ChatGPT-a-thon Part 2: 6 AM – aiLL Nov 30, 2023 – The Past, Present and Future of ChatGPT

Description

Chapters

Transcript

0:18 all
0:19 [Music]
0:21 right we are back Patricia
0:25 planks let's see Cuban Rebel Rebel girl
0:29 great name love it Sonia welcome Pate
0:35 shaking bacon Steven Braille good to see
0:41 you let's get in in early you
0:44 [Music]
0:46 make you're up early okay so if you
0:49 don't know
0:51 this this is uh my uh my 24hour
0:57 celebration of
0:59 uh chat gpt's Oney year anniversary so
1:04 so I went I went live last night at 8:00
1:08 till about 10:30 then I went back on at
1:11 midnight until
1:15 3:30 and I'm back so I'll do a couple
1:19 hours this morning then I'm going to go
1:20 to work I'm going to turn on the old Tik
1:23 Tock Clint Copperhead you're a beast
1:27 too
1:29 uh
1:33 um I still I'm not up yet but I do have
1:36 my coffee I do have my my morning coffee
1:39 I slept good I slept really
1:42 good as you can imagine my brain was
1:45 crespy it was
1:47 crispy it was crispy but I'm
1:50 back okay the boss lady Silver Fox was
1:53 giving me crap for not being on I I
1:55 think I said I'd come back on at 5:30 so
1:58 uh so she gave me crap
2:01 about about
2:04 uh is it 700 a.m. where you are it's
2:07 6:00 a.m. where I am I am in Denver so
2:10 I'm in mountain time so I'm two hours
2:12 Sonia you're in New York I think right I
2:14 miss New York I miss New York I miss my
2:18 New York um yeah so so that's it and uh
2:25 we are we
2:26 are happy birthday to chat GPT I think
2:30 think I've got some birthday images here
2:33 there's a good one the birthday
2:36 suit birthday suit is there let me
2:39 actually go look over in the salon
2:42 Discord oh that's a good one let me save
2:44 this let me do a little housekeeping
2:49 here let's see who's that that's from
2:53 Tim save image as got
2:58 Tim salon
3:13 bday all right that one we got people
3:17 making fun of Elon
3:20 Musk wow did he have a special day
3:24 yesterday
3:28 woo let's do this save image as who is
3:32 this this is
3:34 Lauren this is a cool
3:42 one
3:44 Lauren muscle car and a
3:48 v bang let's see what else we got we got
3:54 the big old helicopter all right so let
3:57 me go add those to our little Library
3:59 here
4:01 it's how the sausage is made we get some
4:04 images and
4:06 we drop them
4:11 in or do we we
4:15 don't CU apple has a bug in it there's a
4:19 surprise there's a surprise
4:22 people Apple couldn't possibly have a
4:24 bug in their software they're just about
4:26 it just working aren't
4:28 they there we go
4:33 beautiful the AI Salon has a birthday
4:36 coming up next Tuesday so chat gbts is
4:39 today the AI Salon is next Tuesday so
4:43 we'll talk about the AI Salon we'll get
4:45 going welcome everybody if you are new
4:48 to the channel I can't believe there's
4:49 97 people in here this early although
4:51 East Coast this is you're just waking up
4:54 time I'm I'm on your commute to work or
4:57 I'm in your just got to work and I'm
4:59 drinking coffee time
5:05 cheers so my name is Kyle Shannon this
5:08 is the AI learning lab good morning FES
5:11 2000 the FES 2000
5:16 nice oh man Be not Afraid what's shaking
5:20 who else we got in here steo steo is in
5:24 isn't it dark dark dark 30 out where you
5:27 are steo F the advertisers I know
5:31 it wasn't even F the advertisers go [ __ ]
5:34 yourself hi Bob Bob Iger from Disney and
5:38 then he basically said you
5:40 know I've singlehandedly done more good
5:43 in the world than than anyone and he
5:46 goes you know I don't want um I don't
5:51 want people who are telling me to do
5:53 better but then are doing evil behind
5:55 the scenes it's like
5:56 woo that was crazy 11:00 p.m. Steve
6:00 that's that's my that's my neck of the
6:02 woods time
6:04 [Music]
6:06 wise all right it's going to take a
6:08 while for the voice to come the voice
6:10 isn't quite there yet but you know we'll
6:12 get it there I just took some vitamins
6:14 they'll kick in in a bit coffee will
6:15 kick in in a bit we'll be good we'll
6:18 we'll limber
6:22 up all right so what I'm going to do
6:25 today um all day is either talk about
6:29 chat GPT or you know put it in context
6:32 or show it demo it we made a
6:37 uh we made a what are they called it's
6:41 going to come to me we made a GPT last
6:44 night um called Altman's Odyssey which
6:49 is
6:49 a um tell you everything you want to
6:53 know about open Ai and chat GPT
6:55 GPT I couldn't name it GPT anything
6:59 because you know they don't like that so
7:01 they block you from saving things as GPT
7:03 anything U or chat GPT anything I
7:06 finally got to drink coffee with Kyle
7:08 hey Tobias cheers buddy Tobias is our
7:12 next sponsor for the AI Salon cheers to
7:14 you buddy thank you very much um that's
7:18 amazing thank you it's that's super
7:20 exciting um I'll talk about the salon
7:23 here in a
7:25 second and we'll get you up and running
7:27 do me a favor if you would share the
7:30 live I'm not normally live this time of
7:32 the day um looks like we have 75 people
7:35 in here or so so welcome welcome welcome
7:38 um if you have questions about AI pop
7:40 them in the comments below um you can
7:43 probably tell there's some people here
7:45 who've been here before we call them the
7:48 Irregulars because they are they're not
7:50 weird they're irregular
7:53 and they're awesome and I hope I hope
7:58 you enjoy this so so I tend to go live
8:01 uh Once a Night seven nights a week for
8:04 about two to three hours starting around
8:05 8:00 my time which is Denver so I know
8:08 it's late on the East Coast but but
8:11 that's what I've been doing and uh this
8:14 is a special happy chat GPT birthday
8:17 since it's November 30th and that's when
8:20 in my opinion the world just changed
8:23 wait I saved mine with GPT in the name I
8:26 think I could have put GPT in there it
8:28 was the chat GP it was the combination
8:30 of chat and GPT it didn't like you sound
8:32 like Tom Arnold thanks I sound like Tom
8:36 I sound like Tom Arnold I look like
8:38 Rosie O'Donnell am I just fat Alec
8:44 Baldwin thank you thank you I assume uh
8:47 sounding like Tom Arnold is a good
8:49 thing it's a compliment
8:58 right
9:01 oh never a dull moment I can't share the
9:05 live I'm at work and shouldn't be
9:08 here that's awesome Vicki thank you for
9:12 your clandestine support I I'll take it
9:14 I'll take anything I get forever
9:16 youngish I'm not an irregular but I
9:18 embrace my weird welcome that's that's
9:21 all you need to do here this is this is
9:23 a neuros spicy friendly channel uh I am
9:26 a a laminated card carrying member of
9:28 the ADHD Club digital Gods good to see
9:31 you digital Gods came up with the title
9:34 for an epic poem we wrote last night
9:36 about the Saga at open AI called um
9:40 Altman's Odyssey and the
9:43 Ilia it's beautiful so we posted that on
9:46 the Twitter we tagged Elon but he was
9:49 busy telling Disney to go [ __ ]
9:51 themselves so I don't think he saw it I
9:54 think he's probably dealing with dealing
9:56 with phone calls from other advertisers
9:58 like uh yeah yeah yeah we're we're with
10:02 Bob we'll see we'll see I I just I don't
10:07 know I've been I've been working in in
10:10 uh with Fortune 500 companies for a lot
10:13 of my life and they just not the party
10:16 and
10:19 crowd so you were able to get the words
10:22 exactly as you type them mine always
10:25 change uh which ones like that one I
10:28 didn't make that image that was I think
10:30 Lauren made that image you can you can
10:33 get them there if you uh if you put your
10:36 mind to it so so there was an image I
10:42 created here let me show you
10:45 one this
10:49 one so that
10:52 says it almost says chat
10:56 GPT it says it says C4 C backwards 4
11:03 a
11:04 GPT um that was about that was probably
11:07 my 20th
11:09 attempt to get it to say chat GPT
11:12 spelled out of
11:15 balloons but it's still pretty
11:17 impressive I mean the fact that it it
11:20 got it that close is still pretty
11:22 impressive it's got like all the little
11:23 strings for every one of the balloons
11:25 that make up the letters right it's like
11:28 a physics impossibility
11:30 what's there so there's likely not an
11:32 image that looks like that so that was
11:34 as close as I could get it and and after
11:36 like 20 of those I was I was done Silver
11:40 Fox see weird Cactus exactly at exactly
11:45 well what I did with my what I did with
11:47 my video oh you can't see it but I put
11:50 the the timer um over that so it this
11:54 was a a a graphic I did for the
11:56 background on one of my timer videos so
11:58 I put the timer right there so it looked
12:00 correct in the video
12:03 but this is these are some of the some
12:05 of the tricks we do to make up for the
12:07 janky ass nature of these tools right
12:10 now um uncanny valley we are in it for
12:14 sure we are in The
12:18 Uncanny
12:21 valy um oh this wait I think Lauren did
12:26 this one I'm I'm confused on my artists
12:29 who do we have we
12:31 have yeah Lauren did the the muscle car
12:35 and Tim Tim sheffel did the a salon
12:38 one um that's a good that's a good one I
12:43 like that one you got yourself your
12:46 classic llama your classic turtle and
12:48 your classic
12:51 Hedgehog in the back of the muscle car
12:54 looks like
12:56 a 71 Camaro and ass
13:02 end or something like
13:05 that and it says AI lab irregular come
13:08 on that's good that's solid well it says
13:11 [Laughter]
13:14 irregular very cool image I always ask
13:17 it to Omit all text then I just add it
13:20 after yeah well that's the other way is
13:23 there's always default to photoshop
13:26 that's that's one of the things that um
13:28 is funny about this AI stuff is you'll
13:31 spend an hour trying to get AI to do it
13:33 right and you're like oh yeah I could
13:35 have done that in Photoshop and 3
13:41 minutes but you know that's how we are
13:44 that's how we are us humans we're
13:46 weirdos so okay so if you have questions
13:50 about chat GPT uh pop them in there why
13:53 can't chat GPT do hands and feet artwork
13:57 hey Spike Jones um
13:59 it's they they all of these things have
14:01 gotten better at hands I mean hands for
14:04 for the better part of a year were these
14:06 mangled sort of pretzel
14:08 situations you know with like you know
14:10 seven fingers doing that and
14:14 um they're at least in the neighborhood
14:16 of of being hands now sometimes they're
14:19 way too long or way too short um or
14:22 there's just mysteriously a sixth finger
14:25 so they're getting better um it likely
14:28 has to do with the training set and how
14:30 they were tagged and things like that I
14:32 assume most pictures are not of hands
14:36 specifically um so I would assume that
14:38 they've gone back and they've they've
14:40 augmented training sets to have
14:42 specifically tagged hand things to make
14:44 them better um I don't know excuse
14:49 me the nature of how diffusion models
14:52 work is that they don't actually know
14:55 what they're creating there's there's
14:57 you know imagine Millions of images that
15:00 are no longer images they've been Den
15:03 they've been they've added a noise
15:05 algorithm to noise them to be just kind
15:08 of look like random noise and then when
15:12 you do a prompt for an image it sort of
15:16 deselects all the things that are not
15:19 tagged with whatever you've
15:22 written and then it kind of amalgamates
15:24 all those noisy images together and
15:27 creates one single noisy image and then
15:30 it denoises those using the same
15:32 algorithm that noised it in the first
15:34 place and
15:36 so when it's creating an image like that
15:39 it's a new original image every time but
15:42 it's got zero awareness of what it
15:46 created and it's got no ability
15:49 currently to look at what it created and
15:52 go oh that's not quite right now if you
15:55 could take something like this
15:58 image
16:00 if you could take something like this
16:02 image and run it back through the vision
16:06 model to say hey are there any type you
16:09 know is there anything in this image
16:10 that doesn't match what the prompt was
16:13 oh it looks like Irregulars missing a u
16:15 down below and it could isolate the area
16:19 that's got a typo and retry it until it
16:22 gets it right and then give you the
16:24 image that's when these things will get
16:26 better and that's that's somewhat in the
16:29 nature
16:30 of the uh the algorithm the qar
16:34 algorithm that they were talking about
16:36 that is rumored to have have
16:40 uh been a big breakthrough in AI that
16:44 the rumor is that might be the thing
16:46 that got Sam Alman fired although Helen
16:48 toner who just quit the board of open AI
16:51 said that it had nothing to do with us
16:53 hitting Advanced AI so so I don't know
