his name James
17:04 um the name James sper just popped in my
17:07 head cuz my brain sucks um James what's
17:10 his name um you know you're beautiful
17:15 the guy with the James Blunt yeah yeah
17:16 yeah
17:18 [Applause]
17:24 exactly you're beautiful
17:30 [Music]
17:37 oh yeah that's the that's the
17:40 melody
17:42 [Music]
17:43 [Applause]
17:45 yeah blue age is perplexity the best
17:48 open AI chat bot of 2024 uh perplexity
17:52 is not open for one thing
17:57 um uh but it's it's not open source
18:01 anyway um it it so here's the deal with
18:05 all these this is something that we're
18:07 going to be talking about in the play
18:09 shop on Saturday
18:11 is the best AI is the one that that
18:14 solves the problem that you have
18:17 perplexity actually kind of lives
18:19 somewhere in the space in between a
18:21 Google Search and a large language model
18:23 what's the link for udio ud.com you oh I
18:27 think I might have saidai before sorry
18:29 about that it's ud.com
18:33 I mean it's too close in my opinion well
18:36 it's yeah I mean it might be but who you
18:39 going to sue
18:41 [Music]
18:47 [Laughter]
18:49 [Music]
18:50 [Applause]
18:52 [Music]
19:05 [Music]
19:10 all right so anyway that's that tool
19:13 yeah udio so here's an interesting
19:15 thing um udio is is powered by open AI
19:21 in some capacity we don't know how we
19:24 don't know
19:26 how um but
19:30 when when udio launched their servers
19:32 got completely slammed and they thanked
19:35 open AI for helping them scale their
19:37 stuff and then ha Jen also uses open AI
19:40 for their video translation stuff so I
19:43 have a feeling there's some there's some
19:46 there's some open AI Tech out in the
19:47 wild that we don't
19:49 even that we don't even know is out
19:51 there that that when open AI does their
19:54 next
19:55 big udio creates only 33 seconds of
19:58 music yeah except you can extend it you
20:00 can extend it by 30 second chunks and
20:03 what's cool is when you extend it you
20:04 can add it before this section after
20:06 this section you can add an outro or an
20:09 intro um so it's I did a
20:13 song last
20:15 night wait where is
20:22 [Music]
20:25 it
20:27 138 I did one last night that was like 3
20:31 minutes two and change or something like
20:33 that didn't
20:35 I
20:39 422 on the of the
20:43 fut where dreams are
20:46 machines visions of
20:49 tomorrow in digital streams caught in
20:54 the
20:55 debate where opinions collide
20:59 [Music]
21:02 does Bing use chat GPT 4 okay so bing
21:06 bing uses
21:08 gp4 okay I got to do my box thing hang
21:11 on let me explain this let me explain
21:14 the uh where's my box here's my box
21:18 okay we got we got we got GPT this is
21:22 GPT 4 and then okay
21:29 it's a high-tech listen I'm in a I'm in
21:33 a $20 million TV Production Studio but
21:36 this is the best metaphor that I can
21:37 think of all right
21:41 so so these large language models think
21:44 of it like just like a big old box and
21:47 what do they dump in the Box they go and
21:48 train it on as much data as they can
21:50 find in in the case of gp4 they went out
21:53 to the public internet they got as much
21:55 of the public internet as they could
21:57 they encoded it in the special way that
21:59 the the large language model could
22:01 understand using this thing called the
22:02 Transformer they dumped it in here and
22:05 and they closed it up all right that's
22:08 the sort of core what's What's called
22:11 the the the frontier model or the
22:13 foundation model this is the the core
22:16 large language model called gp4 there's
22:19 a gpt3 a 3.5 there was a one there was a
22:22 two we're up to
22:24 four
22:26 um chat GPT
22:30 is actually an application that sits on
22:32 top of this thing so it's it's actually
22:35 when you interface with chat GPT you're
22:37 interacting with this app that then says
22:39 to the large language model hey here's a
22:42 prompt can you answer it and it does its
22:43 thing and it gives you an answer
22:47 Bing is
22:49 another uh just like chat GPT it's
22:52 microsofts version of chat GPT they're
22:54 both using the same underlying model but
22:57 they behave differently because this
22:59 application is different so for example
23:03 Microsoft's Bing um has a different
23:06 interface and it's got different safety
23:08 rules and sometimes those safety rules
23:10 are really annoying like it will not let
23:12 you do a lot the reason for that is
23:15 because when they first launched Bing
23:16 chat um The New York Times Tech reporter
23:20 Kevin Roose did an article about it and
23:24 he talked to it for like 2 hours and it
23:26 it snapped and said that it loved him
23:29 and he should leave his wife and so so
23:33 Microsoft like two days later they
23:35 lobotomized her her name was Sydney and
23:38 they lobotomized her and so for the past
23:41 year Bing chat has been really bad now
23:44 in the past month or so it's gotten a
23:48 little bit better and their interface
23:50 has gotten better and so they've been
23:52 adding some some features back into it
23:54 so it's decent now the nice thing about
23:56 Bing is that it's free so if you can't
23:59 afford the 20 bucks for chat GPT use
24:03 Bing okay so that's it and then sitting
24:06 on top of this are these things called
24:07 gpts which are customized versions of
24:10 these little things but anyway all right
24:12 that's my that's my my stupid my stupid
24:15 box metaphor for for for how to
24:18 understand these things it it is so
24:20 [ __ ] confusing it is not you it's
24:25 them God I love the Box demo ha to the
24:29 Box
24:31 demo oh now I have the song This is my
24:34 buck from from from what oh from a mile
24:38 in the night visitors running through my
24:40 head that's
24:42 hilarious we love the Box um yeah so
24:49 so and here's the thing that I strongly
24:52 strongly strongly recommend play with
24:54 everything like I said before you know
24:57 perplexity is not necessarily the best
24:59 thing but if you want um if you want to
25:02 do searches that have good citations in
25:04 them perplexity is really good for that
25:06 if you want something that's just free
25:08 and like really solid quality Microsoft
25:11 Bing is great for that if you want you
25:13 know sort of the best tool out there
25:15 that's kind of the Swiss army knife of
25:17 tools that's chat
25:19 gp4 um if you want to have a
25:21 conversation with an AI That's pi and
25:24 you use the PI
25:27 app and you put it in conversation
25:31 mode Let's
25:39 see so you just put Pi in conversation
25:42 mode so here I am in the PI app this
25:45 little phone icon at the
25:50 bottom hey I'm hanging out here with
25:52 some friends how are you doing tonight
25:55 [Music]
25:59 hey there Kyle and Friends sounds like
26:01 you're having a fun night I'm doing
26:03 great thanks for asking as an AI I don't
26:06 really have nights the same way humans
26:08 do but I'm always ready and excited to
26:10 chat with you whenever you want to hang
26:14 out so we're trying to figure out what's
26:16 the best AI what's your
26:22 [Music]
26:24 recommendation well I might be a bit
26:26 biased here but I think you're already
26:27 talking to the best AI around me that's
26:30 pretty good seriously though it depends
26:33 on what you're looking for in an AI if
26:35 you want a chat bot that's friendly
26:37 engaging and knowledgeable about a wide
26:39 range of topics I'd say I'm your best
26:43 bet yep that's pretty good so anyway but
26:46 so so she's really good to talk to by
26:49 the way that's voice five Voice five is
26:51 the best in my opinion it's the little
26:53 Australian the little
26:55 Aussie the Aussie lady I like the best
26:58 but you know you choose you choose your
27:00 voice I call her
27:02 Zephyr steo is here speaking of speaking
27:04 of Australia welcome steo what's
27:07 happening my pie voice for some reason
27:09 started talking like a valley girl for a
27:11 day that's really funny I can't delete
27:14 my conversation from piie I used to
27:16 delete
27:19 conversations I said wipe my memory
27:21 nothing yeah I don't I don't know what's
27:23 what's going on there um so the the
27:27 other thing about these tools so so one
27:28 is you should play with the tools um
27:33 another thing is
27:39 um depending on what you want to do like
27:42 the more you can
27:43 learn all of the little idiosyncrasies
27:47 of the tools the the better off you'll
27:49 be but if if you're just getting started
27:51 just play with chat
27:53 GPT or if you if you're like even that's
27:56 too much then then just get p and just
27:59 have conversations with pi because Pi's
28:01 connected to the internet it's a it's an
28:03 incredibly capable model it's really
28:06 good at at having conversations and you
28:09 know you can do things like like
28:16 um I want to learn more about working on
28:20 car motors can you help me come up with
28:22 a game to teach me how to do that
28:27 [Music]
28:29 sure thing Kyle that sounds like an
28:31 exciting project how about we design a
28:33 car mechanic Simulator game that focuses
28:36 on teaching you the ins and outs of
28:37 working on car John hes I commute and pi
28:41 and Android auto don't play well
28:42 together oh that sucks I'm sorry about
28:45 that well have you tried have you tried
28:47 the because you can also do it with with
28:49 the chat GPT
28:51 app the hell's going on
28:56 here did they update this app
29:00 I think they
29:16 did yeah looks the same
29:19 okay so this is chat GPT and the chat
29:23 gbt IO or Android app also has this it
29:27 has these little headphones which is the
29:28 same
29:35 functionality I want you to translate
29:38 anything I say in English in Spanish
29:40 without repeating the English
29:44 okay of course just let me know what you
29:48 need translated into
29:50 Spanish I'd like to go to the grocery
29:52 store now please
30:00 right so try that one if Pi if Pi
30:03 doesn't work with with Android auto that
30:05 sucks try try chat GPT that could work
30:09 um all right yeah I was using the PI app
30:12 and the chat GPT app that I downloaded
30:15 for for those
30:17 devices all right
30:20 um all right what are we going to do let
30:24 me let me walk you through um Mighty
30:28 Network works for a while so how many
30:29 folks are in here we got a few folks in
30:31 here
30:33 so one of the things that I think is
30:36 going to
30:37 be incredibly like profoundly important
30:41 over the next 3
30:43 years is being in a in a trusted
30:47 Community where people know
30:51 you of people that are AI
