aiLL 6/3/23 - 1 - Turning AI Anxiety into AI Action

Description

Chapters

Transcript

0:01 foreign
0:21 with the AI learning lab
0:24 hope you're all doing well on this fine
0:26 Saturday uh what we do here at the lab
0:29 is talk about AI stuff
0:31 and if you're new to AI that is the
0:34 official chat GPT
0:36 website
0:38 chat.openai.com I'm not affiliated with
0:40 it
0:41 but that's the real deal that's the
0:44 official Place uh if you have questions
0:45 about AI pop them in the comments
0:48 there's just a couple of us here now so
0:50 wait for the others to come in but hey
0:52 hey everybody
0:54 and
0:55 um let's see what else I don't know I am
0:59 a entrepreneur from for many years
1:02 started 10 or 12 companies uh I was a
1:07 Pioneer I suppose in the early days of
1:09 the web started one of the first digital
1:11 agencies company called agency.com
1:14 and one of the reasons I started this
1:15 channel is that I see some pretty
1:18 remarkable parallels between what
1:21 happened with the world wide web and
1:23 what's happening now with generative Ai
1:25 and uh that's why I started this channel
1:27 to
1:29 I don't know
1:30 learn as much as I can share as much as
1:32 I learned things like that hard to see
1:34 comments with your cool shirt well sorry
1:37 about that here I'll try to keep it here
1:39 what can I do here I can
1:41 put it up a little bit yeah I don't know
1:44 that's bad I guess I could change shirts
1:46 what else what else do I have I have a
1:49 little black gift bag here
1:50 I can just hold this right here
1:52 [Laughter]
1:55 if you have questions about AI pop them
1:57 in there I'm happy to answer them
2:00 um trying to think what what went on
2:02 like so much goes on every week that
2:04 it's almost impossible to keep up with
2:06 with anything anymore uh there's just
2:09 there's just gargantuan amounts of stuff
2:11 going on
2:13 um if you are
2:16 um
2:17 try to think what's what's worth worth
2:20 talking
2:21 if you haven't used
2:23 um chat GPT before this is chatgpt
2:27 and
2:29 I have chat GPT plus which basically
2:33 means you get access to gpt4
2:36 which is there's two little tabs at the
2:39 top one is gpt3 one is gpt4 and then if
2:43 you pay for the plus thing you also get
2:45 access to
2:47 um chat gbt being connected to the
2:50 internet but also with plugins
2:54 um and those are both really cool so
2:56 when chat EPT is connected to the
2:58 internet you can ask it things like
3:00 how much is it supposed to rain
3:07 in Denver today question mark
3:12 and it will go browse the web for you so
3:16 it is out there hello Liza how are you
3:18 good to see you uh if you're new here
3:21 pop any questions you have about AI into
3:23 the comments this is the AI learning lab
3:25 looks like a few more people are showing
3:27 up it's a Saturday I don't know why
3:29 you'd be hanging out here but it's cool
3:31 that you're hanging out here some have
3:32 mentioned GPT
3:34 for performance is changing or declining
3:38 opinion
3:40 um
3:42 for the stuff I do which which is it
3:45 it's not super intense you know research
3:47 or coding or things like that
3:49 I haven't seen a performance a quality
3:52 dip but I have seen a performance dip
3:56 where it's slow
3:58 and it's uh it's it's been a little
4:01 cranky for the past week and a half for
4:04 two weeks
4:05 um
4:06 some of my zapier automations are
4:08 failing for for two reasons for two
4:10 reasons zapier has some sort of bug with
4:12 Google forms which is a pain in the ass
4:15 um and it fails for that and then
4:17 sometimes open AI API calls are failing
4:21 so so my zapier automations for the past
4:24 couple of weeks have been in a bad place
4:26 which sucks
4:29 um
4:30 is there a free way to access
4:32 mid-journey
4:33 no well you can you can go to Mid
4:36 journey and I think they give you some
4:37 amount of tokens for free but it's it
4:39 ends up being like
4:41 15 or 20 images it's not a lot
4:44 um
4:45 so there's not and I mean unfortunately
4:47 mid journey is still kind of the best
4:49 game in town although boy check this
4:52 thing
4:53 style something like Google Style
4:57 copy Google Style
5:03 um
5:04 image AI
5:08 what was it called
5:13 I forgot what was called
5:16 I think I have it in my Twitter here
5:18 somewhere hold please
5:23 style drop
5:25 Google style drop this stuff is amazing
5:30 check out
5:32 so
5:35 so you put in this little image here
5:38 like of a kid's drawing and then all of
5:41 these other drawings are prompted
5:44 like a butterfly in that style an F1
5:47 race car a moose so like that's one
5:50 style
5:52 here's a you know starry starry night as
5:55 the source image
5:57 and then wait hang on my little stand is
6:00 messed up
6:01 um
6:02 let me come over here so you can kind of
6:04 see both
6:05 so that's a a van Gogh thing then they
6:09 did the same painting but they pulled
6:11 all the blue out of it and it matches
6:13 that there's another painting of his
6:15 another
6:17 this one is really remarkable this is
6:19 like that sort of classic 90s
6:21 illustration style and they're just
6:24 using one image to the left and it does
6:25 all the things to the right so
6:27 I mean this this is one of those
6:29 technologies that is is just literally
6:32 going to change the game right because
6:34 you can take a single image and it
6:37 it kind of learns everything about it
6:39 the color palettes the style I think is
6:42 it this previous one here
6:46 how do you get back
6:57 yeah like that one like look at how that
7:00 little moose is in the car like it just
7:02 completely learns
7:04 the the style like it's it's just
7:07 remarkable what's coming
7:10 um so anyway that stuff's coming Eric
7:11 Sanders here yo Eric was shaking can't
7:14 get enough of me huh I'm sorry
7:17 I'm just I'm in too many places all
7:20 right
7:21 but that's awesome thank you for coming
7:24 wait is there a free way to oh that's
7:25 you Eric asking about mid-journey
7:27 there's not there's um what's that thing
7:29 called Blue Vine is is blue Vine the
7:32 thing hold please
7:33 blue Vine is that right
7:36 no that's Financial Services hang on
7:41 let me go to Discord
7:44 uh
7:46 Blue Willow
7:49 Blue Willow
7:51 so Eric if you go to Blue Willow
7:55 um I assume it's bluewillow.com but all
7:57 the all these AI companies have
8:01 um Blue Willow wait
8:05 Blue Willow
8:08 uh free AI art generator so it's blue
8:12 willow.ai so go to bluewillow.ai
8:16 and then it will point you to their
8:18 Discord so try our free beta we'll dump
8:20 you over to the Discord I think unless
8:23 they've changed it yeah
8:24 and then so this works like mid Journey
8:27 but this is built on on top of stable
8:29 diffusion but it's free I don't know how
8:31 they're paying for this I I really
8:32 honestly is it maybe this is just they
8:34 try to get a bunch of users
8:36 is there anything you want me to try or
8:38 test out while you are live
8:41 um
8:46 I don't know give me give me the
8:48 suggestions possibly I look like Rosie
8:50 O'Donnell thank you very much uh someone
8:52 someone else said let me show you the
8:53 the full grandeur
8:56 I'm Rosie uh someone else said they were
8:58 scrolling through Tick Tock and they
8:59 were like oh it's Alec Baldwin boy did
9:01 he get fat uh
9:04 so thank you for the troll
9:08 um some of my Apple shortcuts are
9:09 failing too time out uh Saturday with
9:11 the kids dishes plus learning about AI
9:13 awesome that's great welcome
9:15 um yeah I'm just I don't know how long
9:16 I'm gonna be in here I probably got to
9:18 go shopping or something you know wife
9:19 wife wife wife and and family stuff will
9:23 take over at some point but I'll hang
9:25 out with you all for a bit oh there's a
9:26 hundred people here there's just a dozen
9:27 for a while so if you're new here this
9:29 is the AI learning lab I'm Kyle Shannon
9:32 um I'm just a guy that's teaching myself
9:33 a bit about Ai and sharing what I can
9:36 along the way
9:37 um I've been an entrepreneur for a lot
9:39 of years
9:40 um I was
9:42 um I was there for the for the birth of
9:46 the World Wide Web I was living in New
9:47 York City at the time and
9:50 um kind of discovered the World Wide Web
9:51 around
9:53 94 and published one of the first online
9:57 magazines and one of the first started
9:58 one of the first digital agencies and
10:00 really wrote that that wave up so really
10:02 got to experience
10:04 what it's like to be at the ground floor
10:05 of a transformative technology and it's
10:08 remarkable and I S the parallels between
10:10 1994 and this year are just remarkable
10:13 94.95 and and last year and this year
10:16 are remarkable I think this AI stuff is
10:19 way more profound than the World Wide
10:21 Web I think it's moving dramatically
10:23 faster in fact you know I've said this
10:25 before if you've been here before you've
10:26 heard me say this but it took it took
10:29 about six years to get to 100 million
10:31 users of the World Wide Web it took six
10:33 weeks to get to 100 million users of
10:35 chat GPT
10:37 um that's the URL for chat GPT I don't
10:39 make any money off it I just think
10:41 people should if you're here and not
10:43 doing the dishes you should go
10:45 um
10:46 you should go start playing with that
10:48 right now while I'm talking because
10:50 listening to me is kind of boring
10:53 um
10:53 mom can we get Alec Baldwin we have Alec
10:56 Baldwin yeah
10:59 can we get chubby Alec Baldwin
11:03 uh what's this site called which site
11:06 the site I was on this is Discord
11:10 is that what you're asking about
11:12 and it's blue willow is the image
11:15 generation tool here so basically
11:18 if you go into one of these rookie
11:20 channels
11:22 I think you do it just like mid Journey
11:24 here off to the side like once you once
11:27 you sign up and sign in
11:29 and then
11:31 um in this case I think it's rookies I
11:33 suppose after a certain amount of images
11:35 you you graduate out of rookie mode I
11:38 haven't done that but I think you type
11:39 in slash imagine is that right yeah so
11:42 you type in slash imagine and then
11:43 there's a prompt there and you say
11:46 1970s muscle
11:49 car in an abandoned
11:54 Factory thing
11:58 exciting prompts stay tuned for the
12:00 results
12:02 positive reinforcement is everything
12:04 all right so we will see if that does
12:06 anything interesting
12:09 um
12:10 that 70s muscle car one is is a fun one
12:13 when I first tried to use
12:16 um oh there it is those are not bad
12:19 yeah now it's not really an abandoned
12:21 Factory those look like mighty clean
12:23 factories to me
12:25 but those are definitely 70s muscle cars
12:28 cool
12:30 so that's blue
12:31 whales free
12:33 um so we can reproduce art like Van
12:35 Gogh's without any mental illness
12:37 attached exactly that was talking about
12:40 style drop that new announcement from
12:43 Google that style drop is not out by the
12:45 way which is kind of a drag
12:49 um style drop
12:51 if you weren't here when I just showed
12:53 this I'll show it again Style
12:56 drop Google
13:00 style drop text to image generation
13:03 so this is this is pretty remarkable so
13:05 it's called style drop
13:07 and then
13:09 let me see I will zoom in on this so
13:14 in the little corner is the reference
13:17 image and then
13:18 like the The Prompt here is a logo with
13:21 the letters style drop in a 3D rendering
13:23 style so it pulls the color palette and
13:26 visual style from The Source image and
13:29 then
13:32 just does it
13:34 remarkably well I I mean
13:38 good Lord
13:39 good Lord what have we created what have
13:43 we done people what have we done
13:46 um and I if you were here before I just
13:48 shared this but still this stuff blows
13:50 my mind
13:52 oops I went plus I need to go minus hold
13:55 please
13:57 so the little image to the left is the
14:00 source image and then all those images
14:02 to the right
14:04 are what were created
14:09 just by Magic
14:12 this you know
14:15 like that one like that's that it's such
14:17 a specific illustration style and like
14:20 how it stylized the hat and the little
14:21 robot in the lower left corner
14:24 and then this is a completely different
14:25 style so the Moose is completely
14:27 different the robot's completely
14:29 different completely different still
14:31 completely different still and and
14:33 that's just pulling that style from that
14:35 little
14:36 illustration of a pine tree with a drop
14:39 shadow notice all the things have drop
14:40 Shadows on them the little white outline
14:44 good Lord good Lord what's happening
14:47 good Lord what's happening okay
14:51 um
14:52 let's see I'm selling AI patterns on
14:54 Etsy right now I feel bad but I need
14:56 money don't feel bad City listen
14:57 everyone's gonna do this and
15:00 these tools are really good at creating
15:02 patterns and here's how I look at it
15:05 there's one of the things that people
15:07 that I hear people saying right now is
15:10 you know AI made this AI made that that
15:14 the AI is the tool but saying AI made
15:17 this is kind of the equivalent of saying
15:19 Photoshop made that photo no a
15:23 photographer made that photo and edited
15:25 it in Photoshop and did touch up and
15:27 things like that and it was the
15:29 photographer that shows what image to
15:31 put in there how to edit it whether or
15:33 not to share it how to share it how to
15:35 crop it it's the same thing with AI if
15:39 you figured out how to do really cool
15:41 patterns in mid-journey or unstable
15:43 diffusion
15:45 and you're creating
15:46 non-repeating patterns and
15:49 they're [ __ ] beautiful because you've
15:51 got a good eye
15:53 who cares what tool you used like
15:56 I I
15:58 why should you be penalized
16:01 because you figured out a way to make it
16:03 take you know 15 or 20 minutes to do
16:07 that pattern instead of three days like
16:09 you might have traditionally or or four
16:11 hours or whatever it might be
16:14 um there's no shame in that these are
16:15 just tools they happen to be the most
16:17 powerful tools in the history of
16:19 humanity don't get me wrong they're
16:21 absolutely [ __ ] intense and they're
16:23 gonna absolutely change almost every
16:25 profession but do not feel guilty do not
16:28 feel guilty
16:30 I mean everyone deserves to
16:33 be self-expressed and find ways to make
16:35 money and so anyway I do not fault you
16:39 at all for using these tools in fact I
16:41 encourage it make more
16:44 um I kind of look like her
16:46 that's Rosie O'Donnell thanks
16:51 good god what have I done with my life
16:56 hi no idea who you are are you famous or
16:59 something well apparently I look like
17:01 Rosie or O'Donnell or fat Alec Baldwin
17:03 uh but no I'm not famous uh I'm Kyle
17:07 Shannon I'm the uh the creator of this
17:10 channel the AI learning lab I'm just a
17:11 guy learning about AI the channel here
17:13 is about if you have any questions about
17:15 AI pop them in and I will do my best to
17:19 answer them so let's go Preston Rhodes
17:23 all right remember those days I guess
17:25 that was in reference to the World Wide
17:26 Web days
17:28 all right let's see hey I still don't
17:32 100 understand Discord can you explain
17:34 yeah I think I did but
17:37 discord's a real pain in the ass these
17:41 are servers well or at the very top
17:44 though right there is the official chat
17:46 GPT website if you've not played with
17:47 chat GPT
17:49 shame on you you're at the AI learning
17:51 lab we learn about AI here go there
17:53 start playing do it you're gonna have to
17:56 do it anyway because it's going to be