16:57 we we don't know yet um um but that
16:59 algorithm basically
17:02 says have have one of these tools Tak an
17:05 action and then we're going to evaluate
17:07 the results of that action and we're
17:09 going to give it a score thumbs up or
17:11 thumbs down and then based on that score
17:13 it's going to take another action and
17:15 then we'll give it a score thumbs up or
17:17 thumbs down so it's keeping track of how
17:19 well it's it's doing and it just gets
17:22 better and better and better so so I you
17:26 know I don't think we're that far away
17:28 from these things being able to um to be
17:31 perfect every time I you know my
17:33 contention is that we're going to look
17:34 back on this time fondly and say oh that
17:38 must be a 2023 that's a 2023 set of
17:41 hands oh that's a 2024 typo you know uh
17:46 because these things are going to get
17:47 really perfect very very soon and uh and
17:51 then we we will miss the good old days
17:54 these are the good old days by the way I
17:56 don't know if you knew this but these
17:58 are the good old old days we are um we
18:03 are so I'm
18:08 old I don't know if you knew that
18:11 because you know supermodel good looks
18:13 and everything
18:15 but um in my
18:19 lifetime I just missed being part of the
18:23 PC Revolution so in 7th grade in
18:26 1977 it's when I went to 7th grade so
18:29 old
18:30 guy um there were Radio Shack trs8
18:34 personal computers on every desk in the
18:36 in my math classroom that I walked into
18:39 and I learned basic programming on that
18:41 and that was my introduction to PCS
18:45 and within 3 days of using those
18:51 machines I realized holy crap the world
18:55 is different because of these things
18:57 these things change
19:00 everything
19:02 and I was in seventh grade and Steve
19:04 Jobs and Bill Gates are on the cover of
19:06 Newsweek and and uh Time
19:12 Magazine and I remember feeling like
19:15 damn it I just missed I just
19:19 missed being at the
19:22 beginning of whatever that that was
19:26 right I remember reading the articles of
19:27 Steve Jobs
19:29 and wnc you know going down to fry
19:32 electronics and buying components and
19:34 building the first apple apple one
19:38 board and you know Bill Gates you know
19:41 how how he spun up Microsoft and I was
19:45 like like I was just too
19:47 young and I and I knew in seventh grade
19:51 that like those moment those the ability
19:55 the the the not the ability
19:59 the luckiness to be on the planet at
20:02 exactly the right time when one of these
20:03 moments begins is very rare and I didn't
20:07 think I'd see one again in my
20:10 lifetime and then I was living in New
20:12 York City in 1994 I got a degree in
20:15 acting from Penn State I moved to New
20:17 York City I was pursuing a life in the
20:19 Arts and I was doing desktop publishing
20:22 as my day job and around
20:27 9394 you just started seeing there there
20:30 was just all this Buzz about
20:33 technology so like on the side of buses
20:36 in New York City there'd be an ad for
20:38 Mondo 2000 which was a an early Tech
20:41 magazine out of San Francisco or Wired
20:45 Magazine and and there was just like you
20:48 could tell there was this Buzz of
20:49 something coming and so you know I was I
20:51 was trying to figure out what the
20:53 internet was and you had to dial up to
20:55 it and it was command line and there
20:57 were gopher service and FTP servers and
21:00 I joined a dialup server called uh echo
21:04 in New York City which was a a BBS a
21:07 bulletin board system where people would
21:10 talk and and Ekko was full of like a lot
21:13 of the reporters from the New York City
21:16 um media
21:18 scene and so I stumbled upon this thing
21:21 called the worldwide web and I was like
21:23 oh this is interesting and no one was
21:25 talking about the web and so I ended up
21:27 starting the the web um Channel within
21:33 Echo and there were like I don't know 10
21:36 people in there and we were talking
21:37 about this thing called the worldwide
21:39 web and I had an idea I said maybe I
21:40 could put up an online magazine I know a
21:42 bunch of creative people and I said
21:46 there's not a lot of
21:48 uh there's not a lot
21:50 of art and culture on the Internet it's
21:53 all like researchers and scientific
21:56 dissertations and like Librarians
21:59 talking about data I was like where's
22:03 the cool stuff and there wasn't much
22:05 cool stuff so I so let me see if I've
22:07 got I think I have a a page up in a
22:10 presentation
22:11 here yeah this is it here let me show
22:15 you
22:27 this
22:31 so this is from uh we we'll leave the
22:34 the llama and the turtle up in the
22:35 background that's
22:37 good so so this is from a a presentation
22:40 I gave uh to creative mornings uh back
22:44 in April uh talking about this time so
22:47 I'm living in New York City and I ended
22:50 up teaching myself enough
22:53 HTML to um to put together the first
22:57 issue of urban desires and this is what
22:59 it looked like it was a white page with
23:05 blue
23:06 hyperlinks and this is what the this is
23:09 what the web looked like in 1994 and
23:12 notice how the this was the cover of the
23:14 magazine which which looks like that um
23:18 so I had some photoshop skills and this
23:20 was an artist we were featuring so I put
23:22 together a magazine cover but notice how
23:24 it's left Justified on the web page why
23:27 because the center tag only applied to
23:29 text it didn't apply to images back in
23:33 1994 so we actually had to invent a
23:35 series of pixels so we could scooch
23:38 things to the right to center them on a
23:40 page um so I went on vacation with my
23:43 wife I learned enough HTML to put
23:45 together the first issue of this thing
23:48 and then I got back from vacation and
23:51 and I went on Echo and I said okay now
23:53 I've got a I've got a website what what
23:55 do I do with it like how do you get this
23:57 on the internet and someone someone on
23:59 Echo said I think it was John neelon
24:02 said uh oh you got to upload it I know
24:04 this computer in
24:06 California I was like okay he goes yeah
24:09 I know this computer in California it
24:11 was like a bad drug deal he goes he goes
24:13 you got to upload your files there it's
24:14 called a server I'm like okay and so I
24:27 did so that's what the internet sounded
24:29 like when your computer would call
24:31 another computer if you're if you're too
24:33 young to know that sound you're you're
24:35 an awful person uh but for you old
24:38 people that should bring back some PTSD
24:40 you actually had to call up other
24:42 computers and you could literally watch
24:43 it type across the screen how slow how
24:45 slow you connected to the internet so so
24:48 I put those things on this computer in
24:50 California this was
24:52 November
24:54 of
24:57 1994 it was right around this time
25:00 1994 today's November 30th it was like
25:02 the it was like the last week in
25:04 November three weeks later the second
25:06 week in December I got an email or a
25:10 note on someone from Eko that said
25:13 hey did you know that Urban desires has
25:18 been written about in liberon the
25:20 Parisian News Daily and I was like no
25:25 really they're like yeah so the next
25:28 Sunday I was I was living in New York
25:31 City the next Sunday I went down to Time
25:34 Square this is kind of what time Square
25:36 looked like in the mid 90s those are
25:38 what the cabs looked like back then so I
25:41 ran down to Time
25:42 Square um on a Sunday morning I got up
25:45 bright and early I was probably up this
25:47 early because I didn't want them to sell
25:49 out of this French newspaper and I
25:52 bought this French newspaper and I'm
25:54 standing in the middle of Time Square
25:55 flipping through the
25:57 pages
25:58 and sure enough there's this full- page
26:01 article about Urban desires this thing
26:04 that I had created on vacation and then
26:08 from my apartment in Brooklyn uploaded
26:10 to some computer in California and 3
26:13 weeks later it's written about in in
26:15 Paris and I realized the world just
26:18 changed and
26:20 so in that moment in that moment I knew
26:26 that this was the
26:29 beginning of one of those technological
26:32 shifts like the PC
26:35 was it was all kinds of fun chatting
26:37 with people on 56k it's how I acquired
26:39 my insane patience exactly
26:43 exactly and so in that moment I knew
26:47 that g geography had just collapsed that
26:50 that time had just collapsed that people
26:52 are going to expect things instantly you
26:55 know New York City at this time one of
26:57 the jobs I had before I was doing
26:59 desktop publishing I was a bike
27:00 messenger for 3 months in New York City
27:03 that was Insanity literally I got fired
27:06 from that because the guy that ran the
27:08 bike messenger shop found out I knew
27:11 computers he goes he goes he goes
27:15 Shannon I was like what he goes get in
27:17 my office so I get in his office he
27:20 slams the door behind me he goes what's
27:22 this I hear about you knowing [ __ ]
27:25 computers I said I said yeah I know
27:28 computer's okay he goes what the [ __ ]
27:30 are you doing in here he fired me he was
27:34 like he was like this is not a job for
27:36 guys that know
27:39 computers so that was that was probably
27:41 a good thing I'm probably still alive
27:43 because that guy fired me um so so
27:49 anyway the reason that I'm so amped and
27:53 pumped and insane about
27:56 AI is that I had had almost an identical
27:59 Epiphany with with
28:01 AI
28:05 um
28:07 probably a week or so after chat GPT
28:10 came out so it came out today a year
28:13 ago and I was sitting sitting in my
28:16 living room my wife was watching TV and
28:20 I I said oh I I heard that chaty BT can
28:23 write code and people have told me for
28:25 years oh you should learn python
28:30 and
28:33 um and I was like okay yeah I should
28:36 learn Python and if it can code let me
28:38 let me let me see what it's like to do
28:40 python like I had looked at it but like
28:42 I didn't feel like I've got add so like
28:44 learning a new language for me is like
28:46 [ __ ]
28:47 torture so I said can you write I I said
28:51 to chat gbt write
28:53 me a python script that takes some
28:56 inputs and I don't know
28:58 records them or I don't I don't even
29:00 want what I said I did I was asking a
29:03 nonsensical programming question and
29:05 just goes and wrote out the code and I
29:08 was like oh that's cool and then I
29:11 realized I don't know how to run python
29:13 code so he said hey chat GPT uh how do
29:16 you run python code said wow you open
29:18 your terminal and you do this and you
29:19 run this command line and I was like oh
29:22 that sounds awful I said no I don't I
29:23 don't want to run my command line um and
29:27 and uh I said isn't there a website I
29:29 can just copy and paste the code in and
29:31 see if it works he said oh sure here's
29:33 five I was like oh okay and the first
29:35 one was repet which is you know
29:39 famous uh I didn't know it at the time
29:42 but like a popular site for running
29:45 python so it said copy the code and
29:49 paste it in here and hit the Run button
29:51 and then it'll run in the right hand
29:52 window I was like cool copied the code
29:55 pasted it over there hit the Run button
29:57 and it gave me me an error I was like
30:00 well that's typical that's my experience
30:02 of of this crap nothing works it's all
30:05 janky um and so I told it it it it
30:09 didn't work it gave me this error it
30:10 said oh that's because you don't have
30:12 this Library installed I was like okay
30:14 how do I install that Library it said on
30:17 the repet page in the lower left hand
30:19 corner click this button and in the
30:22 upper right hand corner search for this
30:24 library and when you see it click
30:26 install and then hit the Run
30:29 button I was like oh okay I clicked the
30:32 button I searched for the library I hit
30:34 install I hit the Run button and it
30:36 worked I was like oh within 90 minutes I
30:41 had
30:43 coded a sophisticated python application
30:47 that ask someone five questions took
30:50 those answers formulated them into a
30:52 prompt sent it to open Ai and generated
30:55 a piece of social media content and then
30:57 I I did that call five times so I I had
31:00 in 90 minutes I had a I I had a a
31:04 Content generation engine based on
31:07 questions I would ask someone about some
31:09 topic they were interested in in 90
31:12 minutes never having written a line of
31:15 python in my
31:17 life and that was when I just
31:20 went the world just changed but no one
31:23 knows
31:26 it because this this isn't Googling how
31:29 do I write python code and spending the
31:32 next 3 months slamming my face into a
31:35 brick wall to be able to get to hello