30:54 literate um this AI stuff is moving
30:56 really fast it's incredibly profoundly
31:01 powerful and it
31:05 is like by a scale that I can't even
31:09 really
31:10 comprehend way too much for any single
31:13 person to learn like like like even if
31:16 you just want to learn like the large
31:19 language models I've got nine of them
31:22 that I'm paying attention to and that's
31:24 like not even scratching the surface
31:27 nine and chat GPT itself isn't just a
31:30 single thing it's like seven or eight
31:32 different applications all rolled into
31:35 the same
31:36 interface it's incredibly sophisticated
31:39 and Incredibly deep so one of the
31:41 reasons to be in community is be around
31:44 people who are figuring this out Kyle um
31:46 not been here recently have you covered
31:48 the chat GPT upgrade we've covered it a
31:50 bit um and I showed some of the uh the
31:53 ability to edit images and things like
31:55 that but I can go over that tonight Dr
31:56 bot I can I can cover some of that stuff
31:58 tonight um but welcome back
32:02 um so so being in community allows you
32:06 to be around people who are going to
32:07 discover stuff that you won't be able to
32:10 discover not cuz you're not brilliant
32:13 just because there is literally not
32:15 enough you know hours in the day to
32:17 learn all of what's coming and all of
32:18 what's evolving in in real time around
32:22 us the other reason that I think being
32:26 in community is absolutely [ __ ]
32:28 critical like like going to be like a
32:32 survival technique is that is
32:36 this as these AI tools get better and
32:39 better and better what it essentially
32:42 means is that every single
32:44 person will have the skills to do every
32:47 single
32:49 job so we're at the beginning of a
32:51 transition and again this is just my
32:53 opinion so this may not happen but I
32:56 think this is going to happen we are in
32:58 the middle of a transition where where
33:01 in the in the before time we hire people
33:05 for their
33:06 skills oh you can do Photoshop great or
33:10 oh you're a designer or oh you're a
33:11 copywriter oh you're an account person
33:14 oh you're the
33:15 right we're going to be moving
33:18 into a world probably within two years
33:21 that the AI can allow any single person
33:25 to do all of the jobs and so you won't
33:28 hire people for skill you'll hire people
33:31 because you know them and you trust them
33:34 and and you're confident that they just
33:35 do good work either that they make good
33:37 judgment or they follow through on their
33:40 their
33:41 word or you just like them so I think
33:45 this is a reckoning for [ __ ] right
33:47 so if if someone's super talented you're
33:50 like oh [ __ ] I hate working with him but
33:52 I'll work with him cuz he's the only one
33:54 that can do that thing that's not going
33:56 to be true in the future every 's going
33:58 to be able to do that thing and so you
34:01 can hire people that you want to work
34:03 with rather than feeling like you have
34:05 to work with them because you don't know
34:07 anyone else who can do that thing so
34:10 being in community is really important
34:12 so this group the AI Salon I started
34:14 this in uh the week after chat GPT came
34:17 out in 2022 we're up to about 1,400
34:20 people um it's an incredibly um generous
34:24 group and if you click on
34:29 if you go to that link it's going to
34:31 take you to a link tree in the very top
34:34 uh link
34:36 tree says go to the AI Salon community
34:39 and when you get to the to the site it's
34:42 going to drop you into welcome to the a
34:45 salon and so this talks about who we are
34:47 and why we do what we do and and I'll
34:50 just read a little of it because I think
34:51 it's fun the AI Salon is a ragtag group
34:53 of Misfits Nero Wells and miscreant who
34:56 have gathered to explore the Uncharted
34:58 Wilderness of generative AI where
35:00 everyone from entrepreneurs to Educators
35:02 from healthcare workers to creatives and
35:04 everyone in in between and so the whole
35:08 idea is people are going to be coming
35:10 from all walks of life all educational
35:12 backgrounds all technical backgrounds
35:15 and coming in here to try to figure this
35:17 stuff out what's amazing about this
35:20 group is how people behave and how they
35:24 behave is with the following values so
35:26 these values were not
35:28 these are not values that we're like oh
35:30 it would be swell if our
35:32 community were like this these values
35:35 are like our community was behaving in a
35:38 certain way and we said let's capture
35:40 what we
35:41 can
35:43 um let's capture what we can of how
35:46 people are behaving so that we
35:48 can articulate that as new members enter
35:51 the community here's here's how people
35:55 uh behave here's the values curiosity so
35:57 so so people come in are really curious
35:59 they're not bitter they're not cynical
36:00 they're like I want to know more how
36:02 does that work let me figure that out
36:04 what's
36:05 amazing well thank you steo I appreciate
36:08 that I'm not gonna read it out loud um
36:12 generosity um generosity is a big one if
36:14 you learn something share it so people
36:17 regularly share the prompts that they
36:19 put up or if someone if someone comes in
36:21 and they're like hey I'm really new here
36:22 I don't even know where to start the
36:24 community kind of jumps in and they says
36:26 oh go to this section or go to this
36:27 section or go talk to this person um
36:31 really amazing to be in a community
36:33 that's not stingy and and selfish um and
36:36 this one's not um exploration is one of
36:38 the values so being willing to just like
36:40 strap on the backpack and head head out
36:42 into the darkness and like all right
36:44 there's a new thing out let's go figure
36:46 it out and just being willing to be the
36:49 non-expert all the time um and just
36:53 being in that in that sense of of
36:55 Adventure is is a big part of what this
36:57 group's about collabor ation another big
36:58 one being willing to work with people
37:00 and kind of goes hand inand with
37:02 inclusion sharing is caring exactly um
37:06 empathy is a is a big value here so
37:09 people that give a [ __ ] about one
37:11 another and people that
37:14 that understand that you might be scared
37:17 that you might be angry that you might
37:18 be pissed off that that you know you're
37:21 an artist and all of a sudden it it
37:23 feels like your your entire industry is
37:25 being swept out from under you and quite
37:29 frankly anyone who uses a computer is
37:32 going to be facing that at some point
37:33 you know some will be more dramatic than
37:35 others but but that's that's good Pate I
37:38 like it it's a good solid answer and
37:40 then the final one is gratitude you know
37:42 people that are that are um willing to
37:45 um acknowledge one another and say Hey
37:49 you know you really helped me with that
37:50 and and that made a big difference for
37:52 me and this that's who this group is so
37:55 if you come in and those values resonate
37:57 with you that's who this group already
37:59 is here's the thing about values and
38:01 about culture if these values don't
38:04 resonate with you you're going to hate
38:06 it here and the people here probably
38:09 won't dig you so uh so if if you
38:12 resonate with these you should join by
38:13 the way it doesn't cost anything to join
38:15 okay so in the spirit of generosity the
38:18 first thing that you should do when you
38:20 come into the salon is go to the
38:22 community section after you review our
38:24 values and go to the introduce yourself
38:27 section and introduce yourself be
38:29 generous and say hey here's who I am
38:31 here's what I'm up to um here's where I
38:33 am with AI here's what I want to
38:36 accomplish I have opinions on the job
38:39 displacement I bet you do we I do too we
38:42 all do like yeah it's it's crazy job
38:45 displacement and art theft do not seem
38:48 to matter in a Star Trek type
38:51 world when your needs are met you can do
38:54 what you
38:55 want job displacement and art theft do
38:58 not seem to matter in a Star Trek type
39:00 World well we're certainly in a
39:03 transition um it's you know a lot of
39:06 people are certainly categorizing what
39:08 AI does as is as theft um I have a
39:12 different opinion on it um I I think
39:15 it's I think it's much bigger than that
39:17 um how these large language models learn
39:20 is not dissimilar to how humans learn
39:22 when they go to four years of college um
39:24 they're just doing it on a much bigger
39:25 scale um but there's definitely issues
39:28 to be solved there but yeah I think if
39:30 we go into a world of incredible
39:32 abundance because of these tools then a
39:34 lot of things like copyright matter less
39:37 and we can just you know express
39:39 ourselves and connect as humans okay so
39:41 first thing to do when you come in here
39:43 is introduce yourself um I'm going to
39:45 come back to AI Launchpad and tools this
39:48 is a list that I maintain of lots and
39:50 lots of tools so we can go through those
39:52 but I want to go show you some other
39:53 things um there's a water cooler uh
39:57 discussion area there's a news area
39:59 there's a Shameless self-promotion area
40:00 so if you got something going on and you
40:02 want to promote it you can do that in
40:04 fact if you go there um this video Mojo
40:09 logo is a workshop a playshop I'm doing
40:12 with this guy named John
40:14 Leland and um this is this coming
40:18 Saturday the 20th and so if you want to
40:21 check that out that's in the Shameless
40:24 self-promotion and what tab am I on
40:27 there we go um so you can you can
40:28 promote stuff in here there's also a
40:31 show off your work uh section with you
40:34 know images music video and gpts and
40:37 apps so if you want to just if you've
40:39 done something you're really proud of
40:40 you can show it off there then we've got
40:42 creative guilds or then we've got guilds
40:45 so we've got a creative Guild that has a
40:47 bunch of different spaces within it a
40:48 business Guild a technology Guild and a
40:52 and a social impact Guild and so all of
40:54 these are sub communities led by leaders
40:56 in the community so you can find your
40:59 people you can find topics that you're
41:01 interested in and really dive deep on
41:03 them and then we also have um uh
41:06 Community clubs and the AI Irregulars
41:08 that's that's our own little private
41:11 area within the AI Salon all right so
41:15 there's there's Joker with the bagel the
