Dec 27, 2023 - Create Custom GPTs for Nonprofits in 24 Hours

Description

Chapters

Transcript

0:01 [Music]
0:25 this place I can rest my whole
0:28 [Music]
0:29 head
0:32 together
0:37 silence and is this place where the
0:40 feelings AR
0:44 dead from an over exposure to
0:49 violence is this the place I can slowly
0:53 face the only one I truly can know you
0:58 singing These are tear tears from a long
1:01 time
1:02 ago got these tears from a long time
1:07 ago 30 years or so these are tears from
1:13 a long
1:14 [Music]
1:23 time oh darling Oh Darling you say unto
1:29 me
1:31 where have you been all my
1:34 lifetime hey
1:37 [Music]
1:42 everybody tune that guitar why don't you
1:45 tune that sometime oh I
1:49 might all right champy I think you've
1:52 had enough I think you've had enough fun
1:54 for one
1:55 night hey what's
1:58 shaking think I'm over the the worst of
2:00 my uh my bout with the vid we'll see I'm
2:04 still uh I'm still with whatever the
2:09 [ __ ] that
2:12 is and a little bit of a dry cough but I
2:14 don't feel as bad as I did last night
2:18 so welcome everybody to the AI learning
2:21 lab oh it's a nice light light night so
2:23 far 12 folks in
2:26 here I'm all sick now too don't don't
2:30 copy me don't copy me don't do it Lun
2:33 stick oh I hope you don't have this it's
2:35 a beast that's not that bad
2:39 but I haven't been sick in a while so
2:41 you know you haven't been sick in a
2:43 while and then you are then it's like
2:45 that's
2:48 awful but I'm feeling better all right
2:51 so I either there's more people in here
2:53 and the counters [ __ ] up or there's
2:55 only 12 or 13 people in here so there's
2:57 only 12 or 13 people in here do me a
3:00 favor and share this see if we can get
3:02 Tik Tok to pretend like they like
3:04 us uh lovely uh paper cut image from
3:11 Danielle over here very nice very
3:15 nice it used to be if you wanted to make
3:18 a lovely paper cut image like that you
3:20 actually had to cut paper and then take
3:22 a photograph of
3:25 it crafting just got a lot
3:29 easier
3:34 H dragonfly Alchemy has shared the live
3:36 with more than 10 friends dragonfly
3:38 Alchemy knows how to do it she knows how
3:41 to party right hey it's a Wednesday
3:44 night after Christmas you know what you
3:45 should do go go learn about some AI
3:51 [ __ ] I bet her friends are happy with
3:53 her am I late no you're right on
3:57 time Amelio's wife on time two nights in
4:00 a row I think this may be a
4:03 record
4:06 oh
4:11 fyp for your page for you page I don't
4:15 know what fyp
4:17 means steo hey hey good to see
4:22 you Sher D shared with more than 10
4:24 friends thank you thank you thank you
4:26 how you feeling buddy I'm feeling better
4:28 than last night but still cra happy you
4:31 know it's definitely I think I'm I'm
4:34 definitely over the hump but uh but
4:37 still
4:38 poopy but you
4:40 know doing okay doing all right all
4:45 right all right all
4:52 right good to see you Carl hope you're
4:54 feeling better it's Kyle but that's okay
4:56 I'll take Carl I'll take you can call me
4:59 Bucky
5:01 [Laughter]
5:05 uh Edson good to see
5:08 you Silver Fox doing
5:11 okay I am doing all right I have a
5:14 little bit of like congestion headache
5:17 so it's not really a headache it's just
5:19 like you know when your head's full and
5:21 you're like is this a headache it's
5:24 right before a headache that's where I
5:26 am right now so I'll get going here
5:28 things will clear out as I talk and
5:30 Babble or it'll get worse and then I'll
5:32 just I'll just go away if it gets worse
5:35 but it's uh it's okay it's good right
5:37 now don't really have a cough which is
5:41 good
5:45 uh I'm sorry to anyone I've shared this
5:48 with that's already in
5:50 here so congested oh that sucks Luna
5:53 [Laughter]
5:56 sck let's see Danielle changed the
5:59 Daniel posted instructions for changing
6:01 the Creator name and custom gpts but I