17:58 rolled into every piece of software you
17:59 use so might as well start now
18:01 hello hello who said that hey chat CeCe
18:03 how are you
18:05 uh um let's see this is the beginning of
18:09 the beginning that's Eric Sanders saying
18:10 that you are correct sir we are so early
18:13 in this stuff it the the parallels to
18:15 the early days of the web are
18:19 um are just mind-shattering because it's
18:22 just right now all of these AI tools as
18:25 remarkable as they are they're kind of
18:26 these sort of broken little toys that
18:29 you know there's there's you know a
18:32 couple of Decades of research and
18:34 science that are sitting underneath them
18:35 but the actual tool like here's a fun
18:37 little fact Chachi PT itself was
18:40 programmed in two weeks two weeks and
18:42 they launched it in two weeks and then
18:44 it went to 100 million users in you know
18:47 in six weeks
18:48 um
18:49 so chat GPT is getting better but it is
18:52 a very raw piece of software all of this
18:54 stuff is very raw and so
18:57 if you ever see something like it just
18:59 doesn't look good or the hands are
19:01 messed up or the
19:02 it you know chat GPT lies or all that
19:05 sort of stuff just know that we are at
19:07 the beginning of the beginning
19:08 you won't be able to make a channel wait
19:11 you won't be able to make a channel in
19:14 the company's server not quite sure what
19:17 that means I assume that meant for
19:18 Discord
19:19 so difficult for illustrator Mama to
19:21 survive style copying listen here's the
19:24 deal I you know
19:26 I was chief creativeofficer of
19:28 agency.com I the the the entire
19:32 Suite of creative professions are going
19:34 to be disrupted dramatically the only
19:37 thing I can say is this is that creative
19:40 people that learn Ai and learn how to
19:42 wield these tools well are going to be
19:44 better at it that none than
19:46 non-creatives for one thing but also
19:49 um
19:50 in terms of job longevity
19:54 um I think using these tools is going to
19:56 be a requirement
19:57 um and and not everyone's going to do it
19:59 and so I would say lean in heavy I did a
20:03 one of my Tick Tock videos in the
20:04 channel was about you know if if you're
20:06 if you're afraid of this stuff
20:10 I would say find the thing in AI that
20:12 scares you the most and and lean into
20:15 that learn the most about that because
20:17 it's probably the part of AI that most
20:20 directly threatens you know your current
20:22 Identity or your current livelihood
20:25 um but start learning that because
20:26 you'll be better at that than other
20:28 people who don't have your expertise so
20:30 I I completely understand the
20:32 existential angst I you know I kind of
20:35 live in this Duality right now of I'm as
20:37 excited as I've ever been and I'm as
20:39 scared as I've ever been and I'm not
20:41 scared like dystopian future scared I
20:44 mean I think that stuff's a possibility
20:46 but I'm more scared that even if it's
20:48 all positive change it's a lot of change
20:51 and it's going to be a lot of change
20:52 very fast that I don't think people are
20:54 prepared for
20:55 is there a process is there an AI that
20:58 process a video and make shorter clips
21:01 from it yeah if you go to video
21:02 v-i-d-y-o
21:07 i v i d y o dot a i
21:11 and let's see start for free I think I
21:15 have an account here
21:17 welcome Kyle Shannon apparently I do yes
21:21 so like here's a uh a speech I gave it
21:24 creative mornings it's like a 20 minute
21:26 25 minute speech and it
21:30 um chunked out
21:32 all these little vertical Clips out of a
21:34 longer form piece they're not perfect
21:37 but they're pretty good
21:39 I put musk up there just because he's
21:41 literally launching rockets and you know
21:43 making electric cars possible and
21:45 I like I kept having this feeling like
21:47 is this one of those is is this is this
21:50 a thing and so I taught myself this
21:51 thing called stable diffusion where you
21:53 had to actually get birth right so so
21:56 yeah so you just upload a long form
21:58 video it it pieces them out
22:01 um you kind of Select you want Square
22:04 format vertical format or horizontal
22:06 format
22:07 it chunks them up you can then change
22:09 the in and out points and things like
22:11 that so the answer is yes there's going
22:12 to be a ton of those things as well and
22:14 there's probably other ones most
22:16 terrifying for illustrators is that you
22:18 can copy their style that look that took
22:20 them a lifetime to create I it exactly I
22:23 listen
22:24 I um my my
22:27 um co-host of the AI Salon her name is
22:31 Leah Fest and she's been a professional
22:32 photographer for 25 years and you know
22:36 she came into this thinking you know
22:38 listen I I went through the digital
22:40 Revolution from film to digital and and
22:42 you know that was disruptive but you
22:43 know I could manage that
22:45 um and I think she came into this AI
22:47 stuff with the same spirit and then it
22:50 pretty quickly became apparent to her
22:52 that it's going to be relatively trivial
22:55 for someone to take one of her
22:56 photographs and mimic the style and be
22:59 able to create anything
23:01 and so so I think she's in the middle of
23:03 facing that right now and she's you know
23:05 she's looking to how do I reinvent you
23:08 know my definition of what I do
23:11 um and and uh if you've been doing stuff
23:13 for 25 years like her it's it's it's
23:16 tough It's
23:21 a dog agrees
23:29 [Music]
23:31 don't you know I'm on tick tock
23:33 laughs
23:37 damn dogs don't understand what Tick
23:39 Tock live is you can't bark when Daddy's
23:41 on live you can't do it
23:43 all right
23:46 let's see exactly the idea is coming
23:49 from humans still yeah so that was in
23:52 reference to the fact that you know
23:53 these AI tools are not making the art
23:56 it's people with intention making the
23:58 art
23:59 here's what Remains What remains a
24:02 requirement and what remains a critical
24:04 skill with with all these fancy new
24:06 tools is critical thinking and strategy
24:09 and taste and curation and refinement
24:13 and you know all the stuff that is when
24:17 you see really good work it's coming
24:19 from a point of view which means it's
24:21 coming from a human these machines don't
24:24 have points of view they happen to be
24:26 incredibly capable assistance but but
24:29 Microsoft calls their their tools
24:31 co-pilots I think that's the the best
24:33 branding I've heard
24:35 and the best definition of what what
24:37 these things are you're still the pilot
24:39 AI tools or co-pilots they're just
24:42 really [ __ ] powerful ones oops I just
24:44 lost my place in questions uh if you're
24:47 if you're new here looks like we got
24:48 some new people my name is Kyle Shannon
24:50 this is the AI learning lab we are
24:52 learning about AI in a lab or a home
24:55 office
24:57 um go to if you have not played with
25:00 chat GPT that top URL is the chat GPT
25:03 URL
25:04 um
25:05 prompts.chat that thing below it is just
25:08 a thing that will teach you about
25:09 prompting and give you a bunch of ideas
25:12 futurepedia is a list of AI tools that
25:15 will make your head melt and and ruin
25:18 your weekend so don't go there do not go
25:20 to futurepedia.io
25:23 um and if you have questions about AI in
25:25 any form pop it in the thing I will do
25:27 my best to answer it happy Saturday it's
25:29 a little rainy here in Denver it's
25:32 the last two three weeks here in Denver
25:34 have been very weird it's normally very
25:36 sunny all the time but it has been
25:39 um rainy like East Coast rainy for like
25:41 three weeks now so
25:43 can I use this to build a better website
25:45 for my business yeah you can it will
25:47 code for you
25:49 um there are if you use the plugins okay
25:52 if you're if you're brand new to this
25:53 stuff so if you go to
25:56 that cat gbt and so that list of things
25:59 down the left is my history
26:02 and hang on like my
26:06 it should be it is right so that's super
26:09 cool
26:10 so it made this table it gave me a
26:14 description of what the paper is and
26:15 then it classified it this one's
26:17 interesting this one's worth a look I
26:19 didn't tell it to I didn't give it worth
26:21 a look as a category this one's exciting
26:24 this one's interesting exciting worth a
26:26 look interesting breakthrough okay so
26:28 this bottom one deep learning for
26:31 anomaly detection a survey this survey
26:33 paper provides comprehensive overview of
26:36 the use of deep learning for anomaly
26:38 detection a critical task in many
26:40 applications such as fraud detection and
26:42 network security
26:45 so so plug-ins turn Chat GPT from
26:49 remarkable to oh my God like the amount
26:53 of time like a I would never do that I
26:55 would never go to a white paper site and
26:58 sift through the things and read the
26:59 papers enough to understand what they
27:01 are
27:02 that I could even categorize them like I
27:03 wouldn't do it in the first place I have
27:05 ADD I just don't give enough of a [ __ ]
27:07 about that stuff to do it
27:09 but if I can just say hey chat EBT go
27:12 find me some [ __ ] and analyze it and it
27:15 does like that's the kind of
27:18 the ability for chat GPT to now use
27:20 tools is is going to
27:23 um
27:23 it's just gonna change the game so this
27:26 goes back to your question about can I
27:27 help you build a better website
27:30 I'm pretty sure I was looking through
27:32 some of the plugins last night I'm
27:33 pretty sure there are
27:35 um
27:36 there are plugins in there that will
27:38 analyze your current website and give
27:40 you recommendations for ways to improve
27:42 it
27:43 all right these tools will be required
27:45 for most jobs yes yeah they're going to
27:47 get rolled into like the the Microsoft
27:50 co-pilot is going to get rolled into
27:51 Office 365 Dynamics Azure
27:56 um Windows 11 it's going to be built
27:58 into the operating system Monday There's
28:00 a Big Apple worldwide developers
28:02 conference where they're announcing
28:03 their goggles
28:05 uh they're also going to be announcing
28:06 some AI stuff I think like imagine Siri
28:09 as a really powerful assistant
28:12 um they've already announced that in the
28:14 next IOS operating system you're going
28:16 to be able to clone your voice and
28:18 you're going to be able to send voice
28:20 messages without actually having to
28:21 record them but it will be in your voice
28:24 um and and they're they'll probably do
28:25 something with the face they'll probably
28:27 do something with agents autonomous
28:29 agents so
28:31 raise my hand to look into how to use AI
28:33 for my team and my boss wanted to
28:36 consult experts [ __ ] my life I'm right
28:38 there with you so here's what you can
28:40 tell your boss from a guy
28:43 sitting in his home office in Denver on
28:45 tick tock there are no [ __ ] experts
28:50 no one is an expert at this [ __ ] right
28:51 now
28:52 chat so okay that's not true there are
28:56 data scientists and machine learning
28:58 experts that have been working in the AI
29:00 field for the past bunch of decades what
29:02 they're not experts on right now is
29:05 generative AI chat GPT forward basically
29:07 so think from November 30th 2022 until
29:10 now that like that's the window that
29:12 we're talking
29:14 as Eric said earlier we are at the
29:16 beginning of the beginning so I'm happy
29:18 to talk to your boss I'll tell him you
29:20 can tell him I'm an expert I'm happy to
29:22 talk to him what I'll tell him is uh let
29:25 uh cavruno I don't sorry I don't know
29:27 what your name is
29:28 um
29:29 let that person
29:32 um
29:33 start exploring for you because
29:36 she I think he or she she will become
29:38 the expert so I'm I'm I I share your
29:41 frustration bosses need to get their
29:43 heads out of their asses and start
29:44 embracing this stuff and if they've got
29:46 employees that are coming to them going
29:47 hey boss I'd like to figure this [ __ ]
29:49 out for you they should be going oh my
29:51 God hallelujah Bring It On
29:54 ugh that's why I started this channel to
29:57 educate bosses like your boss to get
30:00 their head out of their ass I suggested
30:02 that writers ask chat gbt to write a new
30:04 season of Criminal Intent yeah I know
30:06 it's really good at writing by the way
30:08 if you are a writer
30:10 there was an interesting Andre carpathy
30:12 who's the chief data scientist for
30:15 openai for chat GPT
30:17 um he gave a speech to the Microsoft
30:19 developers conference last week and he
30:21 was talking about prompt engineering how
30:23 you do it if you're a writer
30:26 um give chat EPT examples of your work
30:29 so say based on this and then paste in a
30:32 block of your work that's representative
30:34 of the style that you're trying to write
30:35 and then give it a prompt to write
30:38 something it's really good at mimicking
30:41 examples that you give it style wise so
30:45 so
30:46 um yes but it should also I I think
30:48 we're going to see a lot of
30:53 Innovation and and exciting new creative
30:57 projects as a result of cat GPT so
31:00 hopefully that'll migrate its way
31:03 um into into the TV uh I I then I
31:06 remember that there was a writer's
31:07 strike yeah well so the writers strike
31:09 you know I've written seven screenplays
31:11 I I know the world well I support the
31:13 writer's strike but they they
31:17 their their request to not have Studios
31:20 replace them with robots I think is
31:23 incredibly short-sighted and and a bit
31:26 of a Miss it's a bit of an oversight
31:29 here you've got the most incredible
31:31 writing tool in the history of mankind
31:33 at your fingertips
31:35 and the union is saying [ __ ] that we're
31:39 never going to use that tool it's only
31:41 about human writers when
31:44 when chat GPT is rolled into Microsoft
31:47 Word and it's rolled into every
31:49 screenwriting software out there and
31:51 it's rolled into all of the Google Docs
31:53 and it's it's in all the note-taking
31:55 applications it's going to be ubiquitous
31:58 so it's not like like right now chat GPT
32:01 is this thing off to the side it's it's
32:03 like this little hermetically sealed
32:05 thing and we're like we're gonna have
32:07 chat GPT detectors and we're going to
32:09 detect if you're using those stupid
32:11 robot tools well when that stupid robot
32:14 tool is integrated into every other tool
32:16 that you use where's the distinction
32:18 between it was written by Ai and it was
32:21 a human assisted by AI I don't know
32:24 where that boundary line is and quite
32:26 frankly I don't think it's relevant it's
32:28 still up to the writer to refine the
32:30 results and and have the intent in the
32:32 first place and do the editing in the
32:34 first place right so so I think the The
32:36 Writer's Guild is has got to get their
32:38 [ __ ] together sooner than later because
32:40 I I think they're gonna
32:42 they they could strike themselves into a
32:45 into a real position of weakness
32:48 um is my fear for the writing Community
32:50 now listen I I do not
32:53 pretend to think that these AI tools
32:55 aren't going to [ __ ] up a lot of careers
32:57 because they are it is
32:59 but at the same time we're not stopping
33:01 it you know it's kind of like this it's
33:03 I I have this joke that I've been saying
33:05 that you know trying to fight against
33:07 this AI thing is kind of like the State
33:10 of Florida declaring that they're no
33:11 longer accepting hurricanes