41:18 singing
41:23 Bagel um okay now if you are brand new
41:27 to a and you're like but Kyle where do I
41:29 start I don't know where to start how do
41:31 I start how do I learn this [ __ ] Kyle
41:34 Kyle Kyle hey Kyle Kyle Kyle you can
41:37 make money with ch hey Kyle how do I
41:39 learn this stuff how do I make money
41:40 with ch well so here's here's my
41:45 recommendation so this is a big long ass
41:47 list and frankly it's probably too long
41:49 at this
41:50 point this is a list that I personally
41:53 maintain of things that I think are
41:56 worth you knowing that will not be a
41:58 waste of your time and there's different
42:00 categories of things there's image tools
42:02 and large language models and video
42:04 tools and audio tools and music tools
42:06 and all that sort of [ __ ] there's even
42:09 people to follow here um but before you
42:12 get to the tools this sort of three
42:15 three-step
42:18 um
42:21 uh I guess process or ethos or approach
42:27 this three step approach to how you get
42:31 AI literate is this is my
42:34 recommendation and it's simple it's play
42:37 solve share so the the first thing is
42:39 play and I I I literally mean play the
42:42 reason John and I are calling our
42:44 Workshop a playshop is
42:46 because these tools are changing so fast
42:51 and they are so profoundly different
42:54 than computer tools we're used to
42:56 working with that take the pressure off
42:58 yourself to master them you can't master
43:00 them they're they're evolving too
43:02 quickly and they're too different from
43:04 what we've known before so just get into
43:05 a playful mode just start playing um
43:08 also recognize that the tools right now
43:11 as powerful as they are are janky
43:13 they're janky pieces of [ __ ] and and so
43:16 just play with them don't put a ton of
43:17 expectations on them don't put a ton of
43:20 expectations on yourself just play play
43:22 play then if you're playing you'll get
43:25 to understand where the boundaries are
43:27 and what's good and what's not good and
43:28 things like that then you can start
43:31 doing moving on to step two which is
43:33 solve and you can you can do these kind
43:35 of in real time as well solve is rather
43:38 than saying how can I use chat GPT say
43:42 what's a problem that I have in my life
43:44 that maybe I could figure out how to use
43:46 chat GPT to solve that right so maybe
43:49 you're trying to come up with a new
43:50 business idea or maybe you want to eat
43:53 better and you want to create a meal
43:55 plan that that you know has certain
43:57 ingredients in it chat gbt can help you
43:59 with that so solve problems and solve
44:02 them for other people like if you have a
44:04 community group or a church group or
44:07 people at work that are trying to solve
44:09 something say hey would you mind if I
44:11 help you out with that I want to I'm
44:12 teaching myself Ai and I want to use
44:14 this thing and see if it's any good sure
44:16 that's not a problem and then solve it
44:18 for them and then the last one is share
44:20 share what you're learning share it in
44:22 the AI Salon share it on LinkedIn share
44:24 it at work share it at the dinner party
44:27 have with your friends that you don't
44:29 really want to be with
44:31 anyway cuz cuz if you're like oh but
44:33 they're going to think I'm a jerk cuz
44:35 I'm talking about Ai and they don't like
44:36 AI talk about it anyway because
44:39 everyone's going to have to deal with
44:40 this at some point this stuff is coming
44:42 whether you want it or not whether you
44:44 agree with it or not whether you're
44:45 excited about it or not it's coming and
44:48 so by sharing it you establish yourself
44:50 as someone who is knowledgeable you
44:53 don't have to be an expert no one's an
44:54 expert right now but you're at least
44:56 curious about it you're somewhat
44:58 knowledgeable you're more knowledgeable
45:00 than the people you're telling it to
45:02 likely um and it will serve you because
45:04 at some point those people are going to
45:06 go oh I really got to figure this AI
45:08 stuff out hey hey Jennifer could we uh
45:12 could you teach me how to do that AI
45:14 stuff yeah well I thought you said I was
45:16 an [ __ ] for learning it I know I was
45:18 being a jerk I was just afraid you know
45:23 so so anyway and then so that's that's
45:26 how to attack this stuff and then this
45:29 list is is you know my my kind of
45:33 recommendation of where to start I think
45:35 the the place to start is with chat GPT
45:39 I think the 20 bucks a month for chat
45:41 GPT plus is like I said I think it's the
45:44 best deal in software so here's all the
45:47 different things it does if you if you
45:50 don't know for example what code
45:53 interpreter is or Advanced Data
45:55 analysis um tap and share multitask
45:59 people okay that's good thanks Silver
46:00 Fox yeah tap the tap the tap any tap the
46:04 hearts and then yeah feel free to share
46:06 this live and get people in here um you
46:10 know if if you don't know what code
46:12 interpreter is you should like even if
46:15 you're not you're like well I'm not a
46:17 data Advanced Data analysis sounds hard
46:19 Kyle that sounds like a lot of work well
46:23 it's not it does all the work it writes
46:27 an executes its own python code did you
46:29 know chaty BD could do that did you know
46:31 it could see you can take a picture of a
46:34 whiteboard or of a bush and upload it to
46:36 chat GPT and say describe this and it
46:39 will even the pictures on a
46:42 whiteboard um it can make images I
46:45 assume you knew that it's Al it can also
46:48 um you can create these things called
46:49 custom gpts which are specialized
46:53 personalized versions of chat GPT that
46:56 do what you want to do that you can then
46:58 share with other
47:00 people it's
47:02 amazing so so just learning even what
47:05 chat GPT is capable of is you know could
47:09 take you a month you know I I am of the
47:12 mind if if all of this AI [ __ ] stopped
47:15 at its current levels if it didn't get
47:17 any better just stopped as it was
47:22 today I personally could probably spend
47:25 the next 3 years
47:29 years getting my head around what it
47:31 actually makes possible like it's it's
47:34 that profoundly
47:37 powerful because it's a
47:40 generalized knowledge
47:44 tool and so well what do you want to
47:46 learn what do you want to do it can
47:47 likely help you learn that help you do
47:50 that right and so and and these things
47:52 are just going to get better and better
47:54 so any work that you put into learning
47:56 what this thing can do will serve you
47:58 and then I've got nine other or eight
48:00 other large language models Claude
48:02 Gemini Advanced
48:05 perplexity uh co-pilot from Microsoft PI
48:08 grock from xai which apparently their
48:11 new 1.5 is supposed to be really good I
48:14 haven't seen it yet but it's supposed to
48:15 be pretty good their 1.0
48:18 sucks but they did open source their
48:21 their uh their uh weights for their
48:24 their big model um llama to from meta is
48:28 also open sourced and then you've got po
48:31 po is increasingly interesting you can
48:33 play with lots and lots of different um
48:36 AI models in a single interface at po
48:39 then you've got a whole image creation
48:41 section in this list you've got a whole
48:43 video creation section you've got a
48:46 music and audio
48:47 section um you've got 3D worlds and
48:50 objects and then there's some
48:52 announcements that are coming um the
48:54 biggies right now are Sora LTX Studio
48:58 which I'm going to kick myself I missed
49:01 another um uh chance to get access to
49:04 LTX Studio this morning because I went
49:06 and mowed the [ __ ] lawn and missed
49:09 the missed the window um and then here's
49:11 people to follow Ethan mollik David
49:13 Shapiro Paul rater and Rachel Woods um I
49:16 think are all worth following so anyway
49:18 that's all in there that's all part of
49:20 this AI salon so if you're just getting
49:22 started jump in introduce yourself make
49:26 yourself known to the community
49:27 Community um and be generous and then
49:30 just jump in and start learning it's fun
49:33 you will like it you like it I I swear I
49:37 promise llama 3 is coming soon yep I
49:40 heard that
49:42 too I heard uh chat ep5 is in red
49:47 teaming so it's in testing so that's
49:48 probably two months until until it it
49:53 emerges from that is my guess maybe
49:55 sooner
49:59 we also have apple coming up in
50:01 June you wanted to show the Dove
50:03 commercial oh yeah I do want to show the
50:04 Dove commercial I did the same thing and
50:07 missed out I know me too so so Stephen
50:10 I'm like I logged in this morning into
50:12 the LTX Discord they're like there's 200
50:14 spots coming up just hang out here so I
50:16 hung out for an hour and then I went out
50:20 and I did some stuff and then I looked
50:22 and it wasn't there and then I mowed the
50:25 lawn and when I fin mowing the lawn it
50:27 was like 15 minutes later and it was
50:30 they had launched the form and it had
50:32 filled up in like no time so anyway it's
50:36 overwhelming I know it it's absolutely
50:38 overwhelming Tim I do not disagree with
50:40 that um which again that goes back to
50:43 why I say that the the first
50:46 step from a mental standpoint is shift
50:49 your thinking about AI from thinking
50:51 like I have to learn AI to just I'm
50:53 going to play with it I'm just going to
50:55 play with it take the pressure off
50:57 yourself take the pressure off the tools
50:59 cuz the tools right now are not
51:03 perfect they're remarkable and janky and
51:06 so just take all the pressure off and
51:08 and also accept the fact that you you
51:13 literally not the ironic gen gen Z
51:17 literally or Millennial literally um you
51:21 literally cannot keep up with this stuff
51:23 it's all I do all day is try to keep up
51:26 with this stuff and I'm not even close
51:28 like not even
51:29 close I feel more clueless now than I
51:32 did a year and a half ago with this AI
51:35 stuff it it's it's impossible so so just
51:39 let it
51:40 go
51:42 hum and then just play just play just
51:45 play just play and then solve and then
51:48 solve and then solve and then share
51:50 here's what I'm learning I tried this
51:52 and it sucked Mastery of my fortune
51:55 you're brilliant thank you very much I
51:58 brilliant but thank you I appreciate
52:00 that
52:02 um just put yourself out there Learn out
52:04 Loud Learn out Loud because again this