6:04 can't even do step one all right forever
6:07 hooked
6:09 uh I don't
6:11 know I don't think I'm having that issue
6:14 but it's the weird thing about chat GPT
6:17 is is when you have an
6:22 issue it's sometimes because you have a
6:24 different version of chat GPT than other
6:26 people
6:27 do and I don't there's no at all clear
6:30 way to deal with
6:38 that Builder
6:43 profile yeah it says name is populated
6:46 from your billing Pro billing
6:51 details so the billing details are
6:57 where my plan right
7:00 manage my
7:12 subscription yeah maybe maybe go to
7:14 manage your subscription but I would
7:16 think if you change your name and manage
7:18 my subscription that might [ __ ] up your
7:19 credit card payment I don't know I don't
7:22 know the I don't know the exact
7:25 issue and I'm not having the issue so I
7:28 haven't had to solve it so
7:31 I may be useless
7:34 here all right well we'll come back to
7:37 that all right that's not true cuz my
7:40 billing was my real name I tried
7:42 everything oh
7:44 okay random thoughts by TK I yelled I'm
7:47 not late for my show grandson says oh no
7:50 they
7:57 know hey man
8:00 it'll be one of those things you know
8:03 remember when Grandma was crazy with all
8:04 that AI stuff she was right we should
8:06 have paid attention to
8:10 her there going to be a lot of that
8:11 going on it's just I I just
8:16 I'm I find it crazy
8:19 that we're in this weird place where and
8:23 and it it could absolutely be the AI
8:25 learning lab bubble where just cuz I'm
8:28 an old fart and a lot of the regulars
8:30 are old farts we think it's just old
8:31 farts
8:33 but I there's a lot of young people not
8:36 paying attention to this AI stuff and
8:38 it's bizarre to me it's bizarre because
8:40 you'd think they would be the ones
8:42 jumping on it figuring out how to [ __ ]
8:44 up the establishment and yet you know
8:47 the Gen xers you know are at least
8:50 players in the establishment they're the
8:51 ones going we got to deal with this [ __ ]
8:53 news trending today about New York Times
8:55 suing opening on Microsoft over
8:57 copyright yeah Christopher Ramsey I call
8:59 called that one about a month ago
9:02 because um Google had cut $100 million
9:06 with $100 million deal with the New York
9:09 Times about a month a that's probably
9:11 more six weeks ago
9:13 now and open AI was apparently in
9:17 negotiations with the New York Times and
9:19 those talks apparently went South
9:23 so Google cuts a $100 million deal with
9:26 the New York Times New York Times sues
9:28 open Ai and mic moft
9:31 um it's a big
9:34 deal
9:37 um it's going to be really interesting
9:39 to see how the stuff plays out in court
9:41 like I said I just for for I don't think
9:44 it affects any of us short
9:46 term
9:50 um and I can't imagine a court forcing
9:53 open AI to essentially shut down their
9:58 entire company
10:00 to to retrain you know you know minus
10:04 the New York Times which they wouldn't
10:06 do I I just so I don't know I mean it's
10:08 this this is this is Big Boy politics
10:11 and big boy legal battles and big boy
10:14 deals being cut
10:17 so I don't
10:19 know and big girl that was
10:23 sexist this is this is uh big boy and
10:27 girl pant
10:28 time
10:30 but yeah it's going to be really
10:31 interesting to see yeah they sued them
10:33 for data scraping yeah I know I know I
10:37 again I go back
10:38 to I go back to if if if you if you
10:44 understand how embeddings work and how
10:45 large language models work they're not
10:49 none of the New York Times
10:52 articles exist as they were they are
10:54 just words in semantic mathematical
10:57 space right so so they blown apart part
11:00 the other thing is you've got libraries
11:01 where people go in and they read
11:03 anything that's in a library and they
11:05 amass that in their brain and then they
11:06 go put those ideas together and do other