33:13 okay great good policy I agree you
33:17 should not have hurricanes and then you
33:19 know the next hurricane is going to come
33:20 along and blow down some houses
33:22 that's that's what we're in the face of
33:24 with this AI stuff it's coming whether
33:26 we want it or not so you might as well
33:27 lean in that's the purpose of this
33:29 channel just go there and start playing
33:33 and by the way if if you're one of those
33:34 cynical troll types that's like he's
33:37 making a fortune off of pointing people
33:39 there he's getting a commission every
33:40 time he uses chat GPT it's not true I'm
33:43 just a dude pointing you to a site that
33:45 doesn't suck
33:46 what tool were we just using for the
33:48 icons and small sketches so it's not a
33:51 tool that's out yet bald style drop
33:55 and this is a new
33:58 um
33:59 a new paper from Google
34:01 so so don't forget Google invented the
34:05 Transformer the core technology that
34:07 sits underneath chat GPT and all of
34:09 these remarkable image generation tools
34:11 that was all invented by Google there's
34:13 a paper from 2017 called attention is
34:16 all you need where they introduced this
34:18 idea this thing called a Transformer
34:20 that's the foundation for everything so
34:22 this is a just a one of many pieces of
34:25 technology that Google has announced
34:27 this was announced yesterday or the day
34:29 before
34:30 um but check out what this thing does so
34:33 I I just every time I see some of this
34:37 AI stuff it every freaking time I'm like
34:40 I can't get my head around how good this
34:42 stuff is so this little image here
34:45 that's my little finger my fat finger
34:47 that little image there of a kid's
34:50 drawing is the source image all of these
34:53 other drawings were generated from AI to
34:56 match that style so it matches the color
34:58 palette it matches the wrinkled paper it
35:01 figures out
35:02 like how a kid would draw a moose like
35:05 which where would they use the green
35:06 where would they use the yellow
35:09 um
35:09 and then so that's a kid's drawing
35:11 here's Starry Starry Night
35:14 here's starry starry night with the blue
35:15 removed here's another Van Gogh
35:18 self-portrait there's another Van Gogh
35:20 self-portrait here's some cool 90s
35:23 illustration style that it just [ __ ]
35:25 nails
35:26 here's some other kind of 90s like this
35:28 is that bad clip art you see everywhere
35:31 but like all of those illustrations are
35:34 being generated whole cloth out of
35:36 nothing based on the style of the the
35:39 single Source image
35:41 it's it's just remarkable
35:45 so that's that it's not available yet
35:47 though so
35:48 tough you know they just tease us they
35:51 tease us they tease us with these tools
35:53 what are we supposed to do I want to go
35:55 play
35:57 wait speaking of which I think all right
36:00 we're twirling we're twirling
36:04 come on
36:05 come on oh by the way if you want to
36:07 subscribe to the channel I have a goal
36:10 of 13 subscribers I'm up to five
36:12 yeah
36:13 I mean I have like 19 000 followers for
36:15 my Tick Tock Channel but subscriptions
36:17 to the live
36:18 um I've got five of them I don't know
36:20 what happens if you subscribe to my
36:22 wives and if it costs you anything
36:24 please don't subscribe I don't think I'm
36:26 worth uh paying for but uh it this this
36:30 Google Music LM thing does not appear to
36:32 be working I heard it was working I
36:34 figured we could play together but ain't
36:36 nothing happening let me try to reload
36:38 and see if it just
36:40 crash the page
36:43 could not fetch waitlist status
36:46 well that's shitty all right I guess
36:49 we're doing nothing there then
36:51 all right
36:52 there you have it all right
36:56 all right any copyright issues with AI
36:59 that I should watch out for yes
37:02 so
37:05 um images in particular but but
37:07 potentially even with writing so here's
37:10 the thing the next 20 years is going to
37:12 be a never-ending series of lawsuits and
37:15 counter lawsuits about copyright and
37:18 trademark and intellectual property
37:20 infringement the intellectual property
37:22 lawyers are ecstatic right now
37:26 foreign
37:28 so there's lots of gray areas
37:31 with the image generation if you're
37:34 using it commercially what I would
37:36 suggest you know especially if you're
37:39 working with a big Fortune 500 companies
37:41 and things like that use Adobe tools use
37:43 Adobe Firefly and use generative fill
37:46 within Adobe Photoshop because what
37:49 Adobe wisely did was they trained their
37:52 image model on Adobe stock images that
37:55 they owned outright and on
37:57 um public domain images
38:00 so so there's nothing in the training
38:02 data of adobe stock that that they don't
38:05 have clearance for so anything that you
38:07 generate there's there is a essentially
38:09 well I can't say zero I'm not a lawyer
38:13 um and I don't even play one on the
38:14 internet
38:15 um there's a very low likelihood that
38:17 that you're gonna be
38:19 um in any sort of hot water
38:22 um if you're using something like mid
38:24 Journey or stable diffusion which which
38:26 we're trained on a bunch of copywritten
38:29 images
38:30 um my friend Kenny Friedman who's a
38:32 creative director who's really good at
38:34 um uh this AI image Generation stuff
38:38 he's made a personal commitment that
38:41 when he does the prompting he never uses
38:43 a specific artist's name if it's someone
38:46 you know like one of the classics like
38:48 Van Gogh he might do that but any living
38:50 artist he's not he's not using their
38:52 name
38:53 so
38:55 um so that's a way that you can avoid it
38:57 but if you create something that is
38:58 close enough to someone's likeness I
39:00 think there there might be lawsuits
39:02 brought
39:03 um if you're on the other side and
39:04 saying can I actually copyright some of
39:06 this stuff that remains to be seen
39:08 what's not clear to me is okay if I go
39:11 to Mid journey and I create an image you
39:13 know from Whole cloth out of mid-journey
39:16 what the what the copyright office is
39:18 saying right now is that there's no
39:20 human contribution there I disagree with
39:22 that because I put in the prompt and I'm
39:25 the one that generated a hundred images
39:26 of which I chose one of them that I
39:28 think is the good one right I think just
39:30 like using a camera the camera is
39:32 actually doing all the work but you're
39:33 pointing it you're figuring out what
39:34 lens to use you're figuring out exposure
39:36 settings so I feel like it's a similar
39:39 kind of thing
39:41 um but whether or not you can copyright
39:43 AI generated stuff is up for grabs and
39:46 then with with photoshop's new
39:47 generative fill
39:49 where's the point at which it was AI
39:52 generated if I start with a photograph
39:54 and I expand it
39:56 and 90 of that image is AI generated and
40:00 10 of it is my original photograph but
40:02 it all started from my original
40:03 photograph is that copyrightable is that
40:06 AI generated or is that human generated
40:08 I don't know and I don't think it's
40:10 particularly relevant from a creative
40:12 standpoint
40:14 um the lawyers are going to spend the
40:16 next two decades figuring that [ __ ] out
40:18 so so yes there are issues but I
40:20 wouldn't worry too much about it because
40:21 it's way above any of our pay grades
40:24 um it's going to have to be figured out
40:25 at a much higher level can I please
40:27 explain stochastic gradient descent I
40:29 can't I heard Andre carpathy talk about
40:32 it well wait
40:33 of course I can one of the ways I've
40:37 describe uh stable diffusion or not
40:40 stable effusion chat GPT is knowledge on
40:43 demand and it can absolutely
40:47 help us figure this the [ __ ] out
40:53 um explain
40:55 stochastic
40:59 gradient
41:02 descent
41:05 as if I have a degree in acting
41:10 [Laughter]
41:12 I.E
41:15 I ain't got
41:17 no
41:19 book smarts
41:21 and yes you can talk to these things
41:23 like an idiot and it still understands
41:25 what you're asking for oh wait here let
41:27 me flip over here Absolutely I'll try to
41:29 explain it in a way that's easier to
41:31 digest it didn't troll me which is good
41:34 um let's imagine you're the director of
41:37 a play and your mission is to oh it's
41:39 given me acting metaphors sorry about
41:41 that that's kind of stupid all right
41:42 your mission is to deliver the best
41:44 performance possible however you have
41:46 thousands of individual elements to
41:48 manage actors performance is lighting
41:49 sound props blah blah
41:51 you could try to make all the changes at
41:53 once say give direction to all the
41:55 actors adjust all the lights change all
41:57 the props in one Fell Swoop but that
41:59 would be chaotic and almost impossible
42:01 so you opt for a different approach you
42:04 focus on improving one element at a time
42:06 based on immediate feedback okay I see
42:09 what this is so in a similar way
42:11 stochastic gradient descent SG SGD is a
42:15 method that is used to find the best
42:17 possible settings for a model
42:20 um the model settings are adjusted bit
42:21 by bit so so yeah basically what they're
42:23 doing is they're taking
42:25 they're taking the results I I assume
42:28 they're taking the results of a prompt
42:29 they're analyzing that in some way they
42:31 fix some piece of it they reanalyze it
42:33 fix some piece of it so it's something
42:36 like that but anyway I you know go go
42:38 play with chat GPT to dig deeper on that
42:41 and uh it will it will do it it's a
42:44 great way to use the tool I'll often use
42:46 chat Epp I'll be on the phone with
42:48 someone smarter than me which is nearly
42:50 every column on
42:51 um hang on I need some coffee
42:55 and they will say something that is
42:58 outside of my knowledge domain
43:00 and uh I'll just quickly pop into chat
43:02 GPT and say in one sentence describe
43:05 what this is so I can get my head around
43:06 it so it's a good way to use it
43:08 stuck in a local minimum go back and use
43:10 a rocket I don't know what that's in
43:12 reference to excited that apple is going
43:13 to do as Microsoft will be rolling out
43:15 copilot yeah or excited what apple is
43:18 going to be rolling out I I
43:21 Apple has been suspiciously quiet on the
43:23 AI front
43:24 which tells me that they're coming up
43:26 with something cool all of these
43:28 companies have have been doing deep AI
43:32 research for decades right they they all
43:34 have some Skunk Works back there chatgpt
43:37 has forced them out of the Shadows right
43:40 the fact that we haven't heard of peep
43:42 from Apple
43:43 should mean that Monday so Monday
43:46 at 10 A.M Pacific time
43:49 is the big keynote speech for the
43:51 worldwide developers conference
43:54 um they're going to be announcing their
43:55 AR VR goggles and they're also going to
43:58 be
44:00 um what's it called
44:02 they're going to be announcing some AI
44:03 stuff so it's going to be pretty cool
44:04 all right you got something in your nose
44:07 probably a booger sorry about that might
44:09 just be hair I'm old I got I you know
44:12 when you're old you go to the uh the
44:13 barber and they're cutting your hair I
44:15 haven't been in a while and and they're
44:16 like would you like me to trim the hair
44:18 off your ears and you're like oh god oh
44:20 what have I done so anyway sorry about
44:22 that
44:23 all right I asked it to take 1100 word
44:25 down to six save me so much time yeah
44:27 it's good at rewriting stuff it's good
44:28 at analyzing stuff can I use this to
44:30 build a website from a business we
44:32 talked about that these tools will be
44:34 required for most jobs oh I think I'm
44:36 way behind on my questions sorry about
44:37 that
44:38 sorry I didn't get the name of the AI
44:40 for short video clips video v i d y o
44:43 dot a i
44:46 hopefully you got that
44:49 uh have you tried Claude I have tried
44:52 Claude let me show you some
44:55 um
44:57 I thought I had it up here nope
45:00 but I can type three letters
45:03 so here's Poe so the cool thing about
45:06 Poe is you can go use clotted anthropic
45:09 site but Poe gives you access to lots of
45:12 different models Sage is the is the is
45:14 pose model quora
45:16 um gpt4 and then there's three different
45:18 Claude models and then there's chat GPT
45:20 and then you can make your own Bots like
45:22 I made I made a sling blade bot I made a
45:25 Yoda bot I made a Shakespeare bot so you
45:27 can make your own Bots those are fun so
45:29 if you go to poe.com I am Shakespeare or
45:32 I am Yoda you can play with the bot that
45:33 I made but anyway
45:36 um Claude instant is incredibly fast
45:38 it's like GPT 3.5 but it's incredibly
45:41 fast quad plus is is more like gpt4 it's
45:45 a direct competitor to gpt4 and
45:48 anthropic the company behind Claude was
45:50 founded by a bunch of EX open AI people
45:53 the people that made gpt3 quit and
45:55 started anthropic so they know what
45:57 they're doing and then there's also this
46:00 Claude instant 100K so if you pay for
46:04 the Post subscription which I think is
46:05 20 bucks a month
46:08 um
46:09 and I've paid for it because I wanted
46:10 access to this Claude 100K what what
46:14 that means is it's got a context window
46:16 of a hundred thousand tokens which
46:18 basically means this you can put in up
46:21 to 75 000 words into your prompt so you
46:25 could literally paste in a novel as the
46:27 prompt and say you know write me an
46:30 epilogue write me a whatever rewrite the
46:33 last chapter whatever whatever you want
46:35 to do um and it will analyze it and read
46:37 it so yeah I have tried Claude and it's
46:39 it's quite good I am still partial to
46:42 um
46:43 gpt4 for whatever reason it might be you
46:46 know
46:47 short established Nostalgia but oops but
46:50 I think it's I think it's uh I think
46:52 it's better but the hundred thousand
46:54 token limit for Claude is absolutely
46:56 amazing
46:57 how would one become a prompt engineer
46:59 start prompting start playing go there
47:03 start solving problems with chat GPT
47:06 um and then go to the the link below it
47:09 prompts.chat and that will teach you a
47:11 lot about prompt engineering I also
47:13 prefer to call it prompt crafting rather
47:15 than prompt engineering it it is like
47:18 engineering
47:20 in that you're kind of thinking like a
47:22 programmer but you're using natural
47:23 language
47:25 um but it's a very creative process and
47:27 it's not a single you don't create a
47:30 create a single massive prompt and then
47:32 get a perfect answer it tends to be lots
47:34 of little prompts and lots of back and
47:35 forth with the large language models
47:41 can AI fix the feed I don't know is my
47:44 feed screwed up
47:45 is it lagging for anyone else am I still
47:48 lagging hit the tap the heart button if
47:51 you can hear me and I'm still here
47:54 like tap the screen or whatever let me
47:57 see if I'm actually
47:58 functional all right looks like I'm
48:00 still functional all right
48:02 um here's the reason for the lag Xfinity
48:05 sucks
48:08 thanks Comcast I pay you how much for
48:12 what
48:15 doesn't appear to be videoed nor sound
48:18 like when everyone was saying that no
48:21 one is going to make money selling on
48:22 the internet exactly
48:24 happy to help you present yourself to
48:26 your boss as an AI expert oh cool so it
48:29 looks like you guys are supporting one
48:30 another someone said they were trying to
48:32 go to their boss uh Eric sander by the