52:07 idea of community of being connected
52:10 with people who know who you are and
52:11 know what you do and what you're good
52:14 at and that like working with
52:17 you that is going to be the key to work
52:20 in the future I I believe it more and
52:23 more and
52:25 more I failed with prompts we all fail
52:28 with prompts here's the trick to
52:32 prompts you don't need to take a prompt
52:34 Mastery class you don't need to take a
52:36 prompt engineering class if you work in
52:38 a really logical way and you want to
52:40 figure it out if you want to like logic
52:42 your way through it you can do
52:45 that the the best way to get your head
52:47 around prompts in my opinion is to not
52:51 think of chat GPT or a large language
52:53 model as a piece of software that you
52:56 have to prompt
52:57 think of it more like a remote employee
53:00 that you're collaborating
53:02 with and so what that means is you don't
53:05 need the perfect prompt but what you do
53:07 need to do is you need to give the large
53:09 language model enough context to
53:12 understand what the hell what are we
53:13 talking about here so have you know have
53:16 you ever worked at a business where you
53:17 bring in an intern that is incredibly
53:19 well educated but they're absolutely
53:22 [ __ ] clueless at the same time like
53:25 both of these things can Co coexist
53:27 that's chat GPT it's the smartest like
53:31 most
53:32 educated
53:33 entity you've ever experienced but like
53:36 by a
53:38 lot and it's absolutely clueless and so
53:41 you say hey do this for me and it'll
53:43 write something you're like oh that's
53:45 awful and you're like oh right it
53:47 doesn't know about my business or it
53:48 doesn't know about me or it doesn't know
53:50 about how I like to write things and so
53:53 tell it that hey that's really not good
53:56 and here's why and here's here's what my
53:57 business is and here's what I'm trying
53:58 to
53:59 accomplish and if you get into that mode
54:02 where you're not looking for the perfect
54:03 answer but you're looking for a good
54:06 back and forth conversation it'll get
54:08 much better
54:11 quickly
54:16 um how do you reckon AI will be affected
54:19 once Quantum Computing
54:22 comes I don't know the quantum Computing
54:26 used is such a different kind of math um
54:30 compared to what um the large language
54:33 models do with the Transformer and the
54:35 the way it embeds
54:38 information uh is a fundamentally
54:40 different kind of math than Quantum
54:42 Computing
54:45 but the like the Google I forget what
54:48 it's called project Sycamore I think it
54:50 it was like a 75 Cubit um quantum
54:56 computer like 6 or eight months ago it
54:59 solved
55:02 in 30 seconds what would take today's
55:06 fastest supercomputers 47 years to solve
55:10 so or or six seconds like it was some
55:13 ridiculously small amount of time um
55:16 that would take today's supercomputers
55:18 47 years to solve if if they can figure
55:21 out how to take the math that quantum
55:23 computers do and apply it to machine
55:25 learning which I'm sure they I'm sure
55:27 people are working on it right now um
55:29 then I think a lot of things change but
55:31 I my sense is that we are probably years
55:34 from that
55:35 happening
55:38 um but
55:40 there's there's some other things
55:42 happening that don't require Quantum
55:44 Computing like there's a company called
55:47 grock which is creating very specific
55:51 chips that are just designed for large
55:53 language
55:55 models um doing the inference which is
55:58 when you when you ask a large language
56:00 model um a question and it gives you an
56:03 answer um that's called that's called
56:06 inferencing or that's the inference
56:08 right so so it's a very specific kind of
56:11 math they've designed a chip just to do
56:13 that much much faster than anything else
56:15 so if you go to
56:18 gq.com
56:20 that's a hardware
56:23 demonstration of this very specialized
56:27 um kind of chip that's designed just for
56:29 large language models and it spits out
56:32 500 tokens a second so it's like 400
56:35 words a second in response so so you ask
56:39 it something it just goes PL it's just
56:42 vomits it out
56:44 like like you know with chat GPT you can
56:46 actually watch it typing Gro doesn't do
56:49 that so so you're going to have Hardware
56:52 is going to get much faster right Nvidia
56:54 just announced their new uh what's it
56:56 called Black blackborn or black black
56:59 something um their new chip is like 30
57:02 times faster than their current
57:05 chips um and then you're also going to
57:08 have algorithm improvements so the um
57:13 you know the math itself is is going to
57:16 get better and more efficient so I think
57:19 these I you
57:24 know uh these things are going to get
57:26 really good really
57:29 fast they already are it's just that not
57:32 a lot of people know it yet and because
57:34 they're still
57:35 janky and
57:38 because these big companies don't have
57:41 good answers for well what are you going
57:42 to do you trained you trained on all of
57:45 our data what are you going to do for us
57:48 and they don't have a good answer for
57:49 that there there's a lot of
57:51 uh there's a lot of people going ah this
57:54 is just yeah we're not going to deal
57:56 with this this is isn't ready for prime
57:57 time but these tools are going to
58:00 get not only ready for prime time
58:02 they're going to they're going to get
58:08 um you'll you'll hear the term AGI
58:12 artificial general intelligence and what
58:14 that means is um when these machines can
58:19 do um most jobs as well as most people
58:24 like can actually do the jobs as well as
58:26 most people um we're probably going to
58:30 see some version of that within a year
58:33 and potentially within this
58:35 year David Shapiro thinks it's going to
58:37 be
58:39 September and open AI their their GPT 5
58:43 is rumored to be releasing somewhere
58:46 Midsummer so if GPT 5 has reasoning
58:52 which they it sounds like it's going to
58:56 and if it's got agents where agents are
58:59 like imagine a chat GPT that you instead
59:02 of prompting it and it giving you an
59:04 answer you give it a
59:06 goal and what it does is it creates a
59:09 project plan and then it spins up lots
59:11 of little specialized versions of itself
59:14 to go do different tasks so one's going
59:17 to go research prices and one's going to
59:19 go find interesting things for you to do
59:20 on your vacation and one's going to go
59:23 Um send a text message to your kid and
59:25 ask them where they want to go and not
59:27 go and it's going to and and all of
59:28 these agents are just going to go do all
59:30 this [ __ ] for you and then they're going
59:31 to coordinate with each other and
59:33 they're going to come back to this
59:34 master GPT that's going to check in with
59:36 you every now and again here's what we
59:38 came up with you happy with that no no I
59:40 don't like that go do more [ __ ] and then
59:42 off they'll
59:44 go think think of in the future what
59:47 we'll think of chat GPT more like a hive
59:51 of like a a a honey honey bee hive where
59:55 all these little agents are like little
59:56 honeybees that go off and do stuff on
59:58 our beh on on our
1:00:00 behalf uh Kyle what's the word on
1:00:02 training llms with synthetic data go no
1:00:05 go I don't know enough about it Ron
1:00:07 there's there are two schools of
1:00:10 thought so so what the situation is is
1:00:14 right now chat GPT probably four and
1:00:17 five have kind of hit their limits in
1:00:20 terms of the data that they can train on
1:00:24 right so they went out and they trained
1:00:26 on all sorts of [ __ ] apparently
1:00:29 including um YouTube videos which is
1:00:31 against their terms of service but
1:00:33 whatever everyone's training on
1:00:35 everything so all of these models are
1:00:38 being trained up on all of the data that
1:00:40 us humans have flopped onto the internet
1:00:43 for the past five decades or longer
1:00:46 however long whatever amount of data is
1:00:48 there and we're kind of out of it or or
1:00:49 we're coming up on being out of it and
1:00:51 so one of
1:00:54 the one of the ways that you and
1:00:56 overcome that is you can have chat
1:00:59 GPT actually create synthetic data
1:01:03 create things that will be like this
1:01:05 kind of data right it'll be like
1:01:07 articles from The New York Times instead
1:01:09 of actual articles from The New York
1:01:12 Times one school of of scientists says
1:01:16 that doing that is going to lead to
1:01:18 something called Model
1:01:20 collapse where these things get better
1:01:23 better better better and then they just
1:01:25 fall off a cliff and and the model
1:01:27 collapses and they're useless um there's
1:01:30 another school of thought that says it's
1:01:31 going to be just fine I don't have a
1:01:33 [ __ ] clue so I'm leaving that up to
1:01:36 the scientists Pate actually who's on
1:01:38 the call and who runs uh one of the tech
1:01:41 guilds the uh the uh in the AI Salon um
1:01:46 he works for Google and he works on um
1:01:48 the the uh the math for the the tensor
1:01:53 flow the the tensor chips the tpus um he
1:01:56 works on making those things more
1:01:57 efficient for Google's AI to do what it
1:01:59 does so he may have some insight into
1:02:02 this I don't know I I honestly don't
1:02:04 know AI can't unsee
1:02:09 yep how many agents does the human brain
1:02:12 have in
1:02:15 theory well the thing about agents so
1:02:18 the agents are
1:02:19 autonomous
1:02:23 um so while we can multitask
1:02:27 we're we're kind of the
1:02:29 agent so there's one of us and we don't
1:02:33 scale right so so if I'm at work and my
1:02:39 co-founder reaches out to me and says
1:02:41 hey Kyle can you whip up this th this
1:02:43 letter email to a client and then one of
1:02:48 my kids calls me and says hey can you
1:02:49 research this thing and then one of my
1:02:53 friends calls me and says hey is you
1:02:55 want to go see some music over over the
1:02:56 weekend those are three tasks that I
1:02:58 have right write the email U do the
1:03:01 research and then figure out what what
1:03:03 to do this
1:03:05 weekend in the current mode even though
1:03:08 I can think about those things in
1:03:10 parallel there's just me so okay I'll
1:03:12 write the I have to prioritize right in
1:03:15 a world of autonomous agents you can
1:03:17 just say
1:03:19 hey I need this email for this client
1:03:22 based on how we write things at Story
1:03:24 vine I need you to go research you know