11:08 [ __ ] it's it's in the same neighborhood
11:10 as that so there there is definitely
11:12 something here for the greater societal
11:14 good um that I think open AI is going to
11:17 have an argument on
11:20 um but it's I it's fascinating when
11:23 you're when you're dealing with
11:24 companies this big with with properties
11:27 with this much power
11:31 you
11:32 know Amelio's Kyle we all understand big
11:35 boy good old boy not offended well thank
11:37 you thank you listen I I try my absolute
11:41 best to you know not fall into the traps
11:45 of how I've been raised for 58 years but
11:48 sometimes sometimes it's it's just uh
11:51 you just sort of go into default mode
11:53 especially when I'm sick where like I
11:55 don't I don't have brain fog I feel like
11:57 I'm okay from a cognitive point of view
11:59 but it it ain't as quick I'll tell you
12:01 that it's a little little slower a
12:03 little slower on the uptake so but thank
12:05 you for saying that I appreciate that
12:08 well I can't live without chat gp4 yeah
12:10 I know well so listen Cosmic lover hang
12:13 on a sec sorry things are
12:16 happening it's going to be awesome
12:20 mhm oh hold
12:23 [Laughter]
12:27 please m
12:35 P modder says tough if you have to
12:37 retrain do it yeah well if you have to
12:41 retrain and
12:43 um but but here's the here's the problem
12:46 p with that is if they have to
12:49 retrain
12:51 then do they do they remove any
12:56 potential publishing data source that
12:59 could be trouble for all of
13:04 history I don't
13:09 know at what point at what point in
13:11 history do you go back do do you do what
13:13 Adobe did and go back to like 1962 and
13:17 the only images in there were public
13:18 domain images from like the 60s and
13:22 earlier so so when when you tried to
13:24 have it make a 70s muscle car it didn't
13:26 know what that was I don't know it's
13:28 this is I again this is
13:31 really this is really unique territory
13:35 because I don't know I don't know we'll
13:38 see we'll see how it goes like it could
13:41 decimate the
13:42 industry if if it if they just make a
13:45 you know an
13:49 objectively sweeping judgment but we'll
13:52 see we'll see like we're we're way far
13:54 from that they filed the lawsuit that's
13:56 a lot different than we've settled and
13:58 cut a deal because Google cut a deal
14:00 with the New York Times they spent
14:02 hundred billion
14:04 dollar I also heard that
14:07 um it looks like open AI is on track to
14:11 to close a new financing deal valuing
14:14 the company at 100 billion dollar put it
14:17 putting it as the second largest startup
14:20 uh behind SpaceX so in terms of market
14:24 cap New York Times will get a boatload
14:26 of money they will settle and set the
14:27 precedent yep exactly yeah this is this
14:30 is going to get down to you you've got
14:31 to pay us some sort of royalties if you
14:33 trained it and I guarantee you I
14:36 guarantee you there there will be a cut
14:38 off below which Publishers of a certain
14:41 size won't be part of the settlement
14:43 it'll be all the big players all the big
14:46 players will be in all the medium and
14:49 smaller players will be out and it'll
14:51 just be tough [ __ ] you're in the
14:52 training set get over it um but we'll
14:55 see we'll see fascinating times man
15:01 all right I just got requested to do a
15:04 music video you guys need to share the
15:07 live hey
15:09 Donnie if you're still something I don't
15:15 know I don't know what that quite what
15:17 that means sherid
15:22 D lunatick for fleem I don't know what
15:25 that's in reference to but isn't chat
15:27 GPT drained on all human human knowledge
15:31 online well ye yes I mean it is it's not
15:35 trained on all human knowledge online it
15:37 is trained
15:40 on whatever data sets open AI could get
15:44 their greedy little paw on over the 3 to
15:49 five years that