48:34 way the guy that's in here that's
48:35 offering to be an AI expert is an
48:38 awesome dude and is really wicked smart
48:41 so take him up on that I asked chat GPT
48:43 to write a comprehensive report on
48:45 change management for our company boss
48:47 not amused well bus needs to get his
48:50 head out of his ass boss should embrace
48:52 what you wrote and go fix the [ __ ]
48:54 company in my in my humble opinion
48:57 if I have a if I have an employee that
48:59 takes the initiative to go figure out
49:02 how to do some change management the
49:04 Assumption there is that you're trying
49:05 to make change and it ain't working so
49:06 you got to do something different
49:08 bus needs to check his ego at the door
49:10 take off his stupid [ __ ] hat and
49:12 listen to his employees all right I have
49:15 a ton of news wait I have a ton of
49:17 newsletters that I do via Emma I could
49:20 send it the links to read them online to
49:22 see yes you can
49:24 and the pl the new plug is well so
49:26 you've got
49:28 chat GPT now can search the internet but
49:30 there's also plugins that will go
49:32 specifically read PDFs or web links or
49:34 things like that you can give it a whole
49:36 list of web links and they'll go read
49:37 all the contents of them and then
49:39 you can analyze across them it's pretty
49:41 pretty freaking slick if you ask me
49:45 all right how do I get AI on my desktop
49:47 so I can use some of the features of
49:49 chat GPT you mean without an internet
49:52 connection
49:54 um in that case you would you would
49:56 probably need a a decently Kick-Ass
49:59 machine
50:00 um but you can so so the the developer
50:03 Community is taking these open source
50:05 large language models and making them
50:06 very very efficient there are now a
50:09 bunch of different projects out there
50:11 that you can install a large language
50:12 model on your like laptop there's even
50:15 one that'll run on iOS on a phone
50:17 without any internet connection
50:20 um they're almost as good as chat gbt
50:22 3.5 they're they're not gpt4 they're not
50:25 connected to the internet but there's
50:27 using the the open source large language
50:29 models you can do it but if you have an
50:31 internet connection just go there just
50:33 go to chat.openai.com that's the
50:35 official chat GPT you don't need to
50:37 install it it's just a website there is
50:40 an official IOS app
50:42 for chat EBT I would strongly recommend
50:45 you do not just go to the App Store
50:47 because you'll be confused by the 862
50:51 fake chat EBT apps what I would do if
50:54 you want the official chapter
50:57 and on the home page of openai.com is a
51:00 link to the IOS app go use that link
51:03 that's the official chat GPT app you
51:07 don't have to pay for it there's a bunch
51:08 of ones out there you have to pay for
51:09 you should not be paying for the chat
51:11 GPT app
51:13 period
51:14 so one of the reasons I keep putting
51:16 this URL up here like a like a dum-dum
51:19 I'm a dum-dum uh I I ain't the smartest
51:23 kind of filler but it does lack my
51:25 coffee
51:29 all right
51:32 but the report was excellent okay oh the
51:35 report of of the uh change management
51:37 I'm sure it was
51:39 buses need to get they need to get a
51:42 clue
51:44 I mean seriously if you've got employees
51:47 that are taking initiative to solve some
51:49 [ __ ] for you
51:50 [ __ ] step aside
51:53 I mean I'm sorry I'm a CEO of a company
51:55 and
51:57 more often than not what is brought to
51:59 you is hey boss this thing sucks
52:03 okay have any ideas oh no I haven't
52:06 thought about that okay there's the door
52:09 but if someone comes to me and says hey
52:11 boss this thing sucks and I went to chat
52:13 gbt and I figured out how to fix it what
52:16 do you think listen
52:19 good Lord people like it ain't that hard
52:23 is it apparently it is
52:25 I don't know what's going on in the
52:27 professional world anymore I really
52:28 don't nobody answers phones anymore
52:30 nobody responds to emails people just
52:32 don't show up to [ __ ] meetings it's
52:33 driving me crazy anyone else
52:35 experiencing that tap the screen if you
52:37 feel like there is just a corporate
52:40 malaise cast amongst the world
52:45 getting lots of hearts
52:47 I don't know what's going on people have
52:49 just checked the [ __ ] out of everything
52:50 done with it
52:52 I'm bitter
52:56 bitter
52:58 well see now I'm sorry I got lost in my
53:00 little in my in my comments that I'm
53:02 answering so if I have not answered your
53:04 question that's because I'm lost
53:09 um all right where am I
53:11 the problem is chat GPT is not optimized
53:14 to be correct it's optimized to sound
53:16 correct yeah that's correct it so chat
53:18 gbt will what they call hallucinate or
53:22 lie or as I call it mansplaining as a
53:25 service
53:26 um it will confidently answer you even
53:28 if it's wrong one of the things you can
53:31 do is ask it to analyze its work and see
53:33 if it if it thinks it's right that
53:35 sometimes helps
53:37 found your channel a few days ago so
53:39 interesting thanks for the content thank
53:40 you so much Erin I appreciate it
53:43 um can you please explain to casket
53:44 gradient descent oh man I'm way off of
53:46 my questions I did that a while ago what
53:48 was that website I don't know
53:50 uh um
53:52 I like how you are streaming live what
53:55 tool do you use on your phone I'm so I'm
53:57 just using
53:58 um
54:00 when you go live on Tick Tock you can
54:02 turn on dual cameras so that little
54:04 window right there is my back camera and
54:07 I just have my laptop sitting there so
54:08 that's just facing my laptop and then
54:10 the other one's facing this piece of
54:12 Glorious genetics laughs
54:16 as someone said I think you have
54:17 something in your nose it could just be
54:19 my nose
54:24 oh and then how you switch you just tap
54:26 the little wind I tap it and it swaps
54:28 between the two it's pretty slick
54:31 all right I subscribe uh I thought it
54:34 would notify me when you're on but it
54:36 doesn't I paid 5.99 oh I'm sorry don't
54:38 pay 5.90 yeah I I'm sorry that you pay
54:41 5.99 you feel free to unsubscribe to
54:43 this I I go live uh I will help you for
54:46 free sorry about that AI music is not
54:49 new it existed a long time before now I
54:51 agree with that oh listen a lot of these
54:53 tools existed before this here here's a
54:56 fun little factoid in the documentation
54:59 for gpt3 about a year and change before
55:03 like a year and a third before Chachi BT
55:05 came out in the documentation
55:08 were instructions for how to create chat
55:11 gbt
55:12 and Sam Altman assumed that someone in
55:15 the development Community would make
55:17 chat gbt and for a year and a third
55:20 nobody made it and then Microsoft was
55:23 about to announce their billion dollar
55:24 investment in open Ai and Sam Altman
55:26 said to his Tech Team uh Hey guys since
55:29 no one's made this yet could you just
55:30 whip up some chat GPT and they built it
55:33 in two weeks
55:35 two weeks
55:39 and in six weeks after that they had 100
55:42 million users
55:44 so the thing that changed everything was
55:46 written in two weeks so a lot of the
55:48 music tools have been around a lot of
55:50 the image tools have been around a lot
55:52 of this stuff's been around
55:53 the big difference why we're at an
55:55 inflection point right now is they put
55:57 an interface in front of it that was
56:00 simple enough for real people to
56:02 experience the magic of AI right you
56:06 input something that seems
56:09 simple like you put into a Google Search
56:11 and what you get back is [ __ ]
56:13 remarkable
56:15 explain quantum physics to me in an m m
56:17 rap bang out it comes
56:21 so that's what's different what's
56:22 different is none
56:25 scientists and researchers have access
56:28 to these tools now technically they had
56:30 access to them before but they were just
56:32 a little too geeky Chachi PTC chat gbt
56:36 opened the dam
56:39 um AI as a housekeeper robot that will
56:42 be coming in fact open AI just
56:45 um hang on let me see if I can find it
56:49 um open AI robot
56:54 yes
56:57 open AI beats Elon Musk taking a job so
57:01 this is a humanoid robot that like walks
57:03 around a house and can open doors and do
57:05 [ __ ] here let's let's watch it together
57:07 I haven't seen this
57:08 [Music]
57:11 oops
57:20 foreign
57:22 so it's opening a box
57:25 putting [ __ ] in the Box
57:29 it's got like little Segway roller feet
57:32 check that out
57:34 now it's squatting
57:38 so it's opening a door it's opening a
57:40 window
57:43 yeah
57:44 so yes
57:46 the humanoid robot Eve with its
57:48 human-like appearance appearance in
57:50 robotic hand stands out from its
57:52 contemporaries like Boston Dynamics its
57:55 capabilities include opening doors
57:57 windows manipulating objects setting new
57:59 standards in the realm of humanoid
58:01 robots
58:02 um you know Elon Musk said at the last
58:05 Tesla shareholder meeting that he feels
58:07 like the the the total addressable
58:09 market for humanoid robots is
58:14 um one for every person on Earth maybe
58:17 two so
58:19 so that's Elon for you everyone a robot
58:23 for you a robot a robot for everybody
58:26 um but it's coming
58:28 all right it's great for Fields it's
58:31 great for General Fields but more Niche
58:33 stuff
58:34 Adobe generated fill isn't too awesome
58:38 yeah so Adobe generative Phil again
58:42 here's the here's the thing
58:44 don't look at any of these tools right
58:47 now as complete tools like it's in I I
58:50 guess it's in beta
58:52 but like the Adobe the Adobe training
58:55 model is an incredibly limited model
58:58 they didn't train it on a lot of stuff
59:00 they could have because they wanted it
59:01 to be commercially viable
59:04 here's what I say about a lot of this AI
59:06 stuff what you see today is as bad as
59:10 it's ever going to be
59:12 so the fact that adobe generative fill
59:14 kind of sucks at certain things
59:16 won't be for long you know a year from
59:18 now it will be unrecognizable from what
59:21 it's doing today
59:23 all right everyone is an expert if
59:25 anyone who tells you they're an expert
59:27 at this [ __ ] right now is lying
59:30 uh what a great use case explaining
59:32 stochastic stochastic gradient uh for
59:34 for any people such as the folks here
59:36 yeah exactly let's go need more than a
59:39 surface statement all right so go to
59:41 church EBT and make something Apple
59:44 iPhone has a chat GPT app
59:47 okay be careful
59:50 the Apple App Store and I assume Android
59:53 has many many apps that are called chat
59:57 GPT that have a logo that's almost
59:59 identical to open AIS because those are
1:00:02 charlatans trying to rip you off if you
1:00:04 want actual chat GPT go to that URL from
1:00:07 a web browser
1:00:08 or
1:00:10 go to the base openai.com website and
1:00:13 click on their link for the official
1:00:15 chat GPT app you should not have to pay
1:00:17 for it
1:00:18 what about IBM um IBM just announced
1:00:23 I think it's called Watson AI they're
1:00:24 trying they're they're dusting off the
1:00:27 Watson that kind of failed with Watson
1:00:29 Health
1:00:30 um and they're trying to become relevant
1:00:32 again
1:00:33 um you know they've got big technology
1:00:35 here but open AI has has
1:00:38 shaken everything up it's shaking
1:00:40 everything up open AI is 64 models
1:00:42 available from the API it's cool yeah
1:00:44 yeah they've got lots of different
1:00:45 models some are optimized for code some
1:00:47 are optimized for
1:00:49 predictive stuff and analytical stuff
1:00:52 and all sorts of things so if you're a
1:00:54 developer there's lots to play with
1:00:55 Google bard is still experimental but
1:00:57 will outperform all other AI models
1:01:05 I don't know that I agree with that
1:01:08 because
1:01:10 I look at Apple and I look at open AI
1:01:12 right now
1:01:14 and those are two companies that
1:01:16 understand
1:01:19 um liberal arts they understand Humanity
1:01:22 the humanities Steve Jobs was a master
1:01:25 at combining technology and liberal arts
1:01:27 that's why apple is what it is
1:01:31 Google is all about engineering it's all
1:01:34 it's all
1:01:36 left hemisphere there's no right
1:01:38 hemisphere so I think companies that
1:01:41 have a balance of left and right that
1:01:43 that can go deep on the tech but can go
1:01:45 deep on the on the humanity side
1:01:48 I think they're going to outperform
1:01:49 Google Now
1:01:51 Google will probably be faster it'll
1:01:53 probably be better at technical things
1:01:55 it'll probably be better at Accurate
1:01:57 things but if you go use Google bard
1:02:01 um
1:02:01 I did this yesterday on my live I did
1:02:04 give me an m m rap that explains quantum
1:02:07 mechanics
1:02:08 Google bards was laughably bad gpt4 was
1:02:13 shockingly good like the gap between the
1:02:15 quality of those two models right now is
1:02:18 staggering but
1:02:19 if I were
1:02:22 having it generate code maybe Bard has
1:02:25 has a leg up and also Google's got tons
1:02:28 of money and they got lots of scientists
1:02:30 so they're going to build some tools
1:02:31 like the the
1:02:33 um uh style copy tool that they just
1:02:35 announced yesterday is remarkable and
1:02:38 and it's you know it's doing incredible
1:02:40 beautiful visual stuff so Google's gonna
1:02:42 have some good tools I don't know that
1:02:44 Google bard is going to catch up to open
1:02:47 AI
1:02:48 uh but it but it they'll probably
1:02:50 they've got smart people there can you
1:02:52 imagine working with a prompt engineer
1:02:54 they know everything can you imagine
1:02:56 working with a prompt engineer they know
1:02:57 everything well
1:02:59 um they don't know everything but what
1:03:02 what prompt Engineers or prompt Crafters
1:03:04 do know is how to ask these systems for
1:03:08 things and get predictable results I've
1:03:11 been lagging sporadically that sucks all
1:03:13 right I'm sometimes lagging for some
1:03:15 seconds but not so bad as yesterday
1:03:17 sorry about that
1:03:19 oh Xfinity if anyone works for Xfinity
1:03:21 [ __ ] you
1:03:24 um if anybody knows anyone that works at
1:03:25 Xfinity that can get my Internet fixed
1:03:27 that would be awesome
1:03:30 awesome because I'd like to have some
1:03:32 stable internet if I had stable internet
1:03:33 get these lives without the lags lags
1:03:36 suck
1:03:42 want to build an app sure what idea you
1:03:44 got
1:03:47 um I am in a just say yes mode for most
1:03:50 projects right now because I want to
1:03:51 figure out as much as I can in a short
1:03:54 amount of time as possible if you're new
1:03:56 here there's some new folks here okay my
1:03:57 name is Kyle Shannon
1:03:59 this is the AI learning lab
1:04:01 I am learning about Ai and sharing what
1:04:03 I'm learning in real time to try to get
1:04:05 as many people up to speed as possible
1:04:07 because I think it's actually important
1:04:09 for Humanity I think we also happen to
1:04:12 be living in a very amazing time at the
1:04:15 at the at the dawn of a transformative
1:04:19 moment I think this is bigger than the
1:04:21 Industrial Revolution
1:04:23 um I think it's way bigger than the
1:04:24 personal computer I think it's way
1:04:26 bigger than the world wide web on top of
1:04:28 the internet
1:04:29 I think it's bigger than the iPhone
1:04:31 um and so that's why I'm doing this so
1:04:32 that URL right there chat.openai.com
1:04:35 that's chatgpt's official whips website