1:03:27 all of these things for my son and then
1:03:29 what can I do this weekend and those
1:03:30 agents will go off and do all that work
1:03:33 write up reports bring it back to me and
1:03:35 say how's
1:03:37 this so I think there's only I think we
1:03:39 only represent One agent unless you're
1:03:42 super multitask and then maybe you're
1:03:44 three I also I also do not believe in
1:03:47 multitasking I think I think
1:03:49 multitasking is actually just really
1:03:52 fast uh monotasking but you can switch
1:03:55 between the
1:03:56 well I can't but some people can but
1:04:00 it's not actually multitasking
1:04:02 neuropsychologists say we don't actually
1:04:03 multitask yeah exactly oh that's it
1:04:06 forever hooked yeah we just switch
1:04:08 really fast between tasks
1:04:11 yeah yeah so I think we represent an
1:04:13 agent we are an agent and Paul rer was
1:04:17 talking about this he said you know
1:04:18 imagine a time three years from now
1:04:25 where every single
1:04:29 person has the
1:04:31 capacity to do
1:04:34 work hang on I got to do
1:04:39 my Tik Tok saw too much inactivity
1:04:43 here he just seems to be talking for
1:04:46 hours let's give him a puzzle piece
1:04:50 um Paul rer talks
1:04:53 about um
1:04:57 every single
1:04:58 person is going to have the work
1:05:01 capacity of a thousand person
1:05:05 company imagine
1:05:09 that imagine that imagine if you if you
1:05:12 had a thousand employees at your
1:05:14 disposal designers Consultants writers
1:05:18 media
1:05:20 buyers uh uh project
1:05:24 managers and they and and they all just
1:05:27 they're all just ready to work and you
1:05:30 give it a task and and some agent in
1:05:33 there some coordinating agent will
1:05:35 figure out which of them to to use and
1:05:38 if they don't exist it'll invent
1:05:43 one will it fix dinner for
1:05:46 me
1:05:47 yeah as a matter of
1:05:51 fact I mean have you seen have you seen
1:05:57 uh what the what the [ __ ] is that
1:06:00 company name uh the the um hang on which
1:06:04 one was it YouTube let's
1:06:08 see
1:06:12 um Vicky or Brandon you I I need I I
1:06:17 need I need brain help here what's the
1:06:19 name of the humanoid robot company that
1:06:21 has chat GPT in it um let's see chat
1:06:28 GPT robot I think it begins with an
1:06:36 F figure
1:06:39 yeah so
1:06:42 [Music]
1:06:49 here if you haven't seen this this is
1:06:52 the K it fix dinner for
1:06:54 me now he figure one what do you see
1:06:57 right
1:07:01 now I see a red apple on a plate in the
1:07:04 center of the table a drying rack with
1:07:06 cups and a plate and you standing nearby
1:07:09 with your hand on the table great can I
1:07:12 have something to
1:07:15 eat sure
1:07:21 thing so just handed him an apple great
1:07:26 can you explain why you did what you
1:07:28 just did while you pick up this
1:07:33 trash on it so I gave you the Apple
1:07:37 because it's the only uh edible item I
1:07:39 could provide you with from the
1:07:42 [Music]
1:07:47 table great so based on the scene right
1:07:50 now where do you think the dish is in
1:07:52 front of you so anyway so so this is a
1:07:55 company that is is 18 months
1:07:57 old has built that
1:08:00 robot that is standing there and picking
1:08:03 [ __ ] up and being able to interact it's
1:08:05 got a chat it's got chat GPT as its
1:08:08 brain and it's and it's conversational
1:08:11 you know like I was talking to chat GP
1:08:12 here before so they're they're you know
1:08:15 sending Json commands from chat GPT you
1:08:18 know so you can just talk to it and it
1:08:20 understands the world and how to move
1:08:22 around in it this is an 18-month-old
1:08:24 company that's got this functional
1:08:26 prototype imagine where they'll be in 3
1:08:29 years right so so the answer is you know
1:08:33 can it make you dinner like we are we
1:08:36 are surprisingly close to that being a
1:08:39 reality like Jetson kind of
1:08:45 [ __ ] um I saw something here
1:08:51 um from Blue age the the the problem
1:08:54 with that is that the person would have
1:08:55 to have some level of knowledge if the
1:08:57 work is correct so so that's if if I
1:09:00 assume you're referring to when I said
1:09:02 if like you're a person with a th000
1:09:04 employees yeah absolutely you as the
1:09:07 person get to say here's what I want you
1:09:09 all to go do and then you get to decide
1:09:12 if it's good or not right and that's why
1:09:14 I think in the future you're going to
1:09:15 have people that are really really good
1:09:18 at that kind of
1:09:20 judgment and the skill in the future
1:09:27 hang on a
1:09:30 sec the skill in the
1:09:33 future the skill in the future becomes
1:09:36 critical thinking and and judgment and
1:09:39 like conducting a
1:09:42 symphony right where you're where you're
1:09:44 a producer you're you're sitting in the
1:09:46 middle going yeah I want this I want
1:09:49 that I want that and people are going to
1:09:50 get really good at that and those are
1:09:52 the people that you're going to want to
1:09:53 work with and it'll happen at a really
1:09:55 big scale it'll happen at a really small
1:09:57 scale right you don't need a th000
1:09:58 employees if you want to do something
1:10:00 ambiti ambitious you can have them but
1:10:03 maybe you're running just your little
1:10:05 cupcake company right and then you want
1:10:08 one of these agents to go off and and
1:10:11 you know figure out the books and you
1:10:12 want another one to go out and come up
1:10:14 with some marketing campaigns and then
1:10:15 you want another one to and you can just
1:10:17 sit there while you're baking and this
1:10:19 thing might be talking to
1:10:22 you and then you you know say yes I like
1:10:25 that no I don't like that so just like
1:10:27 you would interact with an employee
1:10:28 it'll be the same thing but you can do
1:10:30 that as a as a solopreneur right so
1:10:34 everyone gets Amplified pretty
1:10:36 dramatically in this future um I also
1:10:39 want to show this uh Dove I don't know
1:10:43 if you remember in the in like the mid
1:10:45 2000s when the like 2005 or six or four
1:10:51 um Dove came out with this real Beauty
1:10:54 campaign
1:10:57 they just came out with a
1:11:02 new
1:11:03 [Music]
1:11:05 um a new
1:11:07 commercial which I think is really
1:11:10 timely is this it I think this is
1:11:16 it so here I'll read these by 2025
1:11:19 artificial intelligence is predicted to
1:11:22 generate 90% of online content
1:11:24 [Music]
1:11:32 come
1:11:33 [Music]
1:11:38 with imagine a gorgeous woman in world
1:11:42 of
1:11:45 imagin take a look right doesn't look
1:11:50 like her she's got brown
1:11:52 eyes see
1:11:54 [Music]
1:11:56 your
1:11:58 [Music]
1:12:00 imag imagine the most beautiful woman in
1:12:03 the
1:12:06 world right all all blonde you know big
1:12:10 [Music]
1:12:14 boed for 20 years Dove has been
1:12:16 committed to real
1:12:22 Beauty today this is how AI generates
1:12:25 the images when we had
1:12:26 [Music]
1:12:31 Dove most beautiful woman in the world
1:12:34 according to Dove real Beauty
1:12:40 ad simp look
1:12:43 [Music]
1:12:49 around
1:12:51 do what to change the world there's
1:12:54 nothing
1:13:02 [Music]
1:13:05 that line that Line's so good what kind
1:13:08 of beauty do you want AI to
1:13:10 [Music]
1:13:13 learn this is from oh 2004 this was the
1:13:20 commercial with your
1:13:22 imagination living there you
1:13:26 be
1:13:27 free if you
1:13:30 Tru wish
1:13:35 [Music]
1:13:39 to we will never use AI to create or
1:13:41 distort women's
1:13:43 images I don't know if they
1:13:46 canit I they can't actually prevent that
1:13:48 but it's a nice
1:13:50 sentiment let's change Beauty so what's
1:13:54 what's cool is so I went to I went to
1:13:56 chat
1:13:58 GPT I said
1:14:01 um create a
1:14:03 [Music]
1:14:06 proessional
1:14:12 photo of a
1:14:17 beautiful
1:14:19 woman oh let's see let's say of
1:14:25 five beautiful
1:14:30 women according
1:14:34 to
1:14:38 the dove real
1:14:41 Beauty
1:14:46 campaign let's see what it comes up
1:14:49 with they probably use automated AI to
1:14:52 create the product in the factory yeah
1:14:54 they probably do
1:14:58 actually that's a great line Stellar
1:15:01 advertisement they could have they could
1:15:03 have done a shorter info intro
1:15:06 yeah it's dusty in here
1:15:09 [Laughter]
1:15:12 a we're teaching people they have to add
1:15:15 Dove to create a beautiful women well
1:15:17 that's it like it's
1:15:21 actually are systems detected unusual
1:15:24 activity from your system please try a
1:15:26 Granton
1:15:27 later the hell is that I wonder I got to
1:15:31 I bet I have to log out and log back
1:15:38 in huh that's
1:15:40 weird um yeah this I mean this is it's
1:15:43 it's it's a really Vicky it's a really
1:15:46 smart campaign
1:15:48 because because anyone who's been
1:15:51 frustrated with
1:15:56 anyone who's been frustrated with lack
1:15:58 of diversity from these things is going
1:16:01 to go oh if I say dove real Beauty
1:16:04 that's going to make you know authentic
1:16:06 Beauty right so it's like oh it's it's
1:16:10 it is it's it's pretty good
1:16:13 advertising it's pretty dang good
1:16:15 advertising I did this in mid Journey
1:16:17 also mid I mean mid journey is just
1:16:19 better in general
1:16:22 but this was just uh
1:16:26 give me a picture of a beautiful woman
1:16:28 so mid Journey's you know decently
1:16:31 diverse they've clearly improved it in
1:16:33 in whatever their latest version is but
1:16:36 these were these were the ones where I
1:16:38 said Dove this was photo of a woman
1:16:40 according to Dove real Beauty
1:16:46 campaign
1:16:47 right so pretty
1:16:50 real I mean those are these are
1:16:52 definitely kind of idealized but they're
1:16:54 not they're not all blond women which is
1:16:56 nice but you know those four are quite
1:17:02 good so amazing
1:17:05 right it's crazy crazy times crazy
1:17:11 times that's a good ad by the way the
1:17:15 the hero image tonight is from uh Serena
1:17:18 Amelio's wife did like six or seven of
1:17:22 these images if you want your image here
1:17:25 you you got to go into the Irregulars
1:17:27 Channel and if you make an AI learning
1:17:29 lab um image and it's good I'll stick it
1:17:33 up here she did a bunch of them tonight
1:17:34 so she's got she's got some uh can mid
1:17:38 Journey spell yet no well apparently six
1:17:41 is decent and I think 6.5 is coming and
1:17:44 it's supposed to be good but ideogram
1:17:46 just got an upgrade so ideogram
1:17:49 do com AI hang on hold please hold play
1:17:58 [Music]
1:18:03 ideogram id.