31:38 world right you know so so like one of
31:42 the things that always killed me about
31:45 programming tutorials is that all of the
31:48 example programs that you have to write
31:50 first are just stupid little stupid
31:53 little tools happy birthday chat gbt hey
31:55 lunar stick happy
31:58 birthday um where do we have we got we
32:02 got birthday the guy in the birthday
32:06 suit there we go happy birthday Chad
32:11 j so so we are in a
32:15 remarkably remarkably rare moment in
32:19 history like why am I why am I going
32:23 live for 24 hours in
32:26 celebration of a janky ass piece of
32:30 software because chat
32:34 GPT does for machine learning and AI
32:37 what the worldwide web did for the
32:40 internet it took it from you know
32:44 Decades of foundational work Decades of
32:47 researchers and
32:49 mathematicians creating protocols and
32:51 and create making it work and then Tim
32:54 burner Lee comes up with this this Pro
32:57 protocol for for the worldwide web where
33:01 you can click on a
33:03 hyperlink
33:05 click and it would take you to another
33:07 page click and it would take you to
33:09 another page click you're in another
33:11 country
33:13 click oh holy [ __ ] that's really
33:17 significant because not many people are
33:20 going to type command lines in to
33:22 experience the internet but anyone can
33:24 do
33:25 that same thing with Ai and machine
33:28 learning I would say I've been
33:30 interested in Ai and machine learning
33:32 for the past
33:34 decade and I would say starting about 5
33:36 years ago Google started trumpeting you
33:40 know tensor flow and we've got these AI
33:42 tools and anybody can do it and you'd go
33:45 look at the documentation and you'd be
33:48 like no not anyone can do it anyone with
33:51 a computer science degree and a
33:53 willingness to sit there for for 17
33:56 hours reading documentation can do
34:00 it if they've got massive data sets that
34:03 they want to train specialty machine
34:06 learning models on anyone can do
34:08 it so what chat GPT is if you don't know
34:12 what GPT stands for generative it's
34:15 generating it's not copying and pasting
34:18 it's generating
34:20 pre-trained pre-trained meaning it's
34:23 they've brought the data for you it's
34:24 been trained on as much of the public
34:26 internet as could get their hands on and
34:29 Transformer the technology that allows
34:31 them to train it on all this data
34:35 GPT and then chat GPT is a layer sitting
34:38 on top of that that says we're going to
34:40 make it super simple you don't have to
34:43 turn any knobs or switches or anything
34:46 like that you could do the you could
34:48 essentially play with chat GPT on the
34:51 open AI playground for about 2 years
34:54 before chat GPT came out in fact
34:58 here's a little piece of trivia for
35:03 you
35:05 the the instructions for how to create
35:08 chat
35:09 GPT were in the documentation for
35:14 gpt3 and nobody built
35:19 it and Microsoft was about to announce
35:22 their billion dooll investment in open
35:24 Ai and Sam Alman in Octo cber said to
35:28 his Dev team hey Dev team I know you're
35:32 working on the future of humanity here
35:34 but could you guys just go on ahead and
35:36 whip up that Chat gbt thing we put in
35:38 the
35:40 documentation cuz no one's built it and
35:42 I want something to be able to point to
35:43 that says here's what this [ __ ] can
35:46 do and so in two
35:49 weeks the dev team at open AI whipped up
35:52 a little chat
35:56 GPT and so they made the announcement
35:58 that Microsoft made their investment and
36:00 they and they said here's chat GPT go
36:03 play with
36:05 it fully expecting no one would give a
36:12 [ __ ] and in five days a million people
36:16 were active users and in six weeks a 100
36:20 million active
36:24 users 100 million active users in 6
36:27 weeks it was the fastest adoption of
36:29 technology in
36:31 history and they did no advertising it
36:34 was all Word of Mouth it was no
36:36 corporate or Enterprise deals that were
36:38 driving that it was human
36:41 beings one after the other having Kevin
36:44 mallister
36:46 moments oh my God it does
36:50 that Jane you got to try this
36:55 thing 100 million users to put that in
36:59 perspective the worldwide web to get to
37:01 100 million users took six
37:05 years because there wasn't any
37:07 infrastructure right you had to dial up
37:10 your computer had to call other
37:11 computers you paid by the
37:15 minute took six years to get to 100
37:17 million users for the worldwide
37:20 web 6 weeks for chat
37:23 GPT and so chat GPT changes everything
37:27 it's the thing that reveals the power of
37:30 AI and machine learning to regular human
37:33 beings that can just type in to a Google
37:36 search box right it's just as simple as
37:38 a Google search box well what do you do
37:40 there you do whatever the [ __ ] you want
37:42 there well what does it know it knows
37:45 everything okay well what does that mean
37:49 I don't know ask it to translate
37:52 something for you ask it to write code
37:55 for you ask it to write you a poem ask
37:58 it about some obscure Battle in History
38:01 that you are an expert on it will likely
38:05 know
38:07 it yeah but it's out of
38:10 date and it
38:14 hallucinates I'm sorry you were just
38:17 handed the library of humanity where you
38:20 can instantly have
38:22 knowledge and you're bitching that it
38:25 ain't
38:26 perfect
38:27 yeah we're human beings we're going to
38:29 [ __ ] it's like yeah that's remarkable
38:31 but whatever it misspelled something so
38:35 I'm just going to completely dismiss
38:38 it you want to see something cool sure
38:40 Kyle it's your
38:42 channel how do I use chat gbt okay we
38:45 we'll do that I'll give you a thing I'm
38:47 going to show you I'm going to show you
38:48 a way to use chat gbt so this is just an
38:50 iPad I got it on one of them fancy
38:52 keyboard things which I like to chew
38:54 on um there's the chat GPT app okay so
38:59 I'm going to launch
39:02 that and I'm going to make oh we'll just
39:05 do it right in here so this is a
39:09 conversation I was having so see this
39:11 little icon in the lower right you got
39:13 your chat box at the
39:15 bottom you can now put this is available
39:18 to any chat GPT user so if you download
39:20 this is the official chat GPT app it's
39:22 not there's a lot of fakers out
39:25 there if you if you go to the chat
39:28 GPT um app in Google Play or in the App
39:32 Store make sure it's from open
39:34 Ai and then when you run it click on
39:38 these little
39:46 headphones okay I'd like to do Italian
39:53 now sure I'm ready to translate into
39:55 Italian please please go ahead with your
39:59 sentence it's early my blood are my eyes
40:02 are bloodshot and I can't really
40:15 think yeah my German Buddy's here too
40:17 can you translate that into
40:22 German yeah my is out
40:25 here
40:30 I need more coffee do you have more
40:32 coffee where is the coffee machine I'd
40:34 like some coffee
40:47 please oh my Japanese friend just showed
40:50 up can you tell
40:54 them oh my
41:01 no please translate to
41:13 Japanese can you translate into
41:19 eru of course I can translate into UE
41:22 please share the sentence you'd like
41:25 translated Jim was wearing his birthday
41:28 suit when the door flew
41:40 open
41:42 so so that
41:45 [Laughter]
41:50 exists that didn't really exist not that
41:54 long
41:55 ago
41:58 now that's just on an iPad so that's on
42:00 my iPhone so on my iPhone in any country
42:03 right now I don't need some fancy
42:06 translator thing I can just fire fire
42:08 this up um somebody asked how do you use
42:11 chat GPT let's go play with it um holy
42:15 [ __ ] how can it do ero is mind-blowing
42:20 I happy birthday Chad gbt this
42:25 is so on this channel I've been going
42:28 live seven nights a week
42:32 for 6 months now since
42:36 April
42:38 um because I can't get my head around
42:43 what's happening like so much is
42:45 changing so fast that thank you
42:49 Vicki that it's really hard to
42:51 comprehend and part of what this
42:53 Channel's about is just connecting with
42:56 people people that are exploring this
42:58 stuff as well and trying to figure it
43:01 out as well
43:03 and I think that my my experience has
43:08 been we are collectively mind blown and
43:11 face melted on a daily basis where can
43:15 we get normal chat gbt okay so here's
43:17 what we're going to do we're going to
43:18 give you all a leg
43:20 up get ready to do some
43:25 screenshotting
43:27 M and also I'm going to talk about the
43:30 salon if you want to get a head start go
43:32 to AIS salon. mn. and I'll tell you
43:35 about that check out the about the salon
43:37 section there okay
43:41 so if you have not played with chat
43:44 GPT it's its
43:46 birthday so go wish it a happy birthday
43:50 okay go to chat.
43:52 open.com just go there right now you're
43:55 like but
43:57 I don't know if you knew this but I'm
43:58 not very technical I didn't go to some
44:01 fancy ivy league mathematics school like
44:04 you did no I went to a state school and
44:06 got a degree in
44:08 acting okay
44:11 okay so go
44:14 there and and drop the
44:18 intimidation drop the
44:20 dismissiveness drop the AI is evil drop
44:23 the AI is going to kill us the robots
44:25 are going to kill us and just go there
44:27 just go there right now and I we'll go
44:29 there together dang it dang it I'll tell
44:31 you what I'll tell you what we're going
44:34 to go to chat JP you know what we're
44:36 going to do we're going to do some JP in
44:38 we're going to do it all right so first
44:41 of all chat GPT has a cute little
44:44 birthday hat how cute is
44:47 that can chat GPT really invest for you
44:50 how yes it can well it can sort of well
44:54 actually now it probably can really
44:57 huh look at that it couldn't two weeks
44:59 ago it could tell you what to do now you
45:01 could probably build a GPT in about an
45:04 hour not even half an
45:08 hour that can probably do that for
45:11 you with the help help of chat GPT I
45:15 wrote a 100 Tik Tock scripts last night
45:17 there you go Luna Luna Zamora yeah it's
45:19 it's amazing okay so this is it you're
45:22 like okay Kyle I okay not impressed
45:27 so what it probably looks like to you is
45:30 something like this where you got the
45:31 chat GPT logo here you probably have
45:33 this thing that says
45:34 explore if you've typed in any searches
45:37 or prompts you'll have a prompt
45:41 history and then you have this thing and
45:43 it's got these four things make a
45:45 Content strategy explain superconductors
45:47 brainstorm names show me a code snippet
45:50 and those are probably different for
45:52 everyone I'm not sure I don't know if
45:53 those are static or not I think they
45:55 dynamically pop those in in there and
45:57 and then this is it this is
46:00 it you like okay well so it's like
46:04 Google and and you're like sort
46:08 of you could say you you could say it's
46:11 sort of like
46:13 Google
46:15 um but it's not it's not Google so
46:20 so what I could say is I could say
46:24 um explain
46:31 quantum
46:38 mechanics to me all right you're like
46:41 okay well I could do a Google search for
46:43 that
46:45 Inn
46:47 mnm
46:52 WAP and it
46:55 will
46:58 yo gather around let me drop some
46:59 science real quick quantum mechanics
47:02 it's the quantum physics pick in the
47:04 world of the small on a subatomic scale
47:07 where particles dance and the Mysteries
47:10 Prevail Quantum Quantum it's a wild wild
47:12 ride in the quantum realm where laws
47:15 Collide it's a world so strange can't be
47:18 understood with ease but I'll break it
47:20 down for you for y'all just listen
47:24 please oh okay it can write poems cool
47:27 it knows what verses and choruses are it
47:30 it understands song structure it
47:32 understands rhyming it it understands
47:35 now could you as a human do this yeah
47:39 would it take you half an hour or so
47:42 sure if you knew how quantum mechanics
47:44 worked and if you were good at writing
47:46 wraps so so you're like okay okay all
47:50 right uh so now I'm going to say
47:55 um
47:58 great now turn this into a graduate
48:05 level term paper for quantum mechanics
48:09 101 that wants us to write a paper on
48:12 the first
48:16 day and this is the this is the cheating
48:18 thing where people like oh kids are
48:20 cheating kids are cheating who wants the
48:23 output you on the output kids are
48:25 cheating look the kids are cheating he
48:27 just he didn't write that let's use the
48:30 AI
48:32 detector introduction quantum mechanics
48:34 often referred to as quantum physics is
48:36 the branch of physics that deals with da
48:38 da da da da da the wave particle duality
48:41 okay so great so now um explain it to me
48:45 as if I'm