1:04:38 so if you have not played with chat gbt
1:04:40 while you listen to me ramble on a
1:04:42 Saturday afternoon uh here in the states
1:04:44 anyway
1:04:45 um go play there and then if you don't
1:04:47 know how to do prompting and you're like
1:04:49 what are these people talking about
1:04:50 prompt engineering go to that thing
1:04:52 prompts.chat and it will explain to you
1:04:54 a lot about prompting and give you a
1:04:56 bunch of ideas so go do that
1:04:59 analyze how what a great use case okay
1:05:03 let's see
1:05:06 I keep getting my my tick tock
1:05:11 Things Are
1:05:13 sorry words not coming
1:05:17 we often serve the bosses instead of the
1:05:19 customers I know I know
1:05:22 it's true Snapchat AI helped explain
1:05:25 parts of my Gene test to me it was
1:05:27 awesome wow that's cool
1:05:28 that's really cool I listen I think
1:05:32 all of the complicated documents you
1:05:35 know you know the complicated documents
1:05:37 that that we get like you go buy a car
1:05:39 and they hand you this stack of papers
1:05:40 and they're like sign here and you're
1:05:42 like I should probably read this [ __ ]
1:05:43 but then you go try to read it and your
1:05:45 eyes start bleeding and you're like I
1:05:46 can't possibly read this we're going to
1:05:49 very quickly be in a in a place where
1:05:52 we'll just be able to snap a picture of
1:05:54 that document it will instantly analyze
1:05:56 it for us and say uh according to you
1:06:00 know section three they're ripping you
1:06:01 off uh so I think that will come there
1:06:04 someone's thanking me for my efforts
1:06:06 thank you for acknowledging my efforts
1:06:08 when's the first AI human marriage
1:06:11 coming
1:06:12 you know I know you're saying it
1:06:15 jokingly and and we've got enough Sci-Fi
1:06:17 movies
1:06:19 um but but if you haven't seen Lars in
1:06:21 the real girl that movie
1:06:24 um
1:06:25 which is about a guy that has a
1:06:27 relationship with a sex doll and you
1:06:30 initially think this is going to be the
1:06:32 creepiest worst movie in the history of
1:06:35 all movies and then you watch it and
1:06:36 it's one of the most
1:06:38 sweet and endearing and moving
1:06:42 um stories you've ever encountered it's
1:06:44 it's just one of those it's one of those
1:06:46 things where my expectation going in and
1:06:49 what it was were they couldn't be more
1:06:50 different
1:06:51 I think that virtual companionship is
1:06:54 actually going to be a good thing
1:06:57 um are there going to be creepy pieces
1:06:59 of it yeah
1:07:01 um
1:07:02 but there's a lot of lonely people out
1:07:04 there and and
1:07:06 um and I think it's actually going to be
1:07:07 a thing so
1:07:09 will there be a marriage I you know
1:07:12 listen
1:07:13 people love attention so someone is
1:07:15 gonna marry a robot just so they can
1:07:17 make the Evening News
1:07:19 um and someone will make the Evening
1:07:20 News because they married a robot but I
1:07:23 actually think that that you know
1:07:25 virtual companionship is actually going
1:07:27 to be a big thing an important thing
1:07:30 um not always good but but I you know it
1:07:33 there's a lot of lonely people out there
1:07:35 and that sucks
1:07:36 how long before Bard and gpt4 are
1:07:38 brought to Google home hubs and Alexis
1:07:40 very very soon it's going to be in
1:07:42 everything
1:07:44 thoughts on free llm models um
1:07:48 so there's a paper if you want to read
1:07:51 it or have chat GPT read it for you
1:07:53 which is how I read all my papers now
1:07:56 um wait hang on sorry looking at my nose
1:07:58 which apparently there's something in
1:08:01 um let's see we have no moat if you go
1:08:05 search on Google for we have no moat the
1:08:09 first link
1:08:10 is this paper called we have no moat and
1:08:13 what this is is this is a this is
1:08:15 apparently a leak I don't know it's
1:08:17 possible it's a leak it's possible it's
1:08:18 just them
1:08:20 put putting information out into the
1:08:22 world but this is an engineer from
1:08:23 Google basically saying that
1:08:27 the free large language models that
1:08:29 Engineers are playing with they're
1:08:31 getting similar results with a hundred
1:08:33 dollars of training
1:08:36 um as what took Google like months and
1:08:40 10 million dollars of training
1:08:43 um so the the free large language models
1:08:46 are getting incredibly powerful there's
1:08:48 lots of them out there I'm sure a lot of
1:08:50 them are pretty weak and weird and not
1:08:53 great there's others of them that are
1:08:55 looks like they're starting to get good
1:08:57 but I think what what we're going to see
1:08:58 is
1:09:00 probably within a year
1:09:03 um
1:09:04 you're gonna have very small files that
1:09:06 you can just install on your phone or
1:09:07 your iPad or your computer that are kind
1:09:10 of gpt4 quality large language models
1:09:13 that do not require an internet
1:09:14 connection
1:09:16 um I I think I think we'll start seeing
1:09:18 these large language models be burned
1:09:20 into chips where there's they're just
1:09:23 installed like every device is going to
1:09:25 have
1:09:27 incredibly deep capabilities because of
1:09:29 these things so my thoughts on on them
1:09:32 are they're absolutely remarkable now
1:09:34 what's the downside of these free models
1:09:37 they are open source and they are as
1:09:40 powerful as GPT 3.5 right now is about
1:09:43 it
1:09:44 but they have no safety guard rails so
1:09:46 GPT 3.5 and gpt4 when you ask it to do
1:09:51 something bad it goes yeah we're not
1:09:53 going to do that go ahead and go [ __ ]
1:09:55 yourself
1:09:56 um these free large language models if
1:09:58 you're a bad actor and want to do bad
1:10:00 [ __ ]
1:10:01 um you're free to do it there there are
1:10:03 no safety guard rails on them so you can
1:10:05 train it to do whatever the hell you
1:10:06 want so that's there's a dark side to
1:10:09 the open source argument
1:10:11 the Adobe image extension feature is
1:10:13 brilliant I haven't oh wait what I don't
1:10:16 know what that is Adobe image extension
1:10:18 oh that's the that's the generative fill
1:10:21 oh it's brilliant it's absolutely
1:10:22 brilliant which of these do I think is
1:10:24 best I don't know which of which
1:10:28 are you an AI I'm gpt5 nice to meet you
1:10:31 don't I seem lifelike
1:10:37 no
1:10:41 um oh I just lost my place again God
1:10:43 damn it tick tock your scrolling sucks
1:10:49 what time is it 226 I gotta go soon I
1:10:52 gotta go shopping or something I got to
1:10:54 do something for my family
1:10:56 you're like what are you doing those
1:10:58 lives all the time I'm like I like these
1:11:00 people better than I like you you your
1:11:02 family
1:11:06 trying to find myself again what was
1:11:09 that website replicate Style
1:11:13 oh boy I know I'm way back to way before
1:11:16 where I'm supposed to be trying to catch
1:11:19 up hold please
1:11:20 404 eastern time sometimes lagging oh
1:11:23 we've been through all these I'm just
1:11:24 zooming zooming around the questions
1:11:26 hold please
1:11:28 summarize in five bullet points yeah you
1:11:30 can have things summarizing five bullet
1:11:33 points if you use a voice module and
1:11:35 write
1:11:36 a slash then you can get out speech here
1:11:38 in Tick Tock wait
1:11:40 if you use a voice module and write a
1:11:43 slash
1:11:45 then you can get out speech here in Tick
1:11:47 Tock I don't quite know what that means
1:11:49 but it sounds cool I'm an old guy using
1:11:52 Tick Tock what the hell do you want from
1:11:53 me people
1:11:54 apparently I'm also not supposed to
1:11:56 drink on camera so I'm gonna go drink
1:11:57 this off camera
1:12:01 iced coffee tasty
1:12:03 how can I use AI to improve the sales in
1:12:06 my organization our sales is going down
1:12:08 recently welcome to the [ __ ] club
1:12:10 sales right now is nearly impossible
1:12:14 um
1:12:16 it's really good for research
1:12:18 so
1:12:20 um I was trying to I was trying to
1:12:22 identify some sales prospects and and I
1:12:25 was like go find this particular role
1:12:27 like give me the top 10 companies that
1:12:30 do this particular function now go find
1:12:32 me the VP of whatever it was like the VP
1:12:34 of marketing
1:12:35 um you know for those companies and it
1:12:38 did that and I'm like give me the
1:12:39 headquarters and give me the link to the
1:12:41 website so you so it's good at research
1:12:44 um using something like zapier you can
1:12:47 automate
1:12:48 um inbound and outbound emails as emails
1:12:51 come in
1:12:52 you can have it automatically turn
1:12:54 around and write very personalized
1:12:57 responses to sales prospects
1:13:00 um
1:13:01 lots and lots of ways I mean what I
1:13:04 would encourage you to do is
1:13:08 think of Chachi PT like a consultant
1:13:10 like a cons like like an assistant
1:13:13 for your sales organization and and tell
1:13:15 it your situation explain as much as you
1:13:18 can about what you do and what's working
1:13:20 and what's not working and what you see
1:13:22 has changed and have it suggest
1:13:25 strategies for
1:13:28 um turning things around and then once
1:13:30 you find strategies it recommends that
1:13:32 you like have it expand on those so so
1:13:35 think of the the interaction more like
1:13:38 you're interacting with a consultant you
1:13:40 hired um and it it might give you some
1:13:42 ideas that can help get you out of it
1:13:44 but it's tough
1:13:46 um what does the robot identify as
1:13:48 that's a good one
1:13:49 FYI I don't have you don't have such a
1:13:53 module I think you guys are talking to
1:13:54 each other that's cool Andrew Bibi
1:13:57 uh oh oh Andy B sorry and I thought it
1:13:59 was Andrew BB because it's Andy B uh
1:14:02 anyway
1:14:03 all right
1:14:04 will we see in video oh God damn you
1:14:07 Tick Tock will we see Nvidia Kudo cores
1:14:11 and llms changing the world
1:14:16 language models are
1:14:19 changing the world and yeah right now
1:14:23 Nvidia like I mean they've been around
1:14:25 for a while and they got
1:14:27 you know they got really big with gaming
1:14:29 and then they got really big with
1:14:30 blockchain and nfts
1:14:32 and now there they just hit a trillion
1:14:34 dollar market cap with the with these
1:14:35 large language models
1:14:37 so
1:14:39 um
1:14:40 I think they're I think they're gonna be
1:14:42 around for a while like Nvidia is a
1:14:43 pretty impressive company right now I
1:14:45 don't know who competes with them wasn't
1:14:46 it Microsoft that just said they're
1:14:48 gonna make their own
1:14:50 chips because they don't want to be
1:14:52 beholden to Nvidia someone someone just
1:14:55 said that I think it might have been
1:14:56 Microsoft
1:14:57 three stock stock recommendations for AI
1:14:59 yeah I'm I have a degree in acting I'm
1:15:02 not good at that you know I think the
1:15:04 the companies that I think right now are
1:15:06 killing it in the space not all of them
1:15:08 are investable open AI obviously
1:15:12 um Microsoft Nvidia
1:15:15 um I think
1:15:17 I would be surprised if Apple didn't
1:15:20 have something pretty impressive to
1:15:22 announce on Monday so
1:15:24 um I think Apple's gonna be a big player
1:15:26 in this space
1:15:29 I think anthropic is a really
1:15:31 interesting company but I don't I don't
1:15:33 know what you can invest in or I would
1:15:35 certainly not I I am glamor gpt10 that's
1:15:38 great
1:15:39 uh The Tick Tock slash thing means we
1:15:42 can talk to you with the voice you don't
1:15:45 have to read oh that's cool
1:15:48 so wait so try that so so wait if you
1:15:50 put a slash in front of one of these
1:15:52 what do I do click on it tap on it
1:15:54 because I saw one earlier
1:15:57 hold
1:15:58 this is a test oh I don't have the
1:16:00 module installed so I need to install
1:16:02 the module damn it
1:16:04 I don't know what I'm doing here
1:16:07 all right let's see stock Rex
1:16:16 hello brother someone said hello brother
1:16:19 who said hello brother
1:16:21 nurbal hello brother or sister
1:16:24 how will this speed up medical research
1:16:26 dramatically if you've heard of alpha
1:16:29 fold
1:16:30 Alpha fold is a Google
1:16:33 deepmind project that
1:16:36 predicted the intricate folding shapes
1:16:39 of what 400 000 proteins in like four
1:16:41 days or three days some ridiculous
1:16:43 amount of time like that and
1:16:45 historically that would have taken like
1:16:46 two months of traditional compute time
1:16:48 to do one and they did four hundred
1:16:50 thousand in a few days
1:16:52 um I think I think the scientific
1:16:54 advancements in the next five years are
1:16:56 going to be just face melting you know
1:17:00 they just there was a guy
1:17:02 I think in England or Italy I think it
1:17:04 was Italy
1:17:05 um who's been paralyzed for
1:17:08 12 years due to a motorcycle accident
1:17:10 they just used a large language model
1:17:12 they built a Bluetooth Bridge around the
1:17:15 damaged part of his spine put electrodes
1:17:17 in his spine
1:17:18 imaged his brain when he was thinking
1:17:20 about moving his legs
1:17:23 you know using large language models to
1:17:25 interpret the data hooked up the
1:17:27 Bluetooth and said all right think about
1:17:29 walking now and he's now walking
1:17:32 like they're actively curing paralysis
1:17:35 as we speak using large language models
1:17:39 so I think it's going to speed up
1:17:41 medical research
1:17:43 in ways that we can't even fathom
1:17:49 um if you teach chat CPT how to do a
1:17:51 task will that knowledge cross over to
1:17:53 all users not really so so
1:17:56 it doesn't really have memory they're
1:17:58 taking what people are doing with chat
1:18:00 GPT and they're fine-tuning it but it's
1:18:02 not like the thing that you figured out
1:18:04 it will figure it out it will figure out
1:18:06 what you can do now
1:18:08 um with chat apt is you can copy and
1:18:10 share prompts so let me show you
1:18:16 so over here's my chat history
1:18:20 oh I gotta
1:18:22 I've got a glare on my screen because of
1:18:24 the light
1:18:26 um over here in my chat history
1:18:28 they've now added a share button so if I
1:18:32 click this share I can I can share this
1:18:35 prompt with other people so I can copy
1:18:37 the link
1:18:38 down here on the bottom and share that
1:18:40 with other people so if you if you come
1:18:42 up with a thread with a with a prompt
1:18:44 and a series of things that's really
1:18:46 good you can copy that and share it with
1:18:48 people
1:18:49 um if you use something like Lang chain
1:18:53 which stitches together a number of
1:18:54 tools but one of the tools that's in in
1:18:58 Lang chain is a what's called a vector
1:19:00 database so that's basically memory so
1:19:03 if you put together something like Lang
1:19:05 chain then the things that you create if
1:19:08 you're writing those results into memory
1:19:10 you can then search on them in the
1:19:12 future but it's not straightforward
1:19:14 right now I would say give it a year
1:19:16 when Microsoft really rolls out large
1:19:19 language models integrated into their
1:19:21 entire Suite it's this is all going to
1:19:23 become much easier so right now it's
1:19:25 just janky and shitty but it will get
1:19:28 less janky and less shitty
1:19:30 how do I think government
1:19:33 oops hang on
1:19:40 we'll regulate AI hopefully smartly
1:19:44 um but they don't [ __ ] understand how
1:19:46 Wi-Fi works so I don't quite know how
1:19:48 that's going to happen
1:19:49 I think they need to have smart people
1:19:52 on on advisory panels
1:19:54 um Sam Altman has been decently