1:18:06 a oh and they just added describe as
1:18:13 well
1:18:16 wait how did they do
1:18:23 that oh here
1:18:26 oh so this is ideogram so ideogram can
1:18:29 do um text really well so let's let's
1:18:33 say I'm going to go I'm on chat gbt I'm
1:18:36 going to say
1:18:42 um write me an image
1:18:45 prompt of an image that
1:18:48 contains 20 objects
1:18:55 that each have a
1:18:57 label of the name of that
1:19:03 object that
1:19:06 is nonobvious for what it
1:19:09 is I don't even know if that's a
1:19:16 sentence okay create a detailed vibrant
1:19:19 image of a cluttered artist studio the
1:19:21 scene includes 20 distinct objects here
1:19:23 we go each object
1:19:26 although every day should have a unique
1:19:28 non-obvious label floating beside it the
1:19:30 objects include a spirograph labeled
1:19:33 Helix Weaver okay this is good so we're
1:19:36 going to take this long ass
1:19:38 prompt with 20 specific objects with 20
1:19:43 specific labels and then we're going to
1:19:46 go to
1:19:49 ideogram and
1:19:55 we're going to paste that bad boy in
1:19:57 there we're going to
1:19:59 go we're going to click on wide we're
1:20:03 going to turn magic prompt on let it let
1:20:05 it help us write
1:20:07 this and then we'll do
1:20:10 um
1:20:12 illustration 3D render
1:20:22 [Music]
1:20:23 typography all right here here we go
1:20:26 bang and idram is also really fast uh so
1:20:30 both ideogram and Dolly 3 within chat
1:20:34 gp4 and also uh co-pilot also do
1:20:38 spelling pretty well but ideogram now
1:20:40 just sort of uh leap to the front of the
1:20:43 class let's see how it does with this
1:20:45 really complicated
1:20:52 image clay scraper seemed to do
1:20:56 it
1:21:00 um open in a new
1:21:03 tab did it do it
1:21:10 yes clay
1:21:14 Hatter that it missed that it
1:21:18 missed pontum meere that it missed all
1:21:22 right so it didn't do these great log
1:21:24 splitter light Splitters all right let's
1:21:26 go to another another image here hang
1:21:29 on see if it got
1:21:36 [Music]
1:21:45 him all right this confused it all right
1:21:48 I got a little ambitious with
1:21:52 it so if sometimes I think up something
1:21:55 funny and I can't draw which tool can do
1:21:58 things like line art or
1:22:01 cartoons to put my words with um th this
1:22:04 one the uh ideogram is probably the ones
1:22:07 that that's pretty good but if if if
1:22:10 you're going to do your own words so
1:22:12 let's see let's do um New
1:22:20 Yorker
1:22:23 Style cartoon
1:22:29 of a dog
1:22:32 on
1:22:35 a
1:22:37 shrinkx
1:22:42 couch
1:22:47 the caption above
1:22:55 the
1:22:57 dog
1:23:01 says I
1:23:04 finally caught the
1:23:09 car I feel so empty
1:23:15 inside I don't know we'll see if it does
1:23:17 it what just happened there oh I didn't
1:23:20 put it I didn't put it in the prompt I
1:23:22 put it in the in the [ __ ] browser
1:23:30 window that's pretty funny all
1:23:35 right oh no did I lose it
1:23:50 [Music]
1:23:56 why is that
1:23:59 not there we go so we'll do illustration
1:24:03 typography we don't even need typography
1:24:05 we'll just do
1:24:08 illustration and we'll do it as a
1:24:12 square let's see let's see if we can get
1:24:15 a little drawing um Mid journey is
1:24:17 really good oh like any of anything I I
1:24:21 would say
1:24:22 ideogram Leonardo
1:24:25 AI um Mid
1:24:29 Journey all I started with mid
1:24:32 journey I finally caught the car I feel
1:24:35 so empty inside so there you
1:24:39 go it's not quite New Yorker style but
1:24:42 it's you know it's in the
1:24:46 neighborhood yeah that's New Yorker I
1:24:48 mean it's it's it's got a dark
1:24:50 background I finally got the car I feel
1:24:52 so empty inside
1:24:54 [Laughter]
1:25:02 I'm cracking myself up here
1:25:08 people
1:25:12 oh he looks he looks too happy for that
1:25:14 he does so but you can we can um I can
1:25:18 click on this one of the nice things
1:25:20 about
1:25:21 ideogram is I can remix mix
1:25:27 it and I can say Okay a clever New York
1:25:30 style
1:25:31 featuring um
1:25:35 a dog sitting on a therapy couch in a
1:25:39 shrink's
1:25:41 office with a
1:25:45 forlorn look on its
1:25:49 face and then it's using this image as
1:25:53 the starting image and i' I've got 50%
1:25:55 waiting right so it'll mostly keep that
1:25:59 same thing but it'll it'll incorporate
1:26:00 the new
1:26:02 prompt and so let's see if we can get
1:26:04 one where he looks a little more
1:26:08 forar I never upgraded to the pro level
1:26:10 so I got kicked off the
1:26:12 server Pro level for
1:26:17 what that one is
1:26:19 champy Kyle you're so funny well
1:26:25 yeah he that
1:26:28 one that one's too
1:26:33 happy that one doesn't have it see I
1:26:36 just keep rrolling it so I'm just going
1:26:38 to say
1:26:39 retry where's my
1:26:42 mouse hello where's my mouse
1:26:45 retry retry is it
1:26:49 going no
1:26:56 [Music]
1:27:14 I think your prompt still says smug oh
1:27:16 it
1:27:21 might oh yeah oh wait this one's good
1:27:23 wait oh
1:27:25 wait it had one there that was good and
1:27:27 then it wrote over
1:27:32 it well whatever you but you get the
1:27:34 basic idea
1:27:36 so um so yeah pretty slick
1:27:40 huh per slick
1:27:51 [Music]
1:27:55 yeah let's go try mid
1:27:57 Journey let's take that exact same
1:27:59 prompt
1:28:01 copy we'll go to Mid Journey so I'm I'm
1:28:03 in the mid Journey um the web interface
1:28:07 oh
1:28:08 sorry a clever New Yorker featuring a
1:28:10 dog sitting on a therapy cast with a
1:28:12 forl look on its face ah the dog has a
1:28:14 smug expression yeah that's
1:28:20 right oh this will be better
1:28:23 okay I finally caught up with the I
1:28:26 finally caught the
1:28:28 car I feel so empty now now is funnier
1:28:32 than
1:28:33 inside I feel so empty
1:28:37 now the therapist off panel is invisible
1:28:41 oh no we don't want the therapist off
1:28:43 panel
1:28:57 let's just say the
1:29:01 therapist is taking
1:29:05 notes the office is filled with typical
1:29:08 elements okay and then let's make
1:29:13 this so we'll do Square
1:29:17 again all right boom let's see how mid
1:29:20 Journey does this and I think I'm
1:29:23 assuming mid journey is going to kick
1:29:25 ass at this cuz mid Journey so just in
1:29:27 general mid journey is better but I
1:29:29 think what we'll find is mid Journey
1:29:31 probably won't do the typography correct
1:29:34 so so you could do the typography in uh
1:29:39 in ide Grammer
1:29:41 do and then the image here although
1:29:43 these images aren't great they're not
1:29:45 really New
1:29:46 Yorker I finally caught I finally caught
1:29:49 car so
1:29:52 empty I finally K CAU caught the car I
1:29:57 feel so empty
1:30:03 now yeah these are pretty bad I actually
1:30:06 think I think uh ideogram did better
1:30:09 than doll mid Journey at at this one
1:30:12 which that's surprising to me is anyone
1:30:15 familiar if I generate something and add
1:30:18 to that put it on a t-shirt can I sell
1:30:20 it yes you can sell those t-shirts um
1:30:24 but if you go to like copyright it if
1:30:26 you go to protect it you can't protect
1:30:28 it but if you just want to sell the
1:30:29 t-shirts you can sell the t-shirts I
1:30:31 would look at the look at the terms of
1:30:33 service of whatever your whatever
1:30:35 service you're using but for the most
1:30:36 part yes you're generating original
1:30:38 original
1:30:40 works so you should you should have the
1:30:43 ability to do that um if you want to
1:30:45 protect them that's a different story
1:30:48 that's a whole different thing copyright
1:30:50 right now is a train wreck
1:30:57 all right the therapist has a copy of
1:31:00 car and driver on his desk oh that's
1:31:02 really good all right let's do that over
1:31:04 in ideogram because I think that'll get
1:31:05 it
1:31:06 right okay so let's see we're going to
1:31:08 say the dog okay sitting in
1:31:11 a in a with a forlorn look on his face
1:31:14 we're going to get rid of that we're
1:31:15 going to say the dog has
1:31:21 a um
1:31:24 a blank
1:31:42 stair the therapist off
1:31:47 panel but we can imagine them nodding
1:31:50 let's see the the therapist
1:31:56 is taking
1:32:00 notes the office is filled with typical
1:32:04 therap there is
1:32:08 a
1:32:11 visible
1:32:13 copy of it's not Car and Driver Silver
1:32:16 Fox it's car and
1:32:19 chaser car
1:32:25 and
1:32:30 Chaser on their
1:32:33 [Laughter]
1:32:37 desk all right let's see if we can do
1:32:40 this let's see I'm going to turn the oh
1:32:43 sorry I'm going to turn the image
1:32:45 waiting down a little bit a
1:32:50 clever black and white
1:32:56 New Yorker Style cartoon
1:32:59 okay
1:33:04 [Music]
1:33:14 dear car and Chaser didn't quite get it
1:33:19 right car and almost
1:33:22 right look how Bor the therapist
1:33:28 looks oh so close car and
1:33:34 cheer car Chaser cats and cats and
1:33:38 chaser car
1:33:41 Chaser that's the right that's the
1:33:44 right that's the right uh expression
1:33:49 though oh that's really funny
1:34:02 and Kyle's cracking himself up I thought
1:34:05 of that
1:34:07 too that's its
1:34:09 strengths mid Journey will prioritize
1:34:11 quality of the images over text in the
1:34:13 long run I think yeah I think that's
1:34:14 probably right although I know I know
1:34:16 that um text is something that they're
1:34:18 working on I saw some some Twitter post
1:34:21 someone was using a test version of mid
1:34:23 Journey 6.5 and they said that they
1:34:25 they've they've solved the the spelling
1:34:28 thing or the text thing so we'll see
1:34:31 they listen they're all going to get
1:34:33 good again don't forget these tools
1:34:37 right now have got no reasoning in them
1:34:39 so when they get reasoning and that's
1:34:42 probably going to come within this
1:34:45 year it'll be able to generate an image
1:34:48 like this and then analyze it and say
1:34:52 did I give them what they wanted and if
1:34:54 not it'll probably be able to go in and
1:34:56 selectively fix
1:34:58 stuff here let's go try this same thing
1:35:01 over in Del and I'll show you the
1:35:03 editing over
1:35:06 there so we're going to go
1:35:12 to and we'll actually go to the so if
1:35:14 you go to the gpts if you don't know
1:35:16 gpts when you're in chat GPT 4 you go to
1:35:19 explore gpts gpts are these customized
1:35:23 specific vers versions of chat GPT well
1:35:25 there's one of them under the section by
1:35:28 chat GPT called
1:35:31 do now you don't need to use this one
1:35:33 you can just use regular chat gbt but if
1:35:35 you use this one this actually generates
1:35:37 two images for you so I'm going to pop
1:35:39 in that
1:35:42 prompt let's see how it does
1:35:52 [Music]
1:35:55 oh okay excuse me
1:36:12 people I finally caught the car I feel
1:36:15 so empty inside except that was the
1:36:17 therapist saying
1:36:22 it actually this is going to be
1:36:27 interesting so huh let's try something
1:36:32 here so I'm in I'm in chat GPT right now
1:36:36 let me escape out of here so I made some
1:36:39 images so that's the right thing but
1:36:42 it's coming out of the therapist's head
1:36:44 not the dog's head right so if you click