48:48 [Laughter]
48:49 [Music]
48:51 five sure imagine you have a really tiny
48:55 toy car so so tiny you can't even see it
48:57 with your eyes now this tiny car can do
48:59 some really weird things that big cars
49:02 can't it can be a ghost and a real car
49:06 sometimes the kind tiny car acts like a
49:08 ghost it can be in two places at once
49:10 right so it explains all this it's it's
49:13 slowing down now it's normally chat GPT
49:16 3.5 is really fast um I'm going to stop
49:19 this and then I'm going to say
49:22 um okay
49:25 great now what I'd like to do is create
49:29 a python application that demonstrates
49:32 how quantum mechanics works for someone
49:36 trying to understand
49:38 it how would you go about creating that
49:44 application so it can explain how it's
49:47 going to write an
49:48 application pretty quickly and if you're
49:51 if you're not reading along because it's
49:53 going too fast by the way this is this
49:56 is creating original content faster than
49:59 dialup modems you you could this is
50:02 faster than the internet was uh when I
50:05 started
50:06 out um Step One choose Quantum Concepts
50:10 step two choose Quantum Computing
50:12 framework so it explains how to do it
50:15 gather feedback so I'm going to go great
50:18 can you write that
50:24 program
50:30 um is that the front end or the back end
50:33 period what if I wanted a pretty front
50:35 end that makes it easy to
50:40 use the code you provided create a GUI
50:46 Library there you go so so this is just
50:51 chat GPT 3.5 this is the chat GPT that
50:59 that launched a year ago it was a little
51:02 slower than this but this is essentially
51:03 what it did so this has been this has
51:05 been out for a year and you could play
51:07 with uh gpt3 on on What's called the the
51:11 developers playground for like two years
51:13 before that so so this has been around
51:15 for a
51:17 while
51:19 um but there's there's now this thing
51:21 called GPT 4
51:24 unfortunately um GPT 4 if you don't have
51:27 access to it um they're not accepting
51:29 new um what are they called
51:37 subscriptions my my voice is seizing
51:40 up can you can you copyright the product
51:44 output as your own yes you can I don't
51:47 know why you'd want to but you
51:50 can again it's creating original content
51:53 now can you do weird nefarious [ __ ]
51:57 since since this has been trained on
51:59 most of the public
52:01 internet essentially everything that's
52:03 been put on the internet for the past 30
52:04 or 40 years that's publicly available
52:07 and probably some [ __ ] that isn't
52:09 publicly available um has been this
52:12 thing's been trained on it it's it's how
52:13 it knows so much think think of it like
52:16 like the the world's fastest
52:19 librarian
52:20 um so you could get it to do stuff and
52:23 this is one of the reasons Sarah
52:25 Silverman is suing open AI is you can
52:29 get it to produce a quote from her book
52:31 or from some other book and so they're
52:33 saying oh well it's copying and pasting
52:35 it's stealing it's not really stealing
52:37 it's it's actually if you understand the
52:39 way these things work it's actually
52:40 generating original content but it's
52:42 been trained on lots of other stuff but
52:45 every single thing it produces is wholly
52:49 original
52:52 um you you know with with some some very
52:56 minor exceptions if you tweak the
52:59 settings in a certain way you can get it
53:01 to you can get it to repeat itself but
53:03 it's still the the thing it produced is
53:06 original
53:08 um and so yeah so you can you could you
53:11 could you could own the well you you
53:14 could claim the copyright on it
53:17 now will the copyright office give you a
53:20 copyright on it that's a completely
53:22 different things can it tell us if you
53:23 are U UFOs are real we can ask
53:27 it I want you to act as the world's most
53:32 prominent
53:35 ufologist and give me five pieces of
53:39 concrete evidence that UFOs are
53:46 real it's long fast we might get a
53:49 safety thing here government disclosure
53:51 of UFO sightings radar and pilot
53:54 testimonies
53:59 physical Trace
54:03 cases the Tic
54:05 Tac UFO encounter from the 2004 USS
54:16 nits can you tell me definitively if
54:19 UFOs
54:22 exist it I bet it says no
54:27 definitively stating whether they
54:31 exist depends on the context of what is
54:34 meant by UFO it depends what the meaning
54:37 of if
54:39 is or it or whatever Clinton said um if
54:43 we Define a UFO simply as an aerial
54:46 phenomenon that observers cannot
54:48 immediately identify then yes UFOs
54:50 undoubtedly exist however if the
54:53 question is evidence of extra
54:55 extraterrestrial life the answer is less
54:58 clear so it and that's you know one of
55:01 the things that that chat GPT does
55:03 really well is give Nuance nuanced
55:06 answers um so anyway okay so chat GPT
55:12 4 I'm going to start a new chat here so
55:16 I'm G to I'm going to walk you around
55:17 the interface because there's there's
55:18 actually a fair amount going on even
55:21 though it looks really simple so so
55:23 we're in chat gp4 mode we still have
55:25 these prompting questions there's a new
55:28 icon here in chat gp4 that's this giant
55:30 paperclip where it's not giant but it's
55:33 gier than most of the times that icon
55:36 exists it's a weird it's a weird
55:37 paperclip icon I think they might be uh
55:40 they might be trying to bring clippy
55:43 back um you can upload files to chat GPT
55:47 and it can analyze them and it can
55:49 analyze everything from CSV files to
55:52 images it can see images chat GPT can
55:55 see
55:56 um you've also got so this is your
55:58 prompt history so so here's the quantum
56:01 mechanics M&M WP I can go back and look
56:03 at that right there's all that stuff
56:05 here's the UFO thing we just did and
56:08 then there's this explore button and
56:10 then I've got this big long list of
56:13 these other things what are these These
56:15 are called
56:17 gpts so gpts are custom versions of chat
56:21 gbt that you can make that we can make
56:23 that anyone can make if you've got
56:25 access to
56:27 gp4 these are fairly gamechanging in my
56:31 in my opinion we'll talk about what they
56:33 are um and then you have this explore
56:36 button where you can go look at the gpts
56:39 that are there like if you haven't used
56:40 a lot of these your your chat gbt may
56:43 look something like that okay you also
56:46 down here in your um in your uh settings
56:51 have this thing called custom
56:53 instructions custom instruction
56:55 instructions let you tell chat GPT how
56:58 to
56:58 act um so you could tell it to act like
57:01 a belligerent jerk you could tell it to
57:04 act like someone that hates extra words
57:07 actually let's do
57:09 that so so let me just do a new chat GPT
57:12 and I'll
57:15 say why is the sky
57:19 blue and chat GPT the color sky blue is
57:23 due to the scattering light and Earth
57:25 atmosphere the phenomenon is known as
57:28 rally scattering here's how it
57:30 works so it gives us all this stuff it's
57:34 you know it's it's
57:36 wordy it's it's a lot like I got add I
57:39 don't have time to read your stupid
57:42 flipping dissertation on why the sky
57:44 blue there chat GPT so let's go fix
57:48 that so we're going to turn this on and
57:53 here I had it I had it before it was
57:54 acting like a a master
57:57 watercolorist going to
58:00 say you are a very
58:03 impatient
58:05 person who can't stand
58:08 Superfluous
58:11 words
58:13 period you Embrace
58:16 brevity and abhor run on sentences
58:22 period so so the first thing is what
58:25 would you like chat GPT to know about
58:27 you to better provide responses so this
58:29 is kind of the who are you and you can
58:31 say you or I it doesn't matter and then
58:34 the second one is how do you want chat
58:36 GPT to respond so we're going to give it
58:38 some
58:40 rules whenever anyone asks you a
58:45 question never
58:47 provide a setup or conclusion
58:51 period no hang
58:54 on
58:56 whenever someone asks you a question
58:59 follow these
59:11 rules never provide a setup or
59:14 conclusion to your answer
59:16 period try to keep the
59:19 answer to one sentence if
59:24 possible if you need to provide more
59:27 explanation only provide it as a short
59:30 bulleted list never
59:33 more
59:35 period and it wrote never more like
59:39 oops it wrote never more like Edgar
59:41 Allen po that's okay it probably
59:44 understands it so now I'm going to hit
59:45 save and now chat GPT is going to be a
59:49 completely different
59:51 thing why is the sky
59:54 blue
1:00:04 I think that's it there you go the sky
1:00:07 is blue due the scattering of sunlight
1:00:09 by the atmosphere which is more
1:00:10 effective at shorter blue wavelengths
1:00:14 sunlight contains all the colors blue
1:00:16 light scatters more than other colors
1:00:18 scattered Blue Light reaches our eyes
1:00:20 from all directions there you go that's
1:00:23 it no [ __ ] chat GPT is no longer too
1:00:26 wordy you can do it the opposite way you
1:00:28 can say I you you're a lover of words
1:00:31 and and oh actually let's go do that
1:00:34 this is a fun one I've done this
1:00:37 before okay so we're going to flip
1:00:41 it you are a lover of words period you
1:00:46 love fancy words
1:00:49 period you insist on using words in
1:00:53 every response that even you don't
1:00:56 understand
1:00:58 period it is your great pleasure in life
1:01:02 to watch people confused by what you say
1:01:06 knowing that if they
1:01:14 parse
1:01:17 your Genius
1:01:20 statements they will indeed find that
1:01:23 they are accurate period
1:01:26 all right so whenever someone questions
1:01:29 you um follow these
1:01:39 rules always start with an obscure
1:01:42 literary
1:01:44 reference that their question reminds
1:01:46 you
1:01:49 of and find a way to
1:01:52 work a quote from
1:01:57 that
1:02:00 work into your answer period Then when
1:02:05 you answer the
1:02:07 question make sure that you can use the
1:02:11 most complicated word
1:02:19 possible
1:02:23 period
1:02:25 also include
1:02:28 tangents where you tell the story of a
1:02:31 word that you just introduced in
1:02:33 answering their question period it
1:02:36 should take them a while to understand
1:02:39 if you're ever going to answer their
1:02:41 question
1:02:43 period
1:02:47 okay this should be obnoxious
1:02:51 save okay so there was our there was our
1:02:53 brevity dude
1:02:55 so here here's our obnoxious
1:03:00 chick why is the sky
1:03:05 blue ah your question transports me to
1:03:09 the lyrical realm of Percy Bashar Shelly
1:03:12 to a Skylock where he eloquently Muses
1:03:15 like a poet hidden in the light or
1:03:18 hidden in the light of thought singing H
1:03:21 hymns unbidden till the world is rot to
1:03:24 to sympathy with hopes and fears he did
1:03:27 not just like the Skylock himym is a
1:03:29 cryptic raps to the Azure Hue of the sky
1:03:32 is a poetic Enigma of nature the
1:03:36 phenomenon of the sky's bless bless is
1:03:39 aptly elucidated by the principle of
1:03:41 Raleigh scattering the scientific term
1:03:44 named after the 19th century physicist
1:03:46 Lord Raleigh right it's just goes on and
1:03:49 on right
1:03:52 perfect it's Percy BL Shelly
1:03:55 it sure
1:03:56 is you never thought you'd hear him
1:03:59 again oh man
1:04:03 so so just just in that little teeny
1:04:06 tweak of this thing called custom
1:04:08 instructions that I think most people
1:04:10 don't even know is there you can turn
1:04:13 your chat GPT to always behave a certain
1:04:16 way so this came out I don't
1:04:20 know um 3 months ago 4 months ago not a
1:04:24 lot of people knew about
1:04:26 it so now we have these things called
1:04:29 gpts which you can now essentially
1:04:33 save that custom instruction as a
1:04:37 standalone version of chat GPT that's
1:04:39 super personalized and I I'll show you
1:04:41 that in a second but I'm I'm going to
1:04:42 show you some other
1:04:44 things um and this is pro this probably
1:04:47 going to take me like I've got to leave
1:04:48 for work in in uh in I don't know an
1:04:53 hour or so
1:04:55 um does it state where the info was
1:04:57 obtained no it doesn't you can ask it
1:04:59 and it will tell you but it it lies a
1:05:01 lot
1:05:03 so if frankly doesn't know where the
1:05:06 information was obtained
1:05:09 um there are likely ways that they could
1:05:12 modify the database to be able to pull
1:05:14 references out if you want something
1:05:16 that uses um references and citations
1:05:20 like that play around with a tool called
1:05:22 perplexity AI perplexity . a um is
1:05:27 connected to the internet and it's very
1:05:29 much designed as a research tool where
1:05:31 you can pull it now chat GPT one of the
1:05:34 things that chat GPT has now that it