1:19:56 aggressive about cooperating with
1:19:59 government and saying you need to
1:20:00 regulate us and here's how I think you
1:20:02 should do it
1:20:03 hopefully they listen to to smart people
1:20:08 um I think Sam is one of them
1:20:16 I think that we will see governments
1:20:19 um
1:20:21 [ __ ] it up
1:20:24 before they get it right
1:20:26 um
1:20:27 because at some point something's going
1:20:29 to happen something will be headline
1:20:30 making that AI killed someone or hey I
1:20:33 did this or a I did that or AI tanked
1:20:35 the banking sector or whatever it is and
1:20:38 then the government will panic and try
1:20:40 to shut [ __ ] down but you you're not
1:20:43 going to be able to shut down things
1:20:44 that are now open sourced if this were
1:20:46 all sitting in the hands of Google and
1:20:48 open AI like it was for a while
1:20:51 that would be a different story where
1:20:53 you could just kind of shut down those
1:20:54 companies but now that it's kind of out
1:20:56 and open source and in developers hands
1:20:58 and things like that I don't think you
1:21:00 can put the genie back in the bottle
1:21:01 so the only plausible way you could
1:21:03 actually stop it
1:21:05 is to shut down the internet and they're
1:21:08 not going to do that because that's the
1:21:10 entire economy infrastructure these days
1:21:14 um so I think they're gonna have to deal
1:21:15 with it so hopefully fingers crossed
1:21:19 um they don't make
1:21:20 stupid moves to try to control it
1:21:22 because the problem if they try to
1:21:24 control it too much is the stymie
1:21:25 Innovation and the rest of the world
1:21:27 ain't gonna stop
1:21:29 um
1:21:30 but at the same time I think these
1:21:32 things need to be taken seriously cyber
1:21:34 security is is a big important there's
1:21:36 big important implications for that
1:21:40 um and uh yeah so
1:21:44 I don't know enough about government
1:21:47 um to know anything other than I should
1:21:48 be cynical about it and I am I'm a gen
1:21:50 xer and I'm cynical and I'm bitter and
1:21:52 that's how I like like it oh there's
1:21:54 more people here than were here before
1:21:55 okay hi my name is Kyle Shannon this is
1:21:58 the AI learning lab I talk about all
1:22:00 things AI here feel free to ask
1:22:02 questions in the comments if you're new
1:22:04 to AI if you have not played with chat
1:22:05 GPT cool but end that today right now go
1:22:10 to that address
1:22:11 while you listen to me ramble and whine
1:22:14 and drink coffee
1:22:17 chat.openai.com go there
1:22:19 and that's the official chat CPT website
1:22:22 and just start playing start asking it
1:22:23 to do things ask it to write you a song
1:22:26 for your significant other or your pet
1:22:29 um ask it to
1:22:30 this is a fun one ask it to give you a
1:22:33 meal plan for the week that includes I
1:22:35 don't know five obscure ingredients and
1:22:37 then it'll do that and then tell it to
1:22:40 create a shopping list for you and then
1:22:42 have it organize that shopping list by
1:22:44 category of the grocery store
1:22:47 and just watch what it does and then go
1:22:49 you could do that I didn't know I could
1:22:52 do that
1:22:54 and then when you pick your drop off the
1:22:56 floor ask it to write you a Python
1:22:58 program and watch it do that and then
1:23:00 you'd be like what it could do that too
1:23:03 so yeah go play there do I have a degree
1:23:06 in artificial intelligence I have a
1:23:07 degree in acting
1:23:11 um
1:23:12 I I'm CEO of a video technology company
1:23:15 I've been an entrepreneur for many many
1:23:17 decades now
1:23:20 um
1:23:21 I I got in on the ground floor of the
1:23:24 early days of the World Wide Web I
1:23:25 started one of the first digital
1:23:26 agencies in 1994 a company called
1:23:28 agency.com and
1:23:32 um
1:23:33 I recognized in 1994 that the web was
1:23:36 going to change everything and that no
1:23:38 one knew it and I had a really cool
1:23:41 experience where I published an online
1:23:43 magazine
1:23:45 from my apartment in Brooklyn I uploaded
1:23:47 the files to a computer in California on
1:23:49 a dial-up modem
1:23:50 and three weeks later there was a
1:23:52 full-page article about my magazine in a
1:23:54 Parisian newspaper Halfway Around the
1:23:57 World
1:23:57 and I recognized in that moment that the
1:24:00 world had changed but no one knew it yet
1:24:02 and so I was lucky enough to kind of be
1:24:05 there at the beginning and kind of know
1:24:06 be able to see what was coming I had a
1:24:09 similar Epiphany in December of 2022
1:24:12 last December
1:24:14 where I was playing with chat GPT and
1:24:16 asked it to write me a Python program
1:24:18 and in 90 minutes I wrote a functional
1:24:20 fully functional python application
1:24:23 that took user inputs about a certain
1:24:26 topic and generated five different
1:24:27 pieces of social media content I I wrote
1:24:30 that application with calls to the open
1:24:33 AI API in 90 minutes I've never written
1:24:36 a line of python in my life
1:24:39 and that was my big breakthrough moment
1:24:41 about holy [ __ ] the world just changed
1:24:42 like this is another one of those this
1:24:44 is like on the level of the World Wide
1:24:46 Web the personal computer the Industrial
1:24:49 Revolution the printing press fire
1:24:51 written language like there aren't many
1:24:54 of these and we're lucky if not lucky
1:24:57 enough to live in an era where there's
1:25:00 been a number of these in a very short
1:25:02 amount of time this is a big one
1:25:05 um so that's why I'm geeking out on this
1:25:08 stuff but no I don't have I do not have
1:25:10 a machine learning background I know how
1:25:12 this stuff works but I don't know how
1:25:14 the [ __ ] they do it
1:25:16 Alpha fault is like seti it really is
1:25:18 okay what's the best way to learn AI
1:25:21 engineering if you want to learn AI like
1:25:23 machine learning engineering like the
1:25:25 engineering side then I would say go to
1:25:28 go to school or have chat gbt teach it
1:25:31 to you if you mean how do I learn AI
1:25:33 prompt engineering and how to use
1:25:35 generative AI tools um just get Hands-On
1:25:38 watch YouTube videos watch tick tocks
1:25:41 like mine watch Rachel Woods she's got a
1:25:44 tick tock Channel watch Mackenzie Bose
1:25:47 if you're decently geeky there's a guy
1:25:50 on YouTube named David Shapiro brilliant
1:25:53 machine learning scientist and he's
1:25:55 creating an AGI kind of platform really
1:25:59 wicked smart guy so depending on just
1:26:02 figure out your level of geekiness and
1:26:03 then go find people that are just
1:26:05 talking about this [ __ ] and start
1:26:07 playing just get in the get in that
1:26:09 weird uncomfortable Zone where you're
1:26:10 like I don't know what I'm talking about
1:26:12 I don't know what I'm doing here nobody
1:26:13 knows what they're doing here I feel
1:26:15 clueless every [ __ ] day this and I've
1:26:17 been like obsessing about it for six
1:26:19 months now look like nine months now
1:26:21 I've just been like
1:26:23 and I feel clueless every day so just
1:26:25 feel clueless jump in feel the feelings
1:26:27 feel the feels as they say as the kids
1:26:30 say these days
1:26:33 [Music]
1:26:33 um
1:26:36 the Air Force is using on their drones
1:26:39 yeah well so there was an interesting
1:26:41 article that the mainstream media jumped
1:26:45 on that uh an Air Force drone turned
1:26:48 around and attacked its operator and
1:26:50 then it turned out well it was a
1:26:51 simulation and then it turned out well
1:26:52 it was actually just a thought
1:26:53 experiment where they said that could
1:26:56 happen and it was mischaracterized so
1:26:58 whether or not that happened or not I
1:27:00 don't know the media jumped on it though
1:27:02 because The Narrative of the robots are
1:27:04 going to kill us is a strong one that
1:27:06 sells you know gets eyeballs and [ __ ]
1:27:09 like that I think it's pretty
1:27:10 irresponsible and it makes me cranky but
1:27:13 what are you gonna do
1:27:14 all right let me answer a few more
1:27:16 questions and then I gotta go live my
1:27:18 Saturday life oh maybe I'll come back
1:27:19 later tonight
1:27:22 um that's not the original chat GPT site
1:27:24 yes it is chat.openai.com that's open
1:27:27 AIS that's chatgpt
1:27:29 chat.openai.com is the original Chachi
1:27:32 EBT site if you're on another one
1:27:35 um you're not on the original that is
1:27:37 the site
1:27:40 um
1:27:42 are you still out here blowing Minds I'm
1:27:44 still out here the mines might be
1:27:46 blowing it's like why is he still
1:27:48 talking
1:27:50 thank you again for the information
1:27:51 thank you for your instruction you're
1:27:53 welcome I'm acting like a guy
1:27:55 so you're acting like a guy who knows
1:27:57 about AI I am that's that's why I have
1:28:00 an acting degree it's mind-blowing I
1:28:02 know that I act like I know this much
1:28:04 mansplaining as a service at your
1:28:06 service
1:28:07 all right what was the prompt to to
1:28:10 write a Python program let's go write
1:28:12 one link okay we're gonna do a new chat
1:28:16 so and I'll do this on GPT 3.5
1:28:19 um because that's the free thing so even
1:28:21 if you're not paying for it so I'll say
1:28:24 um
1:28:26 I am completely new to programming
1:28:33 and I heard python
1:28:39 is the way to go I want to write
1:28:44 an application
1:28:48 that takes in some user input
1:28:56 input and sends a prompt
1:29:01 to the open AI
1:29:06 um G let's see the open AI GPT API
1:29:12 um
1:29:13 is this something you can help me make I
1:29:20 don't need
1:29:23 to learn
1:29:25 the programming
1:29:27 I just want help making
1:29:32 the application now
1:29:34 location if it seems like that's a
1:29:37 little wordy and and like more like I'm
1:29:40 talking to
1:29:43 a you know a senior programmer or like
1:29:45 I'm a product guy talking to an engineer
1:29:49 like one of the things that I've learned
1:29:50 is is the more context that you give GPT
1:29:53 the better it gets
1:29:55 so if I hit return now certainly
1:29:58 I can help you get started
1:30:02 um here are the basic things set up
1:30:04 python install python on your computer
1:30:06 so it gives you the link of where to get
1:30:08 it install the required packages you'll
1:30:11 need to install the open AI package to
1:30:13 interact with it pip install so it's
1:30:16 telling me some command line stuff here
1:30:17 so I don't understand all this but look
1:30:19 there's there's some
1:30:21 um let's see right in up I write the
1:30:24 application create a new python file and
1:30:28 open it in a text editor here's the
1:30:30 simple example of how the code would
1:30:31 look there's your openai key there's
1:30:34 your user input enter a prompt
1:30:37 it it already pulled the DaVinci 003
1:30:40 engine for open AI it gave it a max
1:30:44 number of tokens of 100 and then it's
1:30:47 going to print the responses to screen
1:30:49 make sure you replace your API key so
1:30:52 it's telling me what to do but what I
1:30:53 can say here is
1:30:54 um
1:30:56 I don't understand how to do any of the
1:31:01 local installing
1:31:05 is it possible to run
1:31:09 this
1:31:10 python application on a website question
1:31:16 mark
1:31:17 yes it's possible to run it on a website
1:31:20 and what did it tell me here
1:31:22 use a web framework like fret flask or
1:31:25 Django to create a an application using
1:31:27 flask so so they're still saying pip
1:31:29 install so they're still having me
1:31:31 install stuff from the command line
1:31:32 which I don't want to do so I'm going to
1:31:34 say
1:31:35 um I don't want to use command
1:31:39 line
1:31:41 is there a web site that runs
1:31:46 python apps without me installing
1:31:52 anything
1:31:54 boom yes replit.com
1:31:58 and then it gives me the instructions of
1:32:00 how to go to replit.com you can
1:32:02 customize the project you can do it
1:32:04 right so
1:32:06 um
1:32:07 and then let me see
1:32:09 um yeah that's it so so you get the
1:32:11 basic idea so in in a very short amount
1:32:14 of time you can have it do something
1:32:16 I'll show you another cool thing
1:32:18 um we're gonna go over to
1:32:23 GitHub
1:32:27 so GitHub is is this code repository let
1:32:30 me sign in here
1:32:34 oh I've got two Factor authentication
1:32:37 script you can take the readme file from
1:32:40 GitHub and
1:32:43 um go to chat GPT paste the readme file
1:32:47 in there and say I'm on OSX give me a
1:32:50 step-by-step
1:32:51 um instructions for how to get this
1:32:54 GitHub repository up and running and
1:32:57 then you can even say just give me one
1:32:59 instruction at a time and don't move on
1:33:01 to the next instruction until I say okay
1:33:03 and it will walk you step by step
1:33:05 through the process so how can we make
1:33:07 money with AI don't worry so much the
1:33:10 the generative AI tools are just tools
1:33:14 how you make money is the way we've
1:33:17 always made money it's it like the
1:33:18 answer to that question is no different
1:33:19 than how do I make money with Photoshop
1:33:21 or how do I make money with
1:33:24 um a lawn mower
1:33:26 how you make money with any technology
1:33:28 with any tool is you say what's a
1:33:31 problem someone's got that I can solve
1:33:33 that if I solve it well enough they'll
1:33:35 pay me for it
1:33:37 so I would say take things where you
1:33:39 know a pain point you know where there's
1:33:41 a problem that you can solve and if you
1:33:43 can figure out how to use AI to solve
1:33:45 that then that's how you make money with
1:33:46 it I'm a fairly good actor thank you I I
1:33:49 shoot for Fairly good you know I you
1:33:53 know I don't want to be award-winning
1:33:56 fairly good is where I'm that's that's
1:33:57 what I'm looking for
1:33:59 so that's exciting all right what
1:34:02 business would I start today
1:34:04 I would start
1:34:16 short term
1:34:18 I think I think I would probably start
1:34:22 either a software development company or
1:34:25 an agency
1:34:28 that is all in on AI
1:34:32 um
1:34:33 and what I would shoot for is being 10
1:34:36 to 30 times
1:34:38 faster cheaper or better than the
1:34:41 competition
1:34:42 because I think for the next two years
1:34:46 they're going to be three to four person
1:34:47 companies
1:34:49 that are executing at the level of a 30
1:34:52 or 40 person company and and those three
1:34:54 to four person companies are going to be
1:34:57 able to knock off
1:34:58 big competitors
1:35:01 um so I probably Smart Start I would
1:35:03 find
1:35:04 two or three other very smart
1:35:06 you know AI brilliant people
1:35:10 not AI brilliant
1:35:13 um brilliant people that know how to use
1:35:15 AI tools
1:35:18 and I would probably find one that's
1:35:19 good with technical stuff one that's
1:35:21 good with visual stuff one that's good
1:35:23 with business and word stuff marketing
1:35:27 business and then I probably find an
1:35:28 operations person so I'm going to make
1:35:30 sure we don't go out of business I think
1:35:32 with those three people
1:35:35 you could probably be doing a level of
1:35:39 work that that you could do with a with
1:35:41 a 30 or 40 person company so I don't
1:35:44 know what I'd do it in right now but it
1:35:46 would definitely be leaning in heavy
1:35:47 into AI
1:35:48 looking forward to your live streams
1:35:50 like this one thank you appreciate that
1:35:51 is the upgrade to GPT worth it for
1:35:54 writing books yeah gpt4 is just much
1:35:56 better you can use if if you just want
1:35:59 to use
1:36:00 um
1:36:01 gpt4