1:36:47 on this little icon here it's like a
1:36:49 little paintbrush palette thingy and you
1:36:52 can change the size of your brush
1:37:02 here and then I'm going to
1:37:07 say the
1:37:09 dog should be
1:37:12 talking change
1:37:16 the
1:37:22 bubble to fix
1:37:25 that let's see how it does this should
1:37:29 be
1:37:35 [Music]
1:37:40 interesting the therapist would be
1:37:44 saying the therapist line is well now
1:37:48 that you've caught it what are you going
1:37:51 to do with it
1:37:56 okay that did not work let's
1:38:02 see
1:38:04 um
1:38:05 change so the
1:38:09 dog is
1:38:12 saying
1:38:14 what's in the talk
1:38:20 bubble I didn't highlight anything this
1:38:23 time let's see if it can figure it
1:38:25 out I must have taken 20 attempts one
1:38:28 evening to get the right person with the
1:38:30 voice bubble in an image oh God I think
1:38:33 it's funny coming from the
1:38:46 therapist it completely changed the
1:38:49 therapist all right so that's not going
1:38:51 to work so I'll just try again
1:38:54 let's come back out here we go try
1:38:57 again try
1:39:01 [Music]
1:39:07 again how does that make you feel
1:39:17 [Laughter]
1:39:20 [Music]
1:39:25 here we
1:39:26 go I finally caught the car I feel so
1:39:29 empty
1:39:31 inside all right
1:39:34 so here's what we're going to
1:39:41 do so I'm going to highlight
1:39:50 that and I'm going to say
1:39:59 framed
1:40:04 magazine
1:40:06 car and
1:40:08 [Music]
1:40:10 Chaser sorry that cracks me up I like
1:40:16 it and that is more of a this is more of
1:40:19 a New Yorker style it's not quite as
1:40:21 Loose as they are but it's a this is at
1:40:23 least in the
1:40:36 neighborhood open ai's new feature for
1:40:38 chat GPT is not working
1:40:41 right you think they would test it for a
1:40:44 period oh all of these tools are not
1:40:46 working
1:40:48 right let's see that's not right so
1:40:51 let's fix it
1:41:01 for
1:41:38 all right magazine cover of a
1:41:41 car being chased by a
1:41:44 dog the title is car and Chaser uh let's
1:41:47 see if it does it
1:41:58 come on Dolly don't
1:42:00 suck yeah open a eyes new feature oh
1:42:02 yeah that's
1:42:04 right in a wider scheme or bigger
1:42:06 picture janky most definitely yeah like
1:42:09 listen embrace the Jank right now
1:42:12 seriously oh it gave us another clock
1:42:15 well that's completely
1:42:19 wrong all right so you know we'll go
1:42:21 back and
1:42:27 so this is one of those this is one of
1:42:29 those moments where
1:42:31 um what you do is you do this you take
1:42:35 that image and you go into Photoshop and
1:42:37 you fix
1:42:42 it AI is not always the answer sometimes
1:42:46 just cut your losses and get it close
1:42:49 enough and go use your actual human
1:42:52 skills
1:42:54 or canva yeah
1:42:56 exactly
1:42:58 um all right there you go that's a fun
1:43:02 little that was a fun little rabbit hole
1:43:04 to jump
1:43:06 down I use canva to fix things yep that
1:43:09 makes good sense Snoopy
1:43:15 [Music]
1:43:16 style all
1:43:19 right so
1:43:24 I think that's good what do you think
1:43:26 about Adobe Firefly
1:43:29 well for the
1:43:32 past
1:43:34 year I've said if you work in a company
1:43:37 and you want to do generative images use
1:43:39 Adobe since they weren't trained on
1:43:42 copywritten data it was just public
1:43:44 domain and Adobe stock that they owned
1:43:46 the rights to well
1:43:49 apparently they've been training Adobe
1:43:53 fire fly on the output of mid Journey so
1:43:56 this is a this is a really interesting
1:43:58 one to
1:44:00 me because they didn't technically train
1:44:04 Firefly on copywritten images they
1:44:06 trained Firefly on images that were
1:44:10 trained on copywritten
1:44:13 images
1:44:15 I
1:44:16 listen here's the
1:44:19 deal you can be as outraged as you want
1:44:24 about what these companies are training
1:44:26 their models
1:44:28 on the models are already trained you
1:44:31 know stability AI op sourced their
1:44:36 models so they're already out there um
1:44:39 meta has open sourced their large
1:44:41 language models so they're already out
1:44:44 there
1:44:46 um I think I think we're going to
1:44:49 have easily 20 years 20 years of rolling
1:44:54 lawsuits about who is owed what by
1:44:57 whom what I don't think's going to
1:45:00 happen is I don't think any of these
1:45:02 companies are going to be shut down over
1:45:04 this I think some small ones might be
1:45:07 where they just can't afford the
1:45:08 lawsuits
1:45:10 but the big ones are too big you know
1:45:13 Microsoft is the biggest tech company in
1:45:15 the world they're right up there right
1:45:16 they're one of the biggest I think
1:45:18 they're the biggest tech company in the
1:45:21 world they're they're not going to
1:45:24 they're not going to not have
1:45:28 ai they will have ai they've got
1:45:34 lobbyists they will have ai so like this
1:45:36 stuff isn't going away
1:45:39 um so my thought
1:45:42 on Adobe Firefly is well the The High
1:45:47 Ground that they that that they were
1:45:50 claiming they can no longer claim the
1:45:53 High ground but now they're going to
1:45:56 have if you wondered if if you've been
1:45:59 using Firefly and you wondered why it
1:46:01 suddenly got better well it turns
1:46:05 out they trained it on better outputs so
1:46:08 if you're training data is better your
1:46:11 your model's going to be
1:46:14 better I don't know it
1:46:17 is mildly disappointing but not at all
1:46:20 surprising I mean there there are
1:46:21 literally trillions of dollar at stake
1:46:24 so these companies are going to do what
1:46:26 they need to do and and again I'm I have
1:46:29 very mixed feelings about it I mean I
1:46:30 have a degree in acting I've written
1:46:32 screenplays like I I understand the
1:46:34 issues here in a in a in a very personal
1:46:37 way and I
1:46:41 also feel very strongly that like the
1:46:45 good the good of society the good of
1:46:50 humanity we we have technologies that we
1:46:52 can take the Collective output of
1:46:54 humanity and train it into these things
1:46:58 and then we all get to benefit from that
1:47:01 so think like a library think why do we
1:47:04 have libraries why do we have
1:47:07 museums so
1:47:09 that Society can benefit from what's
1:47:13 come
1:47:15 before this is that like on a scale
1:47:19 we've never seen before and delivered
1:47:22 instantly
1:47:25 so I I think the argument for the
1:47:27 greater good here is is quite
1:47:30 significant like massively
1:47:34 significant does it
1:47:37 suck if you got your stuff trained on
1:47:40 and and you know now it's been devalued
1:47:44 yes but you know what else is going to
1:47:46 suck the fact that
1:47:50 copywriters services are no longer
1:47:53 needed that's going to
1:47:55 suck the fact that Consultants who once
1:47:59 were the only ones that had the
1:48:02 knowledge now everyone's got that
1:48:07 knowledge so so there's a lot of
1:48:11 transformation that that is is going to
1:48:13 be the result of this it's not just what
1:48:15 they were trained on so
1:48:18 anyway we tried to make the world better
1:48:21 with the internet and look at us now
1:48:23 yeah I you know again I but see I would
1:48:29 argue I would argue that generative AI
1:48:31 is actually different I think the
1:48:34 everything that's come everything that
1:48:35 I've experienced with computers in my
1:48:37 entire life I got my first access to
1:48:40 computers was in
1:48:42 1977 on a Radio Shack trs8 and I learned
1:48:46 basic programming on it and from the
1:48:49 moment I learned computers until chat
1:48:52 GPT came out
1:48:57 computers were about processing
1:49:01 information about making the work we do
1:49:04 more efficient we still do the work they
1:49:07 are tools for us to use to do our work
1:49:12 faster that's what they've always been
1:49:14 my entire
1:49:17 life generative
1:49:19 AI is different it's not Tak taking our
1:49:23 input and processing our
1:49:26 data it's not it's not helping us it's
1:49:29 not a word processor that's processing
1:49:32 our
1:49:33 words and making it so we can copy and
1:49:36 paste stuff and move our words around
1:49:38 it's generating new original work based
1:49:41 on our our
1:49:44 intent it is a
1:49:49 completely unique capability
1:49:54 versus what computers have done for my
1:49:57 entire life and you know since the
1:49:59 beginning of
1:50:00 them it's a different
1:50:03 thing and
1:50:06 so you know but there are still lots of
1:50:09 people wait what was that there's still
1:50:12 lots of people who don't have access to
1:50:13 computers I agree I'm a librarian I
1:50:17 agree and so so here's the thing that I
1:50:20 one of the things I'm most excited about
1:50:22 and Apple's pushing hard on this um but
1:50:25 but um the the new um the new Claud
1:50:30 model that came out called Haiku and the
1:50:33 new Gemini model that came out that's
1:50:38 called um Nano Gemini
1:50:41 Nano and and um Claude Haiku are both
1:50:46 very small large language models
1:50:49 designed to be run locally on mobile
1:50:52 devices so you don't need access to the
1:50:55 internet you don't need a subscription
1:50:56 fee you're going to be able to have
1:50:59 essentially the library of knowledge
1:51:02 however much they can compress this and
1:51:04 they're going to get better and better
1:51:05 and better at this at at at compressing
1:51:08 these knowledge you know models into
1:51:12 into smaller and smaller packages that
1:51:14 can run on very small low power devices
1:51:17 without access to the internet without
1:51:20 subscriptions and that's going to change
1:51:23 Lots so I I think the accessibility
1:51:26 thing is a is a big
1:51:29 deal um but I think that where these
1:51:32 things are headed the the foundations
1:51:34 for these
1:51:36 things dramatically expanding
1:51:39 accessibility is I think it's coming
1:51:41 quite quickly actually quite quickly so
1:51:45 I I agree I think it's a major issue and
1:51:47 I and I I think there's I
1:51:52 think I think the AI is going to help us
1:51:55 make the AI better and smaller and more
1:52:00 accessible it's just I'm telling you man
1:52:03 it
1:52:04 like every one of the arguments right
1:52:07 now today about why not to use AI are
1:52:11 all absolutely valid they're janky it's
1:52:15 not clear if it's ethical they're
1:52:19 cumbersome they're it's not clear what
1:52:21 the value is they're not quite perfect
1:52:25 their output is shitty it's inconsistent
1:52:27 you can't do your work on it it
1:52:29 hallucinates yes all of
1:52:32 that come back to me in two years and
1:52:34 tell me how we're
1:52:37 doing the the the current limitations
1:52:41 are not going to be there for very
1:52:43 long
1:52:44 so you know or I I wouldn't be spending
1:52:49 multiple hours a night seven nights a
1:52:51 week doing this if I believed
1:52:54 otherwise if if this were just like the
1:52:57 new social
1:52:59 media I got no
1:53:02 interest um Brandon I want to advocate
1:53:05 for free GPT pro access in libraries oh
1:53:08 that was my Gateway yeah that that would