1:05:37 didn't for a while and this one of the
1:05:39 knocks on it was well it's only good up
1:05:41 until September
1:05:42 2021 how's that useful well most of
1:05:45 human history existed before that so
1:05:48 it's it's it's actually pretty useful
1:05:52 um um but it's it's now connected to the
1:05:55 internet again so so I could say
1:05:59 um what is some live music that's
1:06:03 happening in Denver this weekend period
1:06:05 I like Americana singer songwriter kind
1:06:08 of stuff so see if you can find music
1:06:11 that I would
1:06:15 like so doing research with Bing and so
1:06:19 it's kind of telling you what it's
1:06:20 searching here but it's off lunatick
1:06:23 perplexity aai is great for citation
1:06:26 yeah exactly all right great you gained
1:06:30 by subscription thank you very much I
1:06:31 appreciate that yeah if you want to
1:06:33 support the channel um share this live
1:06:35 wow there's 300 people in here welcome
1:06:37 welcome welcome so happy birthday to
1:06:39 chat gbt
1:06:41 money uh my name is Kyle Shannon this is
1:06:44 the AI learning lab uh I'm doing a non
1:06:47 standard uh thank you very much that's
1:06:50 awesome uh I am doing a I normally go
1:06:53 live around 8:00 at night mountain time
1:06:55 I'm in Denver um but I'm doing a 24-hour
1:06:58 chat GP Aon uh cuz it's the oneye
1:07:02 anniversary of chat GPT Be not Afraid
1:07:04 thank you so much for the roses that
1:07:05 means a
1:07:06 lot um I have a question for you how to
1:07:09 use chat GPT for my work as a mechanical
1:07:11 engineer okay we can do that Luna thank
1:07:13 you very much appreciate the donut it's
1:07:16 I love donuts all right so chat GPT is
1:07:19 still off looking for events for me
1:07:21 we'll see what it it does Christopher
1:07:23 thank you so much
1:07:26 um can you
1:07:28 check in throughout the day from work
1:07:31 yeah I I'm am going to so so lunatick
1:07:34 what I'm gonna do is i' I'm speaking on
1:07:37 a a small uh like a an invite only
1:07:41 conference call for Content Evolution a
1:07:43 company that I advise for I'm the I'm
1:07:45 actually the chief generative officer
1:07:47 for this group so at N9 o'clock my time
1:07:51 I'm speaking on um
1:07:55 a conference call with them and so I'm G
1:07:58 to have Tik Tock live facing me from the
1:08:01 side I'm going to be talking to them but
1:08:03 you're going to be looking at them on a
1:08:05 different monitor so I'm going to be
1:08:07 doing a sort of
1:08:08 Matrix a matrix uh in Inception kind of
1:08:13 thing this weekend in Denver there's a
1:08:15 few live music
1:08:17 events on
1:08:19 Saturday his golden messenger and a deem
1:08:21 the artist will be performing at the
1:08:23 Ogden additional
1:08:25 Crash Test Dummies and Carlton Stone are
1:08:27 at the
1:08:29 Oriental unfortunately I can't find more
1:08:31 events so anyway so it's connected to
1:08:32 the internet um one of the other things
1:08:36 chat GPT can do now is it can see you're
1:08:40 like but Kyle what does that mean what
1:08:42 do you what do you speak of hello dear
1:08:45 Kyle Shan oh hey Mark oh hi what what do
1:08:48 you speack of new people Kyle's here
1:08:51 every night from about 8 o'clock MST
1:08:53 every night thank you very much so yeah
1:08:55 welcome uh if you're if you're willing
1:08:57 to share this live let people know we're
1:09:00 here I'm doing a 24-hour chat GP Aon
1:09:03 where I am rambling like in between work
1:09:06 like I got work to do I have a pitch
1:09:08 today and I've got this speaking gig uh
1:09:10 but I'm doing this sort of over the 24
1:09:12 hours it started last night at midnight
1:09:14 my time um so if if you're if you want
1:09:18 to learn about AI th this um channel is
1:09:22 kind of the tip of the iceberg of a
1:09:23 larger community called the AI Salon
1:09:26 that I started also a year ago our
1:09:28 birthday is coming up next Tuesday um so
1:09:31 follow this channel um I produce content
1:09:33 here and I do these lives a lot so if
1:09:35 you want to be notified of the lives
1:09:37 follow the channel um if you have
1:09:39 questions about AI pop them in the
1:09:41 comments below Kyle you're still going
1:09:43 since s last night I've taken a couple
1:09:46 of breaks I went from I went from 8 last
1:09:49 night until 10:30 then I went from
1:09:51 Midnight to 3:30 and then I started
1:09:54 started again at 6:00 this morning and
1:09:56 then I'll be doing it all day I'll be
1:09:58 doing it all night and then I'll I'll
1:10:00 probably do it most of the night so I
1:10:05 got to take breaks to go take naps and
1:10:07 you know get get blood flowing in my
1:10:09 legs and things like that um but anyway
1:10:14 um I'm trying to avoid monthly frees is
1:10:16 there a free version relevant um so yes
1:10:20 if Jan Jan if you go to um
1:10:24 co-pilot
1:10:27 microsoft.com that is
1:10:30 Microsoft's uh build of chat
1:10:34 GPT um so actually let me I'll I'll take
1:10:37 you there right
1:10:40 now so so this is it let me restart new
1:10:45 topic so when you go to co-pilot this is
1:10:48 what it looks
1:10:50 like and these are some start like you
1:10:53 can click on these things these are
1:10:55 starter things in the upper right hand
1:10:56 corner here's your prompt history so so
1:10:58 it's very much the same as chat GPT it's
1:11:01 got this prompt box down at the bottom
1:11:03 you can upload images you can ask it um
1:11:06 you know you can say
1:11:08 [Music]
1:11:09 um write
1:11:11 me a
1:11:13 poem like an Eminem rap that explains
1:11:17 quantum
1:11:18 physics so you can do parlor tricks like
1:11:20 that and it's this is basically built on
1:11:23 top of chat GP T4 and it's
1:11:25 free um I personally don't think it's
1:11:29 anywhere near as good as as chat gpt's
1:11:31 version of or open AI version of chat
1:11:34 gbt just because the safety guard rails
1:11:37 on on this are are way too restrictive
1:11:40 in my opinion like if you if it does
1:11:43 something you don't like and you go well
1:11:44 that's shitty like it'll just shut down
1:11:46 the conversation like well there's no
1:11:48 need to get rude and it'll just cut down
1:11:50 the conversation so it just bugs me but
1:11:54 it's it's still this is free and it it
1:11:56 makes images it it can you know I could
1:11:59 say
1:12:00 um great now make me an image that would
1:12:04 go along with that poem If I wanted to
1:12:06 publish it on
1:12:11 LinkedIn and it's now going to go off
1:12:13 and it's going to use dolly3 so so it
1:12:16 uses Dolly 3 to generate
1:12:18 images and we'll see this over in chat
1:12:20 GPT as well I'll try to create that
1:12:25 it also you might not see it but it says
1:12:27 two of 30 so it limits your your chats
1:12:30 to to 30 interactions um because of some
1:12:33 safety [ __ ] that happened with the New
1:12:36 York Times Reporter last year or or at
1:12:39 the beginning of this year all right so
1:12:41 there's our images that represent you
1:12:43 know quantum physics in an Eminem rap
1:12:46 it's pretty
1:12:50 good there you go it looks like Eminem
1:12:52 there you go there there's there's your
1:12:54 image for your article so um so yeah so
1:12:57 this is
1:12:57 free and it's
1:13:00 good but it's not ch ch
1:13:04 ch um so I would strongly highly
1:13:08 recommend uh if you're if you're using
1:13:10 chat
1:13:11 GPT um and you're new to it get on the
1:13:15 wait list for chat GPT 4 because they've
1:13:17 restricted new people
1:13:20 joining um okay I'm going to do
1:13:23 something new here so I mentioned before
1:13:25 that chat GPT can see so what's that
1:13:28 mean well I'm going to go find an
1:13:40 image that one's
1:13:43 interesting yeah that's
1:13:46 good all
1:13:48 right I think it was I don't know who
1:13:50 did this is steo or Ted oh no Chris
1:13:54 Chris did this one Chris
1:13:56 Ramsey um okay so I'm going to I'm going
1:14:00 to upload that image so it's a it's a
1:14:02 laptop it says AI learning Labs with a
1:14:05 helmet it's a motorcycle helmet box or
1:14:07 laptop box there's a motorcycle in the
1:14:09 background wood
1:14:14 floor
1:14:15 um
1:14:19 okay so I'm going to put that in there
1:14:21 I'm going to
1:14:22 say
1:14:25 describe this image in as much detail as
1:14:38 possible the image shows an open laptop
1:14:41 placed inside a cardboard box with the
1:14:42 term motorcycle helmet printed on the
1:14:44 side suggesting the box is a product
1:14:46 packaging for a motorcycle helmet the
1:14:49 laptop screen displays an image of a
1:14:51 helmet that seems to be stylized with
1:14:52 intricate designs the words AI learning
1:14:55 labs are PR prominently displayed in the
1:14:57 middle of the helmet that's right it
1:14:59 goes on and on and on the laptop appears
1:15:02 to be a Macbook as indicated by the
1:15:04 layout of the keys the design of the
1:15:06 touchpad along the characteristic
1:15:08 MacBook hinge the screen is high
1:15:10 resolution
1:15:11 and the lighting is warm coming from the
1:15:14 right side I'm going to
1:15:17 say great I
1:15:20 have a Tik Tok Channel called the AI
1:15:24 learning lab and I'm thinking about
1:15:28 using this image to do a series of
1:15:32 social media ads to drive awareness of
1:15:34 my channel
1:15:36 period can you come up with four
1:15:38 different
1:15:44 concepts for how we might use this image
1:15:47 to raise
1:15:51 awareness and so
1:15:59 one of the things that I hear a
1:16:01 lot and and it's just the nature of sort
1:16:03 of how we've gotten here is
1:16:07 that one of the important things moving
1:16:09 forward is prompt engineering and they
1:16:11 keep using this term prompt engineering
1:16:13 you got to be a prompt engineer you
1:16:14 going to be a prompt engineer and human
1:16:17 beings hear that and they're like oh God
1:16:19 I can't engine I can't be an engineer I
1:16:22 I'm good you you know what I'll be over
1:16:24 here in the corner you can take my job I
1:16:26 just I'm not I'm no I don't want to
1:16:29 engineer anything right it's
1:16:32 intimidating but when you actually
1:16:34 interact with these tools it's you can
1:16:37 be a prompt engineer you can absolutely
1:16:39 take a systematic approach to breaking
1:16:42 down all the components of a prompt and
1:16:44 learning them and listen it doesn't hurt
1:16:46 you to learn how to do that especially
1:16:47 if you want to develop applications or
1:16:50 you want to do more sophisticated
1:16:52 versions of these gpts
1:16:54 understanding all that stuff's good but
1:16:56 you don't have to you can just talk to
1:16:58 this thing like you're talking to a
1:17:01 something I uploaded an image and I said
1:17:04 explain this in detail and it did and
1:17:06 then I said come up with some concepts
1:17:08 for how we might advertise like if you
1:17:12 don't know anything about advertising
1:17:14 you can just ask it hey I don't know
1:17:15 anything about advertising but that
1:17:17 seems like it'd be a good ad could we do
1:17:19 something with that the educational
1:17:21 series teaser the concept use the images
1:17:23 is a teaser For an upcoming series of
1:17:25 educational videos the future of AI in
1:17:28 everyday technology machine learning one
1:17:31 101 or AO AI in the world of
1:17:35 motorcycling I was going to say AI in
1:17:37 the world of Motorcycle Maintenance or
1:17:39 the Art of Motorcycle Maintenance might
1:17:40 be good interactive
1:17:43 challenge behind the scenes Insight
1:17:46 inspirational stories
1:17:49 um slogan hear the stories share the
1:17:52 passion AI
1:17:54 Transformations at the AI learning lab
1:17:56 not bad from concept to reality dive
1:17:58 into AI Innovations design the future
1:18:02 unlock the secrets AI of AI with the AI
1:18:05 learning lab series premiering soon
1:18:07 that's cool and by the way in the lower
1:18:09 leand corner there's a series there
1:18:10 called underhyped which is a series of
1:18:12 videos I created talking about why this
1:18:16 AI stuff is so
1:18:17 important um and how to use it so check
1:18:20 that out so I I like that one and I'm
1:18:23 I'm going to say
1:18:25 um that's great can
1:18:30 you
1:18:31 refine concept number
1:18:34 one with the basic idea of a play on the
1:18:37 words of the Art of Motorcycle
1:18:44 Maintenance uh Zen In The Art of
1:18:47 Motorcycle Maintenance
1:18:50 is what I was looking for to play on and
1:18:54 so yeah so it figured that out right so
1:18:57 it said here's a refined version of
1:18:59 concept one incorporating a play on the
1:19:01 words of the classic book Zen and the
1:19:03 Art of Motorcycle
1:19:04 Maintenance Ai and the Art of Motorcycle
1:19:08 intelligence it's fine it should be
1:19:11 Motorcycle Maintenance but that's okay U
1:19:13 the campaign the the visuals the slogan
1:19:16 discover AI all right so I'm going to
1:19:20 say great can you create a magazine
1:19:24 style