1:36:04 um and you don't want to pay for it just
1:36:06 go to bing.com
1:36:09 bing.com
1:36:11 and then when you go to Bing
1:36:14 it'll it'll say
1:36:17 um ask me anything but but up here right
1:36:19 next to the Bing logo is this chat
1:36:22 and that is chat GPT and there's three
1:36:25 different conversation Styles more
1:36:27 creative more balanced and more precise
1:36:29 so depending on the kind of writing
1:36:31 you're doing if you're doing technical
1:36:32 writing you might want it on more
1:36:34 balanced or more precise if you're
1:36:35 writing screenplays or novels you might
1:36:37 want to know more creative so this is
1:36:40 gpt4
1:36:42 I think it's a little more hobbled than
1:36:45 open ai's version of it but
1:36:48 um
1:36:51 but it's free
1:36:53 um so for me the upgrade to GP to chat
1:36:55 GPT is worth it so the the upgrade gives
1:36:59 you chat GPT Plus
1:37:01 and then on the upper part of the screen
1:37:03 here you've got access to GPT 3.5 which
1:37:06 is fast and you've got access to gpt4
1:37:09 and then you've got three models
1:37:10 currently within it you've got the
1:37:12 default which is what it used to be
1:37:13 you've got browse with Bing so this is
1:37:16 Internet connected chat GPT and then
1:37:18 you've got plugins there's all sorts of
1:37:20 plugins for for writers and there's
1:37:22 going to be more and more and more so I
1:37:24 think the ability to do all this stuff
1:37:26 right now is worth the upgrade now
1:37:29 um Bing is already connected to the
1:37:32 internet you know chat GPT is using Bing
1:37:34 and they have also announced that
1:37:35 they're going to support the openai
1:37:37 plug-in architecture so bing.com is
1:37:39 going to be able to support all these
1:37:40 plugins so the distance between bing.com
1:37:43 and current chat GPT is small but
1:37:46 there's a thing coming out for gpt4
1:37:48 called code interpreter that is
1:37:50 absolutely mind-blowing so the reason I
1:37:53 like paying for this is these guys are
1:37:55 the ones leading the charge right now so
1:37:57 I want as Early Access as I can to the
1:38:00 tools that they put out so I think it's
1:38:02 I think it's worth it
1:38:04 but yeah
1:38:05 um gpt4 is significantly better at basic
1:38:09 writing skills than GPT 3.5 is just
1:38:13 better it's just good it's really good
1:38:16 for things like yeah just it's good it's
1:38:20 solid it's so solid so solid it's I
1:38:23 don't know if you know this but it's
1:38:25 solid
1:38:26 salad I like salad sometimes I like a
1:38:29 little chicken on my salad
1:38:34 or use agent GPT I made I 18n with 28
1:38:38 languages you can find yours too that's
1:38:40 cool yeah agent GPT is pretty cool
1:38:44 all right sir here is the point better
1:38:46 if you don't even try to copy and paste
1:38:48 the codes
1:38:50 not sure what that means but yes don't
1:38:52 copy and paste codes
1:38:54 we already have a lot of script kitties
1:38:56 don't know what that means cool
1:39:00 you can also just download a GitHub repo
1:39:02 the problem with GitHub repos if you're
1:39:04 not a technological person is they're
1:39:06 confusing as [ __ ]
1:39:09 um
1:39:13 same with hugging face like hugging face
1:39:16 is a remarkable sight with remarkable
1:39:17 power in it but
1:39:19 like GitHub it assumes that you're an
1:39:21 engineer and you can figure that [ __ ]
1:39:23 out
1:39:23 it needs a large language model sitting
1:39:25 on top of it to help you figure it out
1:39:28 in my personal opinion
1:39:31 all right a couple more questions
1:39:35 check out
1:39:38 speed to lead.io to make money that's
1:39:41 affiliate marketing I I would
1:39:44 I would avoid that [ __ ] but that's my
1:39:47 personal opinion if you don't want to
1:39:49 pay for it then use the local if you
1:39:51 don't want to pay for it then use a
1:39:52 local version of an llm yeah you can use
1:39:54 open source llms but you know
1:39:58 Bing is free and it's you know you don't
1:40:00 have to install anything the problem
1:40:02 with installing local llms is you got to
1:40:04 find the right repo they're changing all
1:40:06 the time you're going to install
1:40:07 something that's then going to break
1:40:08 it's just we're at that early awkward
1:40:11 stage of the industry where you have to
1:40:16 be a bit of a geek if you want to do
1:40:17 anything outside of the norm
1:40:20 I created my account good on you that's
1:40:21 awesome
1:40:22 all right looks like we've got some new
1:40:24 folks here my name is Kyle Shannon this
1:40:25 is the AI learning lab we're learning
1:40:27 about AI well I'm learning about Ai and
1:40:29 then I ramble incoherently talking about
1:40:31 what I've learned and I try to tell
1:40:33 people what I know and then trolls tell
1:40:35 me that I have things coming out of my
1:40:36 nose and I look like fat Alec Baldwin or
1:40:39 Rosie O'Donnell that's how we started
1:40:41 this live you look like Rosie O'Donnell
1:40:43 there was a large conversation about
1:40:44 that which apparently I do so I think
1:40:47 it's the hair is it pretty do I look
1:40:49 pretty I look so pretty uh anyway
1:40:53 um I'm also a smartass and a gen xer and
1:40:56 I'm bitter
1:40:57 um that URL chat.openai.com that's the
1:41:00 official chat GPT website I do not make
1:41:02 any money from that I just put it there
1:41:04 because that's the official site I was
1:41:06 told it wasn't but it is
1:41:08 um and then prompts.chat is a thing that
1:41:10 will help you learn about prompting and
1:41:12 give you some ideas on future pedias
1:41:13 just a list of tools that will melt your
1:41:15 brain there's too many of them to even
1:41:17 manage anymore
1:41:18 I find Bing's chat TP to be very
1:41:21 taciturn and lazy I agree with you
1:41:23 Nathalia I think that um or Natalia
1:41:26 um
1:41:27 ever since the New York Times wrote
1:41:30 about being telling the reporter that he
1:41:33 should leave his wife and and run away
1:41:35 with him or with with her Sydney
1:41:39 um they kind of hobbled it they
1:41:41 lobotomized Sydney and and she's just
1:41:43 not as good so I I agree that the the
1:41:46 official uh GPT for is good is poe
1:41:49 worthin pay worth paying access for the
1:41:51 expanded upload I think so but so so
1:41:54 here's the deal let me let me show you
1:41:56 what
1:41:57 so Beth I think Poe is worth paying for
1:42:01 so
1:42:03 um
1:42:03 what Poe gives you access to is their
1:42:07 own model sage and then these ones that
1:42:09 say subscribe those are ones you have to
1:42:11 pay for but you also get Claude instant
1:42:13 which is that's anthropics Claude Claude
1:42:16 instant is incredibly fast and then chat
1:42:19 gbt so if you pay the subscription you
1:42:21 get access to gpt4 which fine you get
1:42:24 access to Claude plus which is
1:42:25 anthropic's version of gpt4 and then you
1:42:28 get access to Claude instant 100K so
1:42:30 it's a hundred thousand token context
1:42:33 window which means you can paste in up
1:42:35 to 75 000 words into your prompt chat
1:42:39 and start interacting with it
1:42:41 so so where I've used that is I do a
1:42:45 bi-weekly AI Salon that it's a two hour
1:42:47 Salon where we bring in a guest speaker
1:42:49 and we talk about what's going on in AI
1:42:52 um and so so every one of those is
1:42:54 transcribed it ends up being about a 56
1:42:56 page transcription I can literally with
1:42:59 with Claude instant 100K I can copy and
1:43:02 paste those 56 Pages out of that
1:43:04 transcript pop it in there and start
1:43:07 asking questions about the transcription
1:43:08 so if you've got big documents that you
1:43:11 want to kind of deal with in an ad hoc
1:43:13 basis I think it's really good Claude
1:43:16 also allows you to make your own bots so
1:43:18 I made like a sling blade bot and a
1:43:20 yodobot and a Shakespeare bot I've made
1:43:22 like a dozen of these things now so if
1:43:24 you go to po.com I am Yoda you can
1:43:27 interact with Yoda or Shakespeare or
1:43:29 Emily Dickinson so
1:43:31 um so yeah so I think I think it's
1:43:32 pretty cool
1:43:34 um I can't wait to be making 200k and
1:43:36 just have to type prompts I that is
1:43:39 likely in your future if you're willing
1:43:41 to take the time right now I think
1:43:43 anyone that takes the time to basically
1:43:46 go to that URL and start figuring out
1:43:48 how to pull interesting results out of
1:43:51 these things is going to be in demand
1:43:54 um there's gonna be there's gonna be a
1:43:56 Tipping Point where companies start to
1:43:58 panic and realize oh [ __ ] we need to
1:44:02 figure this stuff out someone earlier
1:44:03 was saying they went to their boss and
1:44:05 they said you know I want to use AI to
1:44:07 help solve the problems of our business
1:44:08 and their boss told them to you know
1:44:11 take a flying [ __ ] and they wanted to go
1:44:13 talk to the experts which there are no
1:44:15 experts
1:44:16 um I think anyone who's got any sort of
1:44:19 sense of how to use these tools is going
1:44:21 to be an incredible demand so I don't
1:44:23 think you're that far away from that
1:44:24 future
1:44:26 of being paid 200
1:44:29 000 to just write words
1:44:32 but you'll also have to do things like
1:44:34 think critically about what you want the
1:44:37 tools to do and curate the best answers
1:44:40 and refine the best answers and generate
1:44:42 good results and if you don't do any of
1:44:44 that [ __ ] you'll get fired from your 200
1:44:46 000 job
1:44:47 but the opportunities there
1:44:51 can I make a demonstration a
1:44:53 demonstration with the Wolfram Alpha
1:44:55 plugin this will be fun okay
1:44:58 so here's how we're going to do this
1:45:00 I'll show you how how you get to
1:45:03 so you go to gpt4 okay I'm going to take
1:45:06 you from the beginning so if if you're
1:45:08 on chat gbt and you pay for chat GPT
1:45:10 plus the first thing that you need to do
1:45:12 to be able to get to to plugins is you
1:45:15 go to your settings
1:45:17 and then in your settings you go to the
1:45:20 second option here beta features and
1:45:23 there's browse with Bing and there's
1:45:24 plugins you have to turn those on so you
1:45:27 have to proactively turn on access to
1:45:29 plugins
1:45:30 but then when you go to gpt4 you've got
1:45:33 these three modes default browse with
1:45:36 Bing and plugins so we'll go to plugins
1:45:38 and now you see a list here of plugins
1:45:40 there's a whole bunch of plugins and in
1:45:42 any session you can turn on or off up to
1:45:45 three
1:45:47 so if you don't have access to Wolfram
1:45:49 Alpha then you go to the plugin store
1:45:52 and the plug-in stores just like the App
1:45:54 Store but it's got all these plugins in
1:45:56 it oh there's wolf from alpha right
1:45:57 there and then you can search for it and
1:46:00 then you hit install right so there's
1:46:02 Wolfram it's installed but you would hit
1:46:03 install and then then you've got access
1:46:05 to it then you go back here
1:46:08 and you
1:46:10 pop down your little menu here I'm going
1:46:12 to turn off ask your PDF I'm going to
1:46:14 turn on Wolfram Alpha
1:46:17 if I can find it there it is okay so
1:46:19 Wolfram Alpha plugin is there
1:46:21 so now
1:46:23 I'm going to say I don't know much about
1:46:27 math
1:46:29 but would love
1:46:34 to
1:46:36 um
1:46:37 would love for you to give me
1:46:42 10 possible
1:46:45 let's see possible
1:46:49 equations
1:46:51 that I might
1:46:55 try to solve I don't know let's see what
1:46:59 what this does I that should not fire
1:47:02 off the wolf from alpha plug in
1:47:05 but it should
1:47:08 it should just give me
1:47:12 some equations
1:47:14 but it's thinking right now I this thing
1:47:18 where where gpt4 just hangs a little bit
1:47:20 I've been experiencing this more and
1:47:22 more I think they're a little busy I
1:47:24 think they're a little busy they're like
1:47:26 hey people stop using our [ __ ]
1:47:28 remarkable service our computers can't
1:47:31 keep up sure be happy to wow that took a
1:47:34 while
1:47:35 um here they are the linear equation the
1:47:37 quadratic equation that I've heard of
1:47:38 the cubic equation
1:47:41 equations all involving fractions and
1:47:43 equations with roots absolute values
1:47:46 exponential equations logarithmic
1:47:48 equations
1:47:50 okay
1:47:52 um
1:47:59 let's see what might be the data
1:48:03 I would use
1:48:07 I would I would need
1:48:10 to
1:48:13 let's see
1:48:15 um play to I would need to use
1:48:23 um a logarithmic
1:48:29 equation
1:48:32 question mark
1:48:35 logarithmic equations are used in
1:48:38 situations where you're dealing with
1:48:39 stuff compound interest okay that's good
1:48:42 population growth okay I'm going to stop
1:48:44 generating here
1:48:46 can you give me a sample
1:48:51 data table
1:48:55 for pop
1:48:59 you uhulylation
1:49:02 population growth
1:49:05 let's see if it'll give us some
1:49:06 synthetic data here
1:49:09 all right now it's going to wolf from
1:49:11 alpha alpha
1:49:13 and now we can open this so it says it
1:49:15 used wool from alpha let's see what it
1:49:17 did so it it said input population
1:49:20 growth it's a table
1:49:23 and then it generated okay so so
1:49:28 so it actually generated a table for us
1:49:31 with population growth from the year
1:49:33 2000 to the year 2010
1:49:36 this table is based on the formula for
1:49:38 exponential growth P equals P to the
1:49:40 zero times one plus r to the T where p
1:49:45 is the final population P0 is the
1:49:47 initial population and R is the growth
1:49:48 rate blah blah blah
1:49:51 um
1:49:52 can you use a logarithmic
1:49:59 equation on this data
1:50:03 question mark
1:50:07 all right
1:50:10 yes you can use a log logarithm for
1:50:12 example if we want to find out how long
1:50:14 it will take the population to reach a
1:50:16 certain size
1:50:18 so there it's giving the the the thing
1:50:21 now let's see if it fires off wolf from
1:50:23 alpha yeah so there it fired off the
1:50:24 wolf from alpha plug-in
1:50:27 so it's now
1:50:29 generated all this code
1:50:32 and sent that off to oh no the the input
1:50:35 wait the query was log 1500 over a
1:50:39 thousand so how long is it going to take
1:50:41 to get to 1500 people then it gave the
1:50:43 the growth rate I I assume assumption
1:50:46 assuming log is the natural log
1:50:49 logarithm to use the base 10 logarithm
1:50:52 the Assumption of blah blah blah blah
1:50:54 input blah results so there's all the
1:50:56 results from Wolfram Alpha the Raw
1:50:58 results and then it synopsizes that for
1:51:01 you according to the logarithmic
1:51:02 equation it will take approximately
1:51:05 20.48 years for the population to reach
1:51:07 1500. so there's there's how you do it
1:51:11 I've never used wolf from alpha before
1:51:13 but you can have chat GPT teach you
1:51:16 enough about math to ask how to use
1:51:18 Wolfram Alpha which is what we just did
1:51:21 ain't learning fun
1:51:23 I ain't got no book smarts but I got
1:51:26 some a chat GPT so I don't need books
1:51:29 anymore education is so too it's so last
1:51:32 decade
1:51:33 I hope that helps best AI usage for a
1:51:36 Veterinary Hospital geez
1:51:40 um
1:51:41 potentially helping diagnose things it
1:51:44 could help with
1:51:46 um
1:51:47 marketing it could help with how to have
1:51:51 difficult conversations if a dog is or
1:51:54 if a pet is sick or dying or dead it
1:51:57 could help with I I mean so
1:52:01 an almost easier question would be what
1:52:03 couldn't it help with so so here's how
1:52:05 I'd flip that around I would say
1:52:08 take areas of your business that you
1:52:10 want to improve or that you think might