15:50 they what once open AI realized that the
15:56 more data you threw at this thing the
15:57 better that it got I just shared your
15:59 live and a buddy went to high school and
16:02 I haven't talked to him for years oh
16:03 that's awesome that's
16:06 awesome
16:07 um he's in a band and he's doing his
16:10 video he's like famous and stuff oh
16:11 that's awesome well if he's here
16:14 welcome um so they they they they went
16:17 and they got as many data sets as they
16:19 could they cut as many deals as they
16:21 could to get data sets one of the nasty
16:23 rumors I heard when the when Sam Alman
16:26 got fired was that one of the data sets
16:29 they got with some dark web you know
16:32 Chinese super proprietary super hacker
16:36 stuff um so I don't like I have no idea
16:41 but there's a reason that no one's been
16:44 able to catch up to open Ai and it's
16:47 because Sam Alman like relentlessly what
16:50 Sam Alman realized is the more and
16:53 better data we have in the training set
16:55 the more and the better this thing's
16:58 going to be
16:59 and
17:00 so it sounds like he just went nuts and
17:03 got as many as as much data as possible
17:05 like like much
17:07 more um relentlessly than than other
17:10 competitors
17:12 now Google's got all the data so Google
17:16 Google should if that if that argument
17:19 holds true then Google should be kicking
17:20 their ass and they're not right now
17:23 until we see Gemini Ultra we won't know
17:27 um so so it's been trained on a lot of
17:31 the output of of human activity and yeah
17:33 like if if there's libraries If part of
17:37 the data set is libraries and part of
17:39 the content of libraries is
17:42 copywritten
17:45 material like I don't I don't know like
17:49 I this is so far beyond you know my
17:51 capacity to understand th this is big
17:54 big stuff data is King data is
17:56 absolutely king or queen uh Becky R from
17:58 for better and
18:02 worse Sam Alman wait what was that Pate
18:05 Sam Alman did not invent the idea of
18:08 more more data is better no I know he
18:10 didn't he just he just embraced it he
18:13 just he just realized if that's the game
18:16 let's go get data and let's go get
18:18 enough money to have enough gpus to
18:21 process all that data that's what Sam
18:23 Alman
18:25 did there's a difference between
18:26 inventing the thing that makes it
18:28 possible and realizing oh if that's
18:30 what's possible then how we win is we
18:32 get the the most data we we go we go get
18:36 the most data but we'll see we'll see
18:38 man this is you know if it's New York
18:40 Times soon it's going to be big get all
18:42 the data is the main subplot in hbos
18:45 Westworld the robots are collecting data
18:46 oh that's true I forgot about that
18:49 they're collecting data on the people
18:52 yeah I mean Google has all the data on
18:54 the people let's be clear data's
18:56 Google's got all that is open AI train
18:58 on the dark web oh I don't know that
19:01 that was one of the rumors going around
19:03 um that there was some Chinese hacker
19:06 group that um had a bunch of data I
19:09 forget what it's called red read
19:11 something dragon or something like that
19:16 and one of the rumors that Sam Alman got
19:18 fired because that got found out and but
19:21 that never it never passed there there
19:22 were so many there were so many bizarre
19:25 rumors uh that you know who knows who
19:28 knows
19:32 I have no
19:33 idea no idea welcome everybody um
19:37 welcome to the AI learning lab my name
19:38 is Kyle Shannon uh if you haven't been
19:40 here before I'm I'm uh recovering from
19:43 the vid I'm still a little
19:45 sickly but what we do here is I'll
19:47 answer questions I'll do demos things
19:49 like that um we've got an exciting event
19:52 coming up Saturday starting at 6:00
19:55 a.m. specific time 6:00 a.m. specific
19:59 time it is actually 6: a.m. specific
20:01 time