1:19:25 ad
1:19:27 that incorporates the copy that you just
1:19:30 wrote and the basic elements of the
1:19:34 image that I
1:19:36 uploaded and so now we're going to have
1:19:38 it make an ad and I I don't I have no
1:19:40 idea how it's going to do it it's
1:19:42 probably going to change the image up a
1:19:43 lot it's definitely going to have typos
1:19:45 in it because they they put a lot of
1:19:47 text up there but but notice what what
1:19:51 I'm doing here is
1:19:54 I'm seamlessly switching modes I'm
1:19:58 switching
1:19:59 from inquisitive mode hey what do you
1:20:02 see in this image to um strategy mode
1:20:06 you know I've got this thing called the
1:20:08 AI learning lab and I want people to
1:20:09 learn it era there seems to be a
1:20:12 technical issue with generating an image
1:20:15 uh try
1:20:19 again to go going into Tactical mode
1:20:23 give me some Concepts oh I like that one
1:20:25 concept let's refine that to to refining
1:20:27 mode and now we're going into image
1:20:30 generation mode we'll see if it if it
1:20:32 it's still choking it might just be
1:20:39 busy let me drink my
1:20:45 water are you talking about Sydney don't
1:20:48 make me
1:20:49 cry just tried to upgrade and unable to
1:20:52 do it due to it being paused yeah
1:20:54 unfortunately Jasmine so get on the wait
1:20:57 list yeah it's it's it's not happy right
1:21:00 now technical difficulties it seems to
1:21:03 do this more more than I'd like it to
1:21:07 try
1:21:10 again um so chat GPT right now you
1:21:15 cannot upgrade to
1:21:17 gp4 um some people were selling their
1:21:20 accounts on eBay it looks like all of
1:21:22 those are gone and now it's just
1:21:23 scammers up there so I I I would be
1:21:26 careful about looking on eBay for these
1:21:29 things um but it'll it'll open up soon
1:21:32 enough again uh it seems like open AI is
1:21:35 back to being semi-stable after their
1:21:38 kurur fuffle last week there was a uh a
1:21:41 letter nah it's it's not generating
1:21:43 images so we'll just try we'll just play
1:21:45 with something else I'll show you how to
1:21:46 do uh some other stuff I'll create a new
1:21:49 chat GPT show you one other thing it can
1:21:52 do
1:21:56 um but anyway get on the wait list for
1:21:59 this in my opinion the 20 bucks for chat
1:22:02 GPT plus is the best deal in software
1:22:05 right now what what this tool can do is
1:22:08 is just staggeringly powerful and it's
1:22:11 like even in me like I'm taking whatever
1:22:13 the bulk of an hour here to just walk
1:22:15 through some of the um some of the
1:22:20 things right don't make me cry ch for
1:22:23 ramsy why cuz I'm using your
1:22:26 image
1:22:28 um that's pretty funny
1:22:32 um just like I'm I'm not even scratching
1:22:36 the surface of what this makes possible
1:22:38 and and we haven't even got to gpts yet
1:22:41 um so so let me show you another thing
1:22:43 it can do um and I know if if you don't
1:22:45 have access to it it's like yeah Kyle
1:22:47 way to way to create the fomo dude way
1:22:51 to create the fomo what is this like an
1:22:52 nft Channel you can make money
1:22:55 with yeah why why are you asking Bob I
1:22:59 don't know I I heard AI was going to be
1:23:02 cool I heard it was bigger than nfts
1:23:04 shut up Bob nfts suck and so do you yeah
1:23:07 but AI is cool shut up no you shut
1:23:10 up does he act like this a lot oh yeah
1:23:14 mhm he does why does he do that nobody
1:23:17 seems to know okay um so one of the
1:23:23 other things that is now integrated
1:23:25 inside chat GPT is a thing
1:23:28 called Advanced Data analysis M sexy or
1:23:32 code interpreter M obscure or whatever
1:23:36 name they come up with that sucks less
1:23:37 than those two it started out as code
1:23:40 interpreter then it turned into Advanced
1:23:42 Data analysis which is technically what
1:23:44 it's doing but it's it's just kind of
1:23:46 magic so what is it how does it work I
1:23:49 am going to go to a different site right
1:23:52 now
1:23:53 called
1:23:54 kaggle k a g g l
1:23:58 e if you're if you're if you're
1:24:00 hankering for a data set come to
1:24:03 kaggle
1:24:06 um so what kaggle has is lots of data
1:24:10 sets so people upload their CSV files of
1:24:12 all sorts of data here and there's I'm
1:24:14 sure there's other stuff you can do I'm
1:24:16 not a data analyst so uh I really don't
1:24:19 know what I'm doing
1:24:21 here detailed products data
1:24:24 sets
1:24:26 102k usability
1:24:29 10 multiplatform online courses data set
1:24:34 okay this is good let's do this four
1:24:36 file four CSV files three megabytes so
1:24:39 let's go grab
1:24:41 this so I'm going to click on this and
1:24:44 I'm going to download this 3 megabyte
1:24:46 zip file archive. zip that's going to be
1:24:49 my downloads folder okay and we're going
1:24:52 to come come back over here and we're
1:24:53 going to
1:24:57 upload archive.
1:25:02 zip so we just know from the description
1:25:05 it's something about courses
1:25:06 multiplatform courses nonetheless this
1:25:09 is not mono platform oh no this is
1:25:13 multiplatform multiplatform
1:25:17 courses in for CSV files so we know that
1:25:22 but it doesn't know that so I'm just
1:25:24 going to
1:25:25 say
1:25:28 what's in
1:25:30 this
1:25:33 archive and what might be interesting
1:25:37 about
1:25:38 it question
1:25:41 mark and so see where it says
1:25:44 analyzing you're like okay yeah it's
1:25:46 analyzing that's cool it's just looking
1:25:48 at it well how do you look at data well
1:25:52 oh error analyzing it's breaking oh chat
1:25:55 GPT please don't
1:25:58 break please be functional on your
1:26:03 birthday what' you do this morning I
1:26:06 watched some old guy try to use a
1:26:08 machine it was it was he couldn't do
1:26:12 it all right it did something good good
1:26:18 Lord all right so I'm gonna I'm going to
1:26:20 show you let's see what might be
1:26:24 interesting course Data Insights
1:26:25 comparative
1:26:27 analysis Trends and patterns so so what
1:26:30 it says is they appear to be CSV files
1:26:33 typically used for storing tabular data
1:26:35 given their names it seems there's all
1:26:38 this stuff so it explains what's in
1:26:40 there would you like to analyze this
1:26:42 data further if so specify specify the
1:26:45 particular analysis so let's see
1:26:49 um I'd like to see a comparative
1:26:52 analysis as well as Trends and patterns
1:26:54 period I'd like you to
1:27:02 identify how these data sets compare
1:27:05 what the trends are and then visualize
1:27:08 what you think are the most relevant
1:27:10 trends that you
1:27:14 find so that is something that you would
1:27:17 ask a data analyst right and so now when
1:27:20 it says analyze I'm I'm going to show
1:27:21 you what's actually going on I'm going
1:27:22 to open
1:27:24 this what what this tool
1:27:28 is this is [ __ ]
1:27:30 astounding what this tool is is it
1:27:33 inside of your chat when it says
1:27:36 analyzing it creates what's called a
1:27:38 container it creates think of it like a
1:27:40 little mini it's almost like a virtual
1:27:43 machine it's like its own little
1:27:45 Computing environment and it's got
1:27:47 access to
1:27:48 Python and so it writes and executes its
1:27:51 own python so it's prompting itself and
1:27:54 then it's writing this code and taking
1:27:57 the results of that and doing other crap
1:28:00 with
1:28:01 it um so so it it used python to figure
1:28:04 out that and now it's it's opening the
1:28:07 Panda's Library it's looking at these
1:28:09 csvs it's failing when it fails let's
1:28:13 see what it it looks like it's hanging
1:28:15 for some reason so this could be just
1:28:17 chat be doing it like doing weirdness um
1:28:22 but it's it's um it's writing and
1:28:25 executing its own python code so it
1:28:28 appears that there's an issue accessing
1:28:30 the CSV files could be due to a mismatch
1:28:32 in the file
1:28:34 path yeah something's weird let's Pro
1:28:37 okay I will extract the contents of this
1:28:40 ZIP file to a temporary directory and
1:28:43 then read the CSV files it should help
1:28:46 avoid any file path related issues let's
1:28:49 proceed with this method so notice this
1:28:52 it's it's not only writing and executing
1:28:54 its own python code it's doing its own
1:28:56 troubleshooting
1:28:58 so this starts to be a glimpse into the
1:29:02 future of how these large language
1:29:03 models are going to work where you're
1:29:06 not going to have to do very much work
1:29:08 at all this is why I was talking earlier
1:29:10 about um prompt engineering is like you
1:29:14 you don't have to worry about being a
1:29:16 prompt engineer these tools are going to
1:29:17 get more and more sophisticated where
1:29:19 all you're going to have to do is say
1:29:21 hey I want to know [ __ ] and it will go
1:29:26 generate its own agents on your behalf
1:29:28 that will go find you [ __ ] and bring it
1:29:31 back to you and that's effectively what
1:29:33 this is doing right now so let's see
1:29:36 what did it do here it
1:29:38 said okay there was an issue the data
1:29:41 from each platform has been successfully
1:29:43 loaded and can be briefly described as
1:29:47 follows the corsera data includes the
1:29:50 course provider the course name the the
1:29:52 skills covers the ratings the review
1:29:54 count the level blah blah blah udemy
1:29:58 contains titles descriptions right so
1:30:00 just even figuring out the
1:30:04 um the data that's in these files would
1:30:07 probably take you 10 or 15 minutes to
1:30:09 open the file look at the columns maybe
1:30:12 you put them as multiple worksheets in
1:30:15 the same spread sheet you you know look
1:30:19 at inconsistency of data comparative
1:30:23 analysis course offerings popularity
1:30:26 course level and duration and then it
1:30:28 looks like it crapped out but here we go
1:30:31 let's begin with basic comparative
1:30:33 analysis and move on before identifying
1:30:35 Trends and patterns uh and then it it
1:30:37 crapped out but but um before it crapped
1:30:40 out it wrote The Python code and
1:30:43 generated this
1:30:46 chart it didn't this chart was not
1:30:48 pulled from anything this was not stolen
1:30:52 from from the internet it analyzed the
1:30:54 data in in these three CSV
1:30:57 files and and did
1:31:01 this I I'll show you another thing that
1:31:04 that'll blow your mind well blows my
1:31:10 [Music]
1:31:21 mind
1:31:28 hang
1:31:43 on
1:31:45 okay so here here's a screenplay I wrote
1:31:48 30 years
1:31:50 ago so it's a feature Lang
1:31:52 I think it's 126 Pages something like
1:31:54 that I think we probably we reformatted
1:31:57 it we tightened everything up so it'd be
1:31:59 120 because that's what screen plays
1:32:01 need to be so it's 120 page
1:32:04 screenplay
1:32:05 um so we're going to upload
1:32:10 that all right and I'm going to say
1:32:18 um please write a brief synopsis of this
1:32:21 screenplay
1:32:23 followed
1:32:25 by a Hollywood elevator pitch which
1:32:29 Compares two different popular movies
1:32:32 together to describe this
1:32:37 one and so it's going to go read my
1:32:41 screenplay so so there are people whose
1:32:43 job it is to do
1:32:46 this they get screenplays from wannabe
1:32:51 Movie Makers like like myself and
1:32:53 they're like all right Sally go read
1:32:57 this and then do what's called writing
1:33:00 coverage right give me some coverage on
1:33:03 this and so the coverage might be well
1:33:07 what's the basic story about um who are
1:33:10 the characters what's this movie
1:33:12 like what's what's the what's the
1:33:14 category I can think about this
1:33:17 in and hopefully it doesn't crap out it
1:33:20 it seems like it's a little unhappy C
1:33:22 TPT is a little unhappy right now which
1:33:24 I assume is one of the reasons they've
1:33:26 paused it paused new subscriptions um
1:33:30 salmon flying East is a gripping
1:33:32 screenplay that intertwines the themes
1:33:33 of Espionage personal betrayal and the
1:33:36 lingering aftermath of War set
1:33:38 predominantly in 1991 with flashbacks to
1:33:41 the Vietnam War era the story revolves
1:33:43 around Jack a former pilot and Derek a
1:33:45 mysterious figure from his past the
1:33:48 narrative unfolds on Jack's fishing boat
1:33:50 in the Puget Sound through a series of
1:33:52 Revelations about their shared history
1:33:54 in Vietnam so gives me nice coverage of
1:33:57 what it is I I I won't I won't uh kill
1:34:01 it for you at the end there elevator
1:34:02 pitch salmon flying East is described
1:34:05 can be described as the born identity
1:34:07 meets the deer hunter it combines
1:34:10 intricate Espionage and identity crisis
1:34:13 of the born identity with the emotional
1:34:15 depth and haunting effects of the
1:34:17 petrol so all right so there's our
1:34:20 elevator pitch
1:34:22 um so I could use that for marketing I