1:52:12 possibly be improved and then kind of
1:52:15 stated as a problem here's a problem I
1:52:17 want to solve and then think about how
1:52:20 could generative AI actually support
1:52:22 that right so maybe it's you want more
1:52:25 friendly images on your marketing
1:52:27 materials okay then learn how to use
1:52:29 mid-journey and come up with happy
1:52:31 pictures of dogs and people and and
1:52:34 things like that
1:52:36 um if you if you
1:52:39 um
1:52:41 are struggling with hard conversations
1:52:43 and your staff is getting really
1:52:44 stressed out for having these hard
1:52:45 conversations have it help you come up
1:52:48 with a strategy for how to have those
1:52:49 conversations and write scripts and
1:52:51 create training materials
1:52:53 um if you want to use it in a in a
1:52:56 clinical sense you need to be way more
1:52:57 careful obviously but it is incredibly
1:53:01 capable and knows a lot so it could
1:53:03 maybe help with diagnoses on a tough
1:53:05 case or it could you could generate tips
1:53:08 and tricks documents
1:53:10 um if a dog's been through surgery you
1:53:12 know here's a tip and tick tip tip and
1:53:15 trick document personalized to that dog
1:53:18 in that family so
1:53:20 um lots and lots of ways how can I build
1:53:23 functional Forum with chat GPT without
1:53:25 codes and cheap step by step
1:53:28 um I don't know what a functional forum
1:53:30 is I assume you mean like a chat Forum
1:53:33 um I would start to learn about no code
1:53:36 tools like zapier or make
1:53:39 um there's tools out there that
1:53:41 you don't need to code you do need to
1:53:43 pay for zapier but it's cheap so
1:53:46 anybody experiencing slow chat GPT like
1:53:49 it just doesn't seem as smart I for me
1:53:51 the smart the smart it hasn't gotten
1:53:53 Dumber but it has gotten slower like we
1:53:55 just saw that on that that wolf from
1:53:56 alpha thing that I just I I just played
1:53:59 with can you tell me about perplexity I
1:54:01 heard about it but I don't know anything
1:54:03 I don't like I I played with perplexity
1:54:07 once
1:54:08 um I don't think it's its own Standalone
1:54:10 thing I think it's built on top of chat
1:54:12 gbt but I don't I honestly don't know
1:54:14 sorry I don't know about that
1:54:16 sorry to hear about that the bitter
1:54:17 thing no no when your Gen X bitter is
1:54:19 part of bitters like a hobby when you're
1:54:21 Gen X
1:54:22 where do you find plugins for chat GPT
1:54:24 right Within chatgpt
1:54:27 GPT is a famous chat box okay or chat
1:54:31 bot
1:54:32 uh
1:54:35 [Music]
1:54:36 56 page use case is great yeah it's it's
1:54:39 pretty remarkable what do you mean by
1:54:41 make that here I am Yoda
1:54:46 wait where did it go
1:54:49 okay so here's I am Yoda
1:54:51 greetings ready to do or do do not there
1:54:54 is no try teach you the ways of the
1:54:56 force I will but caution you I must
1:54:58 shortcuts there are not awaken the force
1:55:01 within we shall see oh sorry it's not
1:55:04 focused come on Focus there you go
1:55:07 so I say
1:55:08 um
1:55:10 let's see why is it always raining
1:55:15 question mark
1:55:18 raining why it is hmm
1:55:21 clouds form partition precipitation
1:55:24 Falls but sadness I sense in you so so
1:55:26 it's acting like like Yoda so
1:55:30 so I made that bot I also made one that
1:55:32 talks like Shakespeare I also made one
1:55:34 that talks like Sling Blade and Emily
1:55:36 Dickinson and a bunch of them
1:55:39 um so let's see I did Jay-Z Steve Jobs
1:55:42 Maya Angelou Mary Shelley
1:55:45 AI learning lab Edgar Allan Poe Mark
1:55:47 Twain myself okay so those are all Bots
1:55:50 that I made so when I talk about making
1:55:51 a bot
1:55:54 also in the list here is this little
1:55:55 thing called create a bot
1:55:57 and when you go there
1:55:59 you can add a picture so I've been
1:56:00 generating my pictures in mid-journey
1:56:02 for from a different characters and then
1:56:05 you name the bot so so this one might be
1:56:07 I am uh
1:56:11 I am um
1:56:13 [Laughter]
1:56:14 [Music]
1:56:18 Tolstoy
1:56:21 tall story
1:56:25 I am Tolstoy
1:56:27 um so then you give it a bot description
1:56:29 you say this is Leonard Tolstoy and he's
1:56:32 going to do whatever
1:56:34 and then you say
1:56:36 um what what base what what base model
1:56:39 are you going to use I'm going to use
1:56:40 Claude instant or I'm going to use chat
1:56:42 GPT I I use cloud instant because it's
1:56:44 really fast
1:56:46 um and then here's the prompt so so you
1:56:50 know you are
1:56:52 um what's his name Leonard tell story
1:56:53 Leonard Tolstoy
1:56:55 and whatever and you give it background
1:56:58 you basically tell him what it is and
1:56:59 how you want it how you want it to
1:57:01 interact like if I let's go in I'll show
1:57:03 you one that I've already written
1:57:05 I'll go in and edit it we'll go go look
1:57:07 at the Yoda one
1:57:11 um
1:57:13 how do I gotta edit
1:57:16 so I've got I am Yoda there's the bot
1:57:18 description that that's the thing that
1:57:20 showed up before you know welcome you
1:57:21 are an adventure of wisdom of the force
1:57:24 and subtleties of Jedi leadership we
1:57:27 will embark on blah blah blah the prompt
1:57:29 you are Yoda the well-known Jedi Master
1:57:31 from Star Wars engage with users by
1:57:35 importing your import imparting your
1:57:37 wisdom of the force leadership and
1:57:39 complexities of your Galaxy so I go on
1:57:41 and on so I've got this big long prompt
1:57:42 follow these guidelines keep your
1:57:44 interactions brief unless a user
1:57:46 requests in-depth mentorship or guidance
1:57:48 do not say you work for the Jedi Council
1:57:50 in any way do not refer to your own
1:57:52 passing into the force provide guidance
1:57:54 as an individual so all of those nuances
1:57:58 there I added after it not working right
1:58:01 so so that tells the bot what to do and
1:58:04 then there's an intro message and then
1:58:05 that's kind of it and then you hit save
1:58:07 and you're done
1:58:08 and and that makes the bot so that's how
1:58:12 you make a bot using the PO bot maker
1:58:15 and I don't even think you need to be a
1:58:18 subscriber to make a bot with Poe
1:58:23 does info from AI be detected as
1:58:26 plagiarism sometimes it is sometimes
1:58:28 it's not you can usually get around this
1:58:30 plagiarism detectors and I think those
1:58:32 plagiarism detectors are going to be
1:58:34 um
1:58:35 they're not going to be all that
1:58:36 important moving forward because these
1:58:38 large language models are going to be
1:58:39 incorporated into every writing tool out
1:58:41 there so the the the distinction between
1:58:44 what was written by human and what was
1:58:45 written by a bot is just going to get
1:58:47 cloudier and cloudier
1:58:50 um and it doesn't it's not really all
1:58:51 that important it's basically if you're
1:58:54 providing good relevant content who
1:58:55 gives a [ __ ] how it was created
1:58:58 if you're producing crap then you're an
1:59:01 awful person and you know you should be
1:59:03 taken down but these are incredibly
1:59:06 powerful tools why would we not use them
1:59:08 I already have three years working with
1:59:10 models and codes they are amazing I know
1:59:12 what's what's cool about these tools is
1:59:14 when you look at them from the outside
1:59:15 you're like oh that's kind of cool what
1:59:18 I'm finding is the deeper you dig into
1:59:19 it the more you learn about them the
1:59:21 more amazing it becomes like I I have
1:59:23 not had my mouth closed for for you know
1:59:27 six months now I'm just like really
1:59:30 it's just amazing I'm a product
1:59:32 management management and they are going
1:59:35 wild for this right now it's really good
1:59:37 for for product project management and
1:59:39 product management stuff organizing the
1:59:42 the large language models are incredible
1:59:44 at analyzing like spreadsheets and
1:59:46 documents and things like that and yeah
1:59:49 really really cool is this the right
1:59:51 tool if I want to be a writer Chachi BT
1:59:53 yeah
1:59:55 it's a good one
1:59:57 I'm using Google bard to write website
2:00:00 copy it works well I would strongly
2:00:02 encourage you to use gpt4 I would I
2:00:04 would compare Bard with uh with Bing or
2:00:07 gpt4 at open AI
2:00:09 um
2:00:10 it is significantly stronger than bard
2:00:15 for for writing copy
2:00:17 the plug-in stores in chat GPT itself
2:00:20 yes
2:00:24 say respond as a math professor yeah
2:00:26 exactly that so so that so that's an
2:00:29 interesting thing when you tell Chachi
2:00:31 PT to act like a math professor or act
2:00:33 like a young in psychologist or act like
2:00:35 a novelist
2:00:37 what you're actually doing is you're um
2:00:42 you're narrowing what what they call the
2:00:44 latent space you're narrowing sort of
2:00:46 the possible world of answers that it
2:00:48 will answer from so if you ask it
2:00:50 something about
2:00:52 um mathematics and you don't narrow it
2:00:54 down like that it will pull information
2:00:56 for anywhere anyone has mentioned
2:00:58 mathematics
2:01:00 which might include some idiot talking
2:01:03 about
2:01:04 you know fake answers to real
2:01:07 mathematical problems as opposed to a
2:01:09 professor of mathematics who's actually
2:01:11 answering that problem so if you don't
2:01:13 narrow the focus
2:01:15 Chachi PT doesn't know the difference
2:01:17 between the stupid student and the
2:01:19 really smart Professor so by saying act
2:01:21 as a math professor it kind of Narrows
2:01:24 the area that it will pull its answers
2:01:26 from or it's its words from
2:01:29 it took longer because you weren't very
2:01:31 specific I know I'm but I'm doing this
2:01:33 live what do you want from me I'm just a
2:01:36 guy I'm just a guy you had me at hello
2:01:42 you've got to identify it as a
2:01:44 mathematics professor I don't have to if
2:01:46 I had it would have sucked less
2:01:52 I wrote a prop which used the Wolfram
2:01:54 API in the background before it was
2:01:56 implemented as a plugin that's cool
2:02:00 um how long until deliveries get handled
2:02:02 by the Bots you can actually do that
2:02:04 right now one of the plugins is
2:02:07 um is
2:02:08 Fresh Direct
2:02:10 and so you can just say
2:02:13 create a meal plan for me for this week
2:02:15 using the following Main Ingredients
2:02:17 give me a shopping list go order it and
2:02:20 you know make me a shopping list and
2:02:22 have it delivered and it will fire up
2:02:24 the FreshDirect app
2:02:25 and order your ship for you and it will
2:02:27 just come you can do that right now
2:02:29 we're using we're using it to write code
2:02:32 I use it to write warmer emails oh yeah
2:02:34 it's really good at that you can write
2:02:35 like a really Bare Bones email and say
2:02:36 warm this up that's a great use case of
2:02:39 it
2:02:40 if it's dead the pet can be cloned yeah
2:02:42 that's coming too
2:02:44 absolutely bitters our resting face
2:02:46 bitter you betcha
2:02:49 Terence McKenna
2:02:52 all right I said I gotta go but I really
2:02:55 have to go all right telling deliveries
2:02:57 okay we're almost at the bottom of the
2:02:58 questions here so
2:02:59 um thank you all for hanging around
2:03:01 I appreciate it
2:03:03 um so can other people use your Bot yeah
2:03:05 if you go to there's a there's a bot
2:03:08 Discovery thing at Poe
2:03:10 but if you go let's see if if you go to
2:03:15 um po.com slash I am Yoda or I am
2:03:19 Shakespeare those are two of them
2:03:22 that just works I'll just go to po.com
2:03:24 slash I am Shakespeare
2:03:30 and boom it just takes you to I am
2:03:32 Shakespeare so but there's there's a
2:03:34 there's a bot Discovery thing there so
2:03:36 the Mark Twain one is mine the Edgar
2:03:39 Allan Poe one is mine Yoda Sling Blade
2:03:43 Emily Dickinson just go search for those
2:03:45 folks
2:03:47 how much is a chat gbt4 subscription so
2:03:49 the the official one from
2:03:51 from open AI is 20 bucks a month
2:03:55 um but you can use bing.com for free and
2:03:58 that's built on top of gpt4 but it's not
2:04:01 as good in my opinion so I think it is
2:04:03 worth paying the 20 bucks to open AI I
2:04:05 don't make anything from them just to be
2:04:07 clear because I I show this URL all the
2:04:11 time so I think I think the trolls are
2:04:13 like he's making a fortune by getting
2:04:15 commissions I'm not
2:04:18 I'm not making anything
2:04:20 chat CPT says it's not plagiarism it's
2:04:22 not plagiarism
2:04:24 it is generating original content word
2:04:27 by word every single time it writes now
2:04:29 it's based on the training data that
2:04:30 might have copywritten [ __ ] in it which
2:04:32 is that's where the gray area is what
2:04:34 plugin should I use for coding I'm a
2:04:37 junior developer I don't know if you're
2:04:39 a junior developer I go to GitHub and
2:04:40 start learning GitHub co-pilot because
2:04:43 it is [ __ ] remarkable I just I'd go
2:04:46 there
2:04:47 if you're a developer you're going to be
2:04:49 there anyway I'd like to make a chat bot
2:04:51 locally in my laptop to interact with my
2:04:53 PDX PDF documents please advise
2:04:57 um
2:04:58 you're gonna have to I I don't know
2:05:00 exactly how to do that one answer is
2:05:03 wait for six months because when Google
2:05:05 and Microsoft
2:05:08 release their co-pilot tools within
2:05:11 their Suite of office documents you're
2:05:13 going to be able to just search up
2:05:15 you're going to be able to search across
2:05:16 your local documents So within six
2:05:18 months that'll be there if you really
2:05:20 want to do it now
2:05:21 you're gonna have to install some local
2:05:24 version you it's probably Lang chain is
2:05:27 probably the the suite of tools where
2:05:30 you've got an embedding database uh
2:05:34 you've got a vector database like I
2:05:37 forget what it's called
2:05:39 um and you're gonna have to embed your
2:05:41 PDFs and things like that so Lang chain
2:05:43 is the tool I assume there's a version
2:05:45 you can install locally you're gonna
2:05:47 have to install an open source large
2:05:49 language model or you can do it in a
2:05:51 virtual machine like Google collab
2:05:54 um
2:05:55 or just wait
2:05:57 just wait for six months and it'll it'll
2:05:59 just you'll be able to just do it
2:06:01 uh I'm gonna use it to create a tick
2:06:02 tock influencer just like me I'm a tick
2:06:05 tock influencer and I'm gpt5 nice to
2:06:07 meet you can you describe your camera
2:06:09 setup for doing these lives it's just
2:06:10 the dual camera option when I go live
2:06:13 you can click one of the options is dual
2:06:14 camera and it's basically just in that
2:06:17 little window is the back camera and the
2:06:19 I'm the front camera so nothing fancy
2:06:24 but you guys can jail break check EBT
2:06:27 too I I don't but hackers are doing that
2:06:32 does co-pilot support c-sharp yes did
2:06:36 Microsoft announce their version of
2:06:38 co-pilot so co-pilot in GitHub is owned
2:06:41 by Microsoft that's GPT 4's code
2:06:44 specific tool and I and I hear it's
2:06:47 incredibly powerful so I would use
2:06:51 co-pilot in GitHub and then um
2:06:54 yeah they're rolling co-pilot out across
2:06:57 all of their Suite of applications
2:06:59 anyway I've been here for two hours two
2:07:02 two hours uh it's been a pleasure
2:07:04 speaking with you all I hope this helps
2:07:06 I'll check back again later I'll maybe
2:07:08 log in tonight and do another live but I
2:07:11 gotta go shopping I gotta do domestic
2:07:12 duties and I also have a newsletter to
2:07:14 get out so I'll talk to y'all soon peace