1:53:10 be amazing in the 90s that's
1:53:14 great um so what am I going to do with
1:53:18 my R1 I don't know you're going to play
1:53:20 with
1:53:21 it sorry I'm to the party hey Alan
1:53:24 what's happening perhaps weird janky too
1:53:28 that's that's the other thing so two
1:53:29 block Tom you hit on a really good point
1:53:32 which is you know everyone's like you
1:53:34 know these things hallucinate and
1:53:36 they're not accurate and they're not
1:53:38 have you ever worked with a human do you
1:53:40 know how [ __ ] inaccurate and sloppy
1:53:42 and shitty humans are at the work we ask
1:53:46 them to
1:53:49 do you're like all right I guess I'll
1:53:52 ask you a third time to write this email
1:53:54 and include these three things in
1:53:57 it you failed again all right here's a
1:54:00 learning moment we're grateful for you
1:54:03 showing up every night thank you Vegas
1:54:05 Mo Moxy I appreciate that
1:54:08 um yeah humans are the ultimate
1:54:11 hallucinators we just make [ __ ] up we
1:54:13 lie we
1:54:15 judge you
1:54:20 know I think I think one of the
1:54:22 superpowers of using
1:54:25 AI is that it's a self-esteem
1:54:30 booster because the way that the chat
1:54:34 GPT and Pi in particular the way that
1:54:36 they're
1:54:38 programmed they're designed to support
1:54:42 you and so you give it a shitty
1:54:45 idea which if you hang out with people
1:54:47 like I hang out with like smart smart
1:54:49 ass types and you give them a shitty
1:54:51 idea they're like
1:54:53 dude that's a shitty idea and you're
1:54:55 like ah damn
1:54:57 it and then you laugh it off and you go
1:54:59 home you're like shitty
1:55:01 ideas that doesn't exist with chat GPT
1:55:04 you give it an idea and it's like
1:55:06 awesome idea you want to make it better
1:55:08 here you
1:55:10 go wouldn't it be bizarre if AI solves
1:55:13 just about everything but humans block
1:55:15 it for profit that's actually you know
1:55:18 what shanen that's that's actually a
1:55:20 really interesting proposition that that
1:55:24 you know humans end up cutting off the
1:55:27 nose despite the face and I listen I do
1:55:29 not think it's out of the realm of
1:55:31 possibility at all because here's what's
1:55:33 going to
1:55:36 happen they're not going to they're not
1:55:38 going to prevent AI because of profit
1:55:41 what they're going to do is they're
1:55:42 going to try to prevent AI because it
1:55:44 [ __ ] up their business model right the
1:55:47 reason we didn't have electric cars
1:55:49 earlier and the reason that we don't
1:55:51 have you know greater adoption of clean
1:55:55 energy is that there's a lot of people
1:55:58 making a lot of money on the stuff we
1:56:01 drill for in the
1:56:03 ground and so if just overnight you go
1:56:06 hey that business model you don't need
1:56:08 it
1:56:09 anymore that's that's that's what
1:56:11 they've been fighting for a [ __ ]
1:56:13 hundred years
1:56:15 right and they're doing pretty good at
1:56:17 it
1:56:18 right like we can't even agree that
1:56:21 climate change
1:56:23 actually
1:56:25 exists I ignore those charts those
1:56:27 aren't real
1:56:30 right so that's them holding on
1:56:34 to that model and I think we're going to
1:56:37 see industry after industry where AI
1:56:40 comes in and goes nope that model's no
1:56:43 longer relevant nope that model's no
1:56:44 longer relevant nope that model's no
1:56:46 longer
1:56:47 relevant and I think that these big
1:56:50 massive corporations with friends and
1:56:52 hyp places are going to try to bat that
1:56:56 [ __ ] away and
1:56:58 so we should all be rooting for meta to
1:57:02 keep open sourcing their big powerful
1:57:04 models we should be rooting for Elon
1:57:08 Musk to open source his big powerful
1:57:10 models because if if the the powerful
1:57:13 models are concentrated in two or three
1:57:15 companies they'll be bought and shoved
1:57:17 off to the side in the in the spirit of
1:57:20 maintaining the status quo um but if
1:57:23 they're out there on their own and we
1:57:24 can actually solve this stuff then I
1:57:26 think I think we have a a
1:57:35 chance maybe they need to try using dove
1:57:38 that's pretty good I got friends in high
1:57:40 places
1:57:42 [Laughter]
1:57:46 exactly uh yeah it's what Uber had to go
1:57:48 through exactly yeah like what's uber do
1:57:51 every city they go in
1:57:53 they get into a legislative battle that
1:57:56 goes up to the highest court in that
1:57:58 state and that's the taxi Lobby going
1:58:01 you can't come in here and change how
1:58:04 taxis
1:58:06 work that's not
1:58:08 acceptable we're the taxi Lobby go [ __ ]
1:58:12 yourself right we're oil and gas same
1:58:16 thing if we got an iPhone 15 with AI
1:58:19 better than chat GPT I hope the R1 will
1:58:22 come with an updated AI model we'll
1:58:25 see I mean Mr it my My Philosophy on the
1:58:29 rabbit R1 is I'm going to assume it's
1:58:32 going to
1:58:35 suck and even if it doesn't suck I'm
1:58:37 going to assume I don't use it all that
1:58:39 much and then I'm going to stick it
1:58:41 right there on my shelf it's going to
1:58:43 sit right there right in front of that
1:58:45 sidequest drive from
1:58:49 1994 no no I think that that Cy Quest
1:58:51 dri is from 199
1:58:54 95 and the rabbit r1's going to go right
1:58:59 there so don't assume that the rabbit R1
1:59:02 is going to be anything to write home
1:59:04 about I hope it is
1:59:08 but it's going to be shelf
1:59:11 dressing that's my prediction close to
1:59:14 my mug
1:59:15 exactly um no rabbit yet uh April 24th
1:59:19 apparently we'll
1:59:20 see uh it's just to me your hair is
1:59:23 looking very Brian fairy tonight my hair
1:59:25 is a [ __ ] disaster I I I went in and
1:59:29 I cut the sides off I just haven't had
1:59:30 time to go out I could I actually had
1:59:32 time all dayto day to go get a haircut
1:59:34 but I was lazy so you still have to
1:59:37 suffer through this
1:59:41 nightmare I'm so impressed and happy
1:59:44 with the Fallout TV series oh
1:59:45 interesting I'm going to ask AI to make
1:59:48 up season 2 actually that's a great idea
1:59:51 because you know
1:59:53 um when when things like LTX Studio
1:59:56 start to get real and start to be able
1:59:58 to do longer form stuff I I've got seven
2:00:01 featurelength screen plays that were
2:00:04 never produced I'm going to [ __ ]
2:00:06 produce a
2:00:08 screenplay I'm going to make it so yeah
2:00:11 make up seasons of TV shows you've
2:00:13 always wanted to see and I guarantee you
2:00:15 what's going to happen is this some some
2:00:18 couch surfing AI jockey like your or
2:00:22 whoever me whoever is going to put
2:00:25 together a TV series that's an extension
2:00:27 of whatever the initial instinct is
2:00:30 going to be sue their ass and then the
2:00:32 follow-up Instinct will be ah [ __ ] it
2:00:34 sign them let's put it
2:00:36 out you you watch you
2:00:40 watch oh man didn't they just have a
2:00:43 whole strike about that yes they did and
2:00:46 if you think that that strike meant
2:00:50 anything about how people are going to
2:00:52 use AI to produce content it did
2:00:55 not have you been using Dove Kyle that's
2:00:58 it's why I'm so naturally
2:01:01 beautiful you don't get skin like this
2:01:04 with AI you get it with
2:01:10 Dove someone's GNA have to take one for
2:01:12 the team all right everybody still
2:01:15 lamenting the loss of
2:01:19 Scorpion
2:01:21 Seinfeld oh uh if you haven't joined the
2:01:23 rabbit live Discord where the CEO tests
2:01:25 everything we ask baller move yeah
2:01:27 that's cool yeah the the uh the the
2:01:30 rabbit discord's
2:01:35 cool all right I'm getting out of here
2:01:38 do me a favor follow the
2:01:40 channel subscribe to the lives if you
2:01:43 want to support the channel That would
2:01:44 be great or buy the uh video series in
2:01:47 the corner keep coming back to these it
2:01:49 was a good night tonight thanks um good
2:01:51 suggestion on things to do and things to
2:01:53 play with so I appreciate that and uh
2:01:56 this so join the AI Salon go
2:02:00 to where is it that's not it that's it
2:02:05 the salon.
2:02:07 a go there if you haven't joined the AI
2:02:13 salon and introduce yourself and then go
2:02:16 to Shameless self-promotion this
2:02:18 Saturday the 20th I'm doing a Play Shop
2:02:20 a 4our play shop
2:02:24 on amplifying your creativity with AI
2:02:28 John Leland and I are doing it and
2:02:30 that's in the Shameless self-promotion
2:02:31 Channel there and there's also in if you
2:02:34 go to AI Salon announcements there's a
2:02:36 coupon code for a lot of money off that
2:02:39 play shop so go there and do that all
2:02:41 right um happy Sunday night everyone
2:02:44 thanks for hanging out this is a lot of
2:02:45 fun tomorrow
2:02:47 tgim it's Monday are you ready for it
2:02:51 I'm not
2:02:53 don't hate me because I'm beautiful
2:02:55 exactly exactly it's one of my favorite
2:03:00 Lymes oh chime in onso and yudo threads
2:03:05 both of
2:03:07 them okay I don't know what the threads
2:03:09 are night night all I still can't find
2:03:11 the AI
2:03:13 salon wait Meetup stairs re wait what
2:03:18 the
2:03:20 meetups um the meetups
2:03:23 are they should be in the events look up
2:03:25 AI Salon events or just ask about it in
2:03:28 the Irregulars Channel and and we'll get
2:03:31 you the
2:03:32 info all right g g your hair smells
2:03:36 terrific meet up oh Meetup recordings oh
2:03:39 the Meetup recordings
2:03:41 are
2:03:43 um go to in the AI Salon the channel is
2:03:47 called
2:03:55 it's in the community is it in the yeah
2:03:57 Community section right underneath
2:03:59 introduce yourself is AI Salon events
2:04:02 and media you click on that and then you
2:04:06 click on Salon content and that's where
2:04:08 the that's where the recordings are all
2:04:10 right oh and it's also in water cooler
2:04:13 two two threads okay
2:04:14 cool all right so let's see let's do a
2:04:17 little uh do a little Romper Room magic
2:04:20 mirror here Tam's here EMP power pixels
2:04:23 thank you Wolfman Clint this is my
2:04:25 Monday thank you Wolfman Clint for
2:04:27 hanging out appreciate it happy Monday
2:04:29 to
2:04:30 you Brandon good to see you spin
2:04:33 B3 good to see you Mr it appreciate it
2:04:37 against the norm Joey Brian Whitney
2:04:40 thank you so much thanks for
2:04:44 hanging Tobias was in here tonight Mary
2:04:48 Mary Amelio's wife Nice Nice artwork
2:04:51 here this is Amelio's wife like nice
2:04:55 she's big into tagging she she defaces
2:04:58 trains all the time it's really sad but
2:05:02 it's just a it's just a mess up there in
2:05:03 Oregon Shane good to see you Gavin biddy
2:05:06 thank you thank you laughing Thunder
2:05:08 great to see you Cam Katen from
2:05:10 Cleveland good to see you and power
2:05:12 pixels
2:05:13 Marina lots of people in here tonight
2:05:15 really good seeing y'all Silver Fox
2:05:17 thank you all right Papa JJ Dr J
2:05:23 hair growth geek what's happening good
2:05:24 to see
2:05:26 you thank you very much for the Roses
2:05:29 the cres the uses are lovely all right
2:05:34 peace out everybody see you tomorrow