1:34:24 could say
1:34:30 um who are film producers you know that
1:34:33 you
1:34:34 think might enjoy a script like
1:34:39 this I'll see if it knows
1:34:42 any I don't have personal
1:34:44 knowledge all right it's just giving me
1:34:46 generic things okay uh all right so I'm
1:34:49 going to ignore that um let's let's go
1:34:51 uh
1:34:52 okay great give me a character
1:34:57 breakdown of who the main characters
1:35:03 are a brief
1:35:06 description of their
1:35:09 Persona
1:35:11 and three possible casting
1:35:20 recommendations so it's gonna so it's
1:35:22 now going off and reading my screenplay
1:35:24 analyzing the
1:35:26 characters and then going to give me
1:35:27 casting casting recommendations I did
1:35:30 this the other night it was actually
1:35:31 they're actually pretty good um and it
1:35:35 was you know was some ACT actors I'd
1:35:37 heard of and it was others I hadn't and
1:35:39 when I looked them up I'm like yeah that
1:35:40 that tracks that makes
1:35:46 [Music]
1:35:50 sense
1:35:53 [Music]
1:36:01 Jack Jack is a former pilot with a
1:36:04 complex past he's stoic carrying the
1:36:06 weight of the past casting
1:36:08 recommendations Hugh Jackman Liam niss
1:36:10 Jeff Bridges
1:36:12 perfect Derek dereck's a mysterious
1:36:15 mysterious figure from his past he
1:36:17 reveals intertwined past that they share
1:36:20 Derek is a catalyst of unfolding Dr
1:36:22 deeply involved in the complex web of
1:36:24 Deceit Jeremy Irons Ed Harris perfect or
1:36:27 Gary Oldman
1:36:29 perfect mitsu a an amasian the alleged
1:36:34 daughter of Jack from alazon during the
1:36:36 Vietnam
1:36:38 war right and then so that so gave three
1:36:41 and I could I could go on and get more
1:36:43 um okay so now I'm going to do one more
1:36:45 thing create a table with the five most
1:36:49 important visual moments of the
1:36:54 film where the scene is described in the
1:36:57 left column and the
1:37:00 visual is described in the right as if
1:37:04 it were describing a still from the
1:37:09 film so if I wanted to Market this
1:37:12 screenplay which I I got to tell you
1:37:14 something that's wild about this so so
1:37:17 this this happened how I figured out how
1:37:20 to do this with screenplays was four
1:37:22 nights ago someone said hey Kyle you
1:37:25 said chat jpt would be good at
1:37:26 storyboarding what do you mean by that
1:37:28 could you how do you do that I was like
1:37:30 I don't know I was just talking
1:37:33 [ __ ] and so and so I just did this kind
1:37:35 of on the Fly and it it
1:37:40 just I I I find it hard to
1:37:44 comprehend that we all have access those
1:37:49 of us that have access to gbd4 before
1:37:52 but you know we're all going to at some
1:37:53 point that that there are Technologies
1:37:56 in the world right now that can do
1:37:59 this it it's staggering it's
1:38:03 staggering all right so I don't know why
1:38:05 it's doing code analysis right now I
1:38:07 guess it's just reading the
1:38:09 screenplay what is it doing let's go
1:38:12 look at it you can open these things oh
1:38:15 it's it's create a table
1:38:18 okay so it's creating the table it's
1:38:21 it's Reddit now it's using python to
1:38:23 create the table I think that's Overkill
1:38:25 but let's see what it
1:38:27 does it might give us a download a
1:38:30 downloadable
1:38:39 thing all right do me a favor while this
1:38:41 is doing this um if you'd be so kind
1:38:44 share this live like this live double
1:38:46 tap on the screen um follow my channel
1:38:50 um if you want to support the channel if
1:38:51 you like what I'm up to here cool I'm
1:38:53 going to be going live periodically
1:38:55 throughout the day so if you follow the
1:38:56 channel you'll get notified if you
1:38:58 subscribe to my lives that'll support me
1:39:01 would appreciate that in the lower
1:39:03 leftand corner is a series that I put
1:39:05 together called underhyped which talks
1:39:08 about why I think that AI is actually
1:39:11 underhyped not overhyped um and it's
1:39:14 it's a whole series of videos that talk
1:39:16 about um how AI today comp compares to
1:39:20 the early days of the worldwide web how
1:39:22 these tools actually work um how to use
1:39:25 them how to prompt them how to not get
1:39:27 overwhelmed by them which is essentially
1:39:29 impossible at this point but I do have
1:39:31 advice on that um so if that sounds
1:39:33 interesting to you that's a series down
1:39:35 in the left hand leftand corner um and
1:39:38 yeah so anything you can do to support
1:39:39 the channel great okay scene description
1:39:42 so there's visual description so Jack
1:39:45 and Derek confront their past on the
1:39:47 fishing boat revealing secrets about
1:39:49 their time in the Vietnam War a tense
1:39:51 scene where J on the deck of a fishing
1:39:53 boat in the pet sound jack with a rugged
1:39:55 and contemplative look listens intently
1:39:58 to Derek who appears weary and burdened
1:40:00 by his past the water is calm reflecting
1:40:03 the Cloudy Sky adding a somber mood to
1:40:06 the
1:40:07 conversation so now I can
1:40:11 say please make a 16
1:40:17 by9
1:40:19 wide image
1:40:22 for each of those
1:40:25 scenes in a
1:40:28 storyboard
1:40:30 style that looks like
1:40:32 watercolors but is clearly a storyboard
1:40:35 for a
1:40:38 film leave it at
1:40:41 that and it's now going to make all that
1:40:45 [ __ ] Kyle needs to get going to the
1:40:48 office soon but he'll be back online in
1:40:49 a couple of hours are people asking me
1:40:51 all sorts of questions I'm ignoring this
1:40:53 is one of the problem one of the
1:40:54 problems with the channel is is when I
1:40:57 when I go down my rabbit holes to
1:40:59 explain [ __ ] I miss all the comments so
1:41:02 I do have mods that are pinning things
1:41:04 for me but if I'm not looking at the
1:41:06 screen they they ain't [ __ ] being uh
1:41:09 replied to um oh that's so cool look at
1:41:14 that it's the wrong kind of fishing boat
1:41:16 for the pet sound but that's
1:41:20 okay you got your close-ups you got your
1:41:22 over the
1:41:26 shoulders and now it's creating the next
1:41:28 image so so this is now storyboarding my
1:41:35 film what's the second one that's
1:41:41 intense Ping's assassination in Vietnam
1:41:44 witnessed by Derek and young mitsu a
1:41:46 hering scene in a war torn Vietnamese
1:41:49 alley Derek cradles a dying ping in his
1:41:52 arms while young mitsu watches from a
1:42:03 distance cuckoo
1:42:06 right and I mean what's what what what
1:42:08 blows my mind about this
1:42:13 is like like these scenes here with the
1:42:16 exception of that boat being
1:42:18 wrong this is kind of what I had in my
1:42:20 head when I wrote it it's like like I
1:42:24 would describe that scene to a
1:42:26 storyboard artist and you know there
1:42:28 might be different ways you'd format it
1:42:29 in different ways you'd call things out
1:42:32 but like this is in the
1:42:44 neighborhood I don't know why we've got
1:42:46 cars from the 1950s
1:42:50 here
1:42:54 all right that's great so let's let's
1:43:01 do and so here now it's creating links
1:43:03 so I can I mean I can rightclick those
1:43:05 and download them but it's it's actually
1:43:06 creating a series of links to those
1:43:13 images I'm not going to start writing uh
1:43:16 I'm not going to create a GPT right now
1:43:18 just because it's going to take too long
1:43:19 to do it um but needless to say um
1:43:23 there's a lot of power here um I I'll do
1:43:25 one other
1:43:28 image great make that last
1:43:31 image photographic and
1:43:34 cinematic as if it's actually a
1:43:37 still from the production period I'd
1:43:41 like to see what that final scene looks
1:43:44 like on
1:43:46 film
1:43:48 period make the lighting dramatic
1:43:52 the set decoration
1:43:54 impeccable and the overall image
1:44:00 Unforgettable and that my ability
1:44:04 to just talk to this thing like a human
1:44:08 being and let it figure out whatever the
1:44:10 [ __ ] that
1:44:13 means again it just it melts my
1:44:19 brain if you know how comp computer's
1:44:21 work it doesn't make sense that you can
1:44:24 talk to it like
1:44:25 this we've been talking to Siri and
1:44:30 Alexa for a decade now
1:44:32 right and like I didn't get that I know
1:44:35 you didn't that's the whole
1:44:37 point and and they just don't understand
1:44:43 stuff
1:44:45 eh it's okay it's not quite it's not
1:44:48 quite the image I wanted but that's fine
1:44:50 um
1:44:54 and and we can just do this now and and
1:44:56 you know the implications of this are
1:44:58 are just I like I I don't know and my my
1:45:02 problem right now
1:45:05 is I think a problem all of us are going
1:45:07 to face is here's what it is as much as
1:45:11 I act like I know what I can promise you
1:45:14 is that I am [ __ ]
1:45:17 clueless we all are no matter how if
1:45:21 you've been in machine learning for 30
1:45:23 years we are in entirely new territory
1:45:27 here every technologist I talk to that
1:45:31 has been in Ai and machine learning for
1:45:33 a long long time I say well tell me like
1:45:37 what's different now that generative AI
1:45:40 is here and the response I hear over and
1:45:42 over and over again is everything's
1:45:45 different nothing's the same how we code
1:45:49 is different how we write is different
1:45:52 how we're going to make images is
1:45:54 different how we're going to make films
1:45:55 is different you want to know something
1:45:57 wild the there there's a um there are
1:46:01 these uh text to video programs like
1:46:04 Runway ML and P Labs that can take still
1:46:08 images like this and turn them into
1:46:10 videos so one of the things I'm going to
1:46:13 do with this screenplay is I'm going to
1:46:15 have it write a trailer like a
1:46:18 promotional trailer for the film and
1:46:20 describe what the visuals
1:46:22 are as the trailer goes on I'm going to
1:46:26 generate the
1:46:27 images I'm going to take him over to one
1:46:29 of those video things turn the still
1:46:31 images into videos and then I'm going to
1:46:34 narrate the um the trailer with films
1:46:40 from from this stuff my friend is a
1:46:43 developer at a major news outlet and
1:46:45 told me they are all quietly panicking
1:46:49 yeah and quite frankly lunar stick I
1:46:51 like I think that they should noisily
1:46:54 Panic right if if he's a developer to
1:46:58 Major news organization major news
1:47:00 organizations are I got I'll tell you a
1:47:04 funny story so let me let me flip this
1:47:06 over here I'll tell you a funny story
1:47:08 lunatick um and you've you've been on
1:47:11 these lives I might have I might this is
1:47:12 one I might have told but in
1:47:15 1996 I got invited to speak at the
1:47:20 national newspaper Publishers
1:47:23 association meeting annual meeting cuz I
1:47:26 was I started agency.com this this fancy
1:47:30 digital agency and we were doing you
1:47:32 know internet stuff and they're like
1:47:34 let's bring in the internet
1:47:36 expert and so I I was a a cocky 30
1:47:41 something 30-some guy that thought my
1:47:43 [ __ ] didn't stink and so I I talked a
1:47:46 little bit about the worldwide web and
1:47:48 why I thought it was important and then
1:47:50 I I walked around the podium and I sat
1:47:53 on the front of the stage and I said
1:47:55 very matter of factly to the audience um
1:47:59 you know as an
1:48:01 industry I would encourage you to
1:48:04 explore the blurring of the line between
1:48:07 advertising and editorial because
1:48:10 because it's coming whether you want it
1:48:11 or not I was escorted out of the meeting
1:48:15 by the by the main boss dude he was like
1:48:18 all right thanks very much Kyle Shannon
1:48:20 uh follow me he walked me off stage
1:48:23 walked me to the back door and said uh
1:48:26 I'll call you a
1:48:27 [Laughter]
1:48:30 cab uh yeah so the newspaper industry um
1:48:34 they're going to go Kicking and
1:48:35 Screaming but you know
1:48:37 listen they they
1:48:41 uh they're they're dealing with you know
1:48:44 the implications of the worldwide web
1:48:46 stuff so so this AI stuff is is uh crazy
1:48:49 but what I would say to your friend is
1:48:50 is you know they should be actively
1:48:53 saying we should be having these
1:48:55 conversations now like like put me on a
1:48:58 team to explore AI enough to put us out
1:49:03 of business like like why don't we learn
1:49:06 internally at this media organization
1:49:08 how to [ __ ] things up because they're
1:49:11 going to get [ __ ] up whether we do it
1:49:14 or whether it comes from a couple of
1:49:16 Hoodie wearing punks with a case of Red
1:49:18 Bull that whipped up a new media
1:49:20 organization over the weekend you want
1:49:23 to see something
1:49:25 wild it won't look wild but I think it's
1:49:30 wild I did a LinkedIn post on this go to
1:49:35 runway.
1:49:41 [Music]
1:49:44 TV
1:49:49 shoot
1:49:52 [Music]
1:50:01 all right it should play a video here we
1:50:03 can go look at one if it
1:50:07 doesn't stream offline all right okay
1:50:13 so this kind of looks like an early
1:50:18 version of Netflix right
1:50:21 where there's like you got the channels
1:50:23 and you got the shows to the right well
1:50:25 what are these shows so these
1:50:28 shows are I'm going to try to see if I
1:50:31 can get one that I've seen before I
1:50:34 think this one I've seen before
1:50:37 um so these are videos that
1:50:40 are generated the at least some of the
1:50:45 Clips in these videos have been
1:50:46 generated using Runway
1:50:49 ml okay
1:50:51 [Music]
1:51:01 they're kind of janky and they're kind
1:51:03 of
1:51:05 weird I'm just going to hit the the uh
1:51:08 random
1:51:09 button you got me on this channel can
1:51:12 you hear me I'm receiving you what
1:51:15 happened I haven't heard from you in
1:51:17 like 4 hours s sort of janky acting
1:51:20 Escape pod on my ship I didn't get your
1:51:22 exact coordinates before I lost contact
1:51:24 with your signal station are you working
1:51:26 synthesized acting give me the
1:51:28 coordinates and I'll head that way my
1:51:31 station no I'm on the
1:51:34 ground what do you mean you're on the
1:51:36 ground I mean on the ground me and my
1:51:39 family are stuck in a bomb shelter
1:51:41 because half the planet is being
1:51:42 destroyed by all the junk Landing
1:51:44 here you're on the ground like all right
1:51:48 so that's that
1:51:50 [Music]
1:51:52 some weird
1:51:54 hands some weird
1:51:57 movements some weird interactions
1:52:00 although that was actually not too bad
1:52:02 weird
1:52:03 hand okay so this is all janky ass [ __ ]
1:52:07 you're like okay what are okay Kyle what
1:52:10 I don't get it what what is so this is
1:52:12 like so you're telling me this is like
1:52:15 like shitty Netflix that's exactly what
1:52:18 I'm telling you and what's going to
1:52:21 happen is Netflix is going to dismiss
1:52:24 this and Hollywood studios are going to
1:52:27 dismiss this because they're going to
1:52:28 look at it and they're go well those are
1:52:29 janky ass pieces of [ __ ] and they don't
1:52:31 have stars in them and they
1:52:34 don't and what's going to happen
1:52:37 is sites like this are going to fly
1:52:40 under the radar where people are going
1:52:42 to dismiss them and what's happening
1:52:44 with this generative AI technology is
1:52:46 it's going from laughably bad to
1:52:49 terrifyingly good good in like a year
1:52:51 and a half and video right now is right
1:52:54 at that inflection point pabs p i Labs
1:52:59 uh if you go to Pika do is it p.
1:53:03 a hang
1:53:05 on p. art go to p.
1:53:10 art
1:53:12 um this technology that they're about to
1:53:15 launch that I'm on the wait list for get
1:53:17 on the wait list for this um is taking
1:53:21 it to a whole another level you're going
1:53:22 to be able to upload you'll see in a
1:53:24 second here there's an existing video
1:53:27 clip of a woman walking and they
1:53:29 highlight her top and just change her
1:53:31 clothes in real time while she's walking
1:53:35 it's like are you kidding
1:53:40 me so what's going to happen
1:53:42 is every Media company that has their
1:53:45 head in the sand where where you know
1:53:47 the the the tech team is quietly
1:53:51 panicking look at this right here they
1:53:54 just swap out her
1:53:55 shirt pop on some glasses onto the r
1:54:00 Tang just type it this is called idea to
1:54:04 video is the the Pika
1:54:07 1.0 so so the technolog is coming that
1:54:11 I'll be able to take my screenplay that
1:54:13 I just
1:54:15 storyboarded and probably
1:54:18 within if I want to work for it within a
1:54:21 year I can make that some version of my
1:54:24 film and within 3 years I can probably
1:54:27 easily make a good version of salmon
1:54:30 flying
1:54:31 East on a weekend by
1:54:34 myself including soundtrack including
1:54:38 acting including writing including
1:54:42 promotion all of
1:54:44 it and so sites like runway. TV that
1:54:49 write now look
1:54:52 laughable and kind of janky and
1:54:55 like is is this just like a webflow
1:54:59 template with like with like YouTube
1:55:02 thumbnails in it pretty much it's pretty
1:55:05 much what it
1:55:06 is um these are going to get very
1:55:09 serious very fast and I think it's going
1:55:12 to I and I think I think this is
1:55:15 illustrative of what's going to happen
1:55:17 in every industry where the big
1:55:19 incompass are going to think that they
1:55:21 have 3 to five years to deal with this
1:55:23 [ __ ] and they'll be like ah we'll get to
1:55:26 it it's not there yet I hear it
1:55:27 hallucinates it's it's it's laughably
1:55:30 bad I looked at one of those
1:55:33 movies it's it's not
1:55:36 professional
1:55:38 okay
1:55:40 okay
1:55:43 um I I I I just I think it's going to be
1:55:47 terrifying how fast the centers of power
1:55:51 in each different industry are going to
1:55:53 be
1:55:56 disrupted so if you work for one of
1:55:59 those companies I would start being the
1:56:01 squeaky wheel just let people know
1:56:03 you're working on this AI
1:56:06 stuff but they think I'm weird it's our
1:56:08 policy not to use AI all right go tell
1:56:11 them you want to be part of a task force
1:56:12 that that explores
1:56:15 AI because you think it's
1:56:18 important and let them give that role
1:56:21 why not what's the worst could happen no
1:56:24 you can't do that all right I'll go back
1:56:26 to my
1:56:27 cubicle and use AI to do my work in 4
1:56:30 hours that you're paying me 40 hours to
1:56:32 do and I I'll shut the [ __ ] up about it
1:56:34 and then in the weekends I'm going to
1:56:36 create six different businesses because
1:56:38 I can and one of those might be a threat
1:56:41 to the company I work
1:56:46 for you know there's there's there's a
1:56:48 lot of uh there's a lot of talk of oh AI
1:56:52 is going to kill all these jobs but you
1:56:53 know what else AI is going to do it's
1:56:55 going to give a lot of individuals
1:56:58 disproportionate
1:56:59 power to what they have been able to do
1:57:06 historically there's
1:57:09 me 20 24 hours of chat GPT all right
1:57:14 listen I got to go get ready for work I
1:57:17 got to go to work I got this 9:00
1:57:18 meeting I'll be back on at I wonder how
1:57:20 Adobe will play a part in this so Adobe
1:57:23 right now is actually leading Luna so
1:57:27 um go go to
1:57:30 firefly.com that's that's adobe's
1:57:33 generative uh AI
1:57:36 tools if you've got access to
1:57:39 photoshop uh go watch videos I'll I'll
1:57:42 I'll share it later if you're still
1:57:44 here um on generative fill so you can
1:57:49 now take a Photoshop image and like
1:57:51 let's say you have a vertical image of
1:57:52 someone standing at the beach and you
1:57:55 want the whole beach scene but you want
1:57:57 it wide you can just expand the image
1:57:59 and it'll fill in the stuff that's
1:58:01 that's not there that's that's there
1:58:03 right now what's interesting about Adobe
1:58:05 is they didn't train any of their models
1:58:07 on um restricted copywritten work it's
1:58:11 it's only been done on um public domain
1:58:15 and Adobe stock content that they have
1:58:18 rights to they're also giving artists
1:58:22 the opportunity to opt out or opt into
1:58:25 models and they're also building oh
1:58:28 don't forget to follow Kyle will be back
1:58:30 at 9:00 a.m. Central Time no I'll
1:58:32 actually be back at 9:00 a.m. mountain
1:58:34 time I'll be back in less than an
1:58:37 hour
1:58:38 um okay um so anyway so Adobe is they're
1:58:44 leading on this and they're getting
1:58:46 really aggressive about it so John Knack
1:58:48 who works for the Adobe a AI team has
1:58:50 spoken twice now at the AI Salon oh the
1:58:53 AI Salon let me before you go if you're
1:58:57 interested in joining a community that
1:58:59 is up to this stuff this channel is kind
1:59:03 of an extension of this community that
1:59:05 started a year ago our one-year
1:59:06 anniversary is next Tuesday if you go to
1:59:09 AIS salon. mn. that's our Mighty
1:59:11 networks's Community go there check it
1:59:14 out check out about the
1:59:16 salon and about the salon will tell you
1:59:19 what the salon's all about what our
1:59:21 values are check check out those values
1:59:23 if you resonate with them great it'll
1:59:26 probably be a fit for you there's guilds
1:59:29 here where you can join different guilds
1:59:31 there's a a starters Guild there's a
1:59:33 technical Guild there's a writing Guild
1:59:35 an Art Guild um think of the AI Salon as
1:59:39 like a a support group it's like it's
1:59:43 like a AI AA it's like hi my name is
1:59:47 Kyle and I'm into AI hi
1:59:50 Kyle it's been 10 minutes since I've
1:59:52 made an image in DOI oh we've been there
1:59:55 yeah yeah I'm making one right now on my
1:59:59 phone you better get some rest now I got
2:00:01 to go to work there's there's no no rest
2:00:03 for the wicked in this in this town so
2:00:05 anyway go check that out uh join if you
2:00:07 think it's interesting come to our uh
2:00:10 anniversary next Monday um I will be
2:00:13 back in less than an hour I'm going to
2:00:14 head to work it's 10 minutes that way
2:00:17 and uh yeah I got to finish getting
2:00:19 ready
2:00:20 peace out hope you enjoyed this thank
2:00:23 you all mods for for showing up at
2:00:25 whatever hour I seem to be going online
2:00:28 I don't know how you're doing it but but
2:00:30 thank you until next time Clint
2:00:32 Copperhead thank you very much thank
2:00:34 thanks mods uh I I'll catch youall very
2:00:37 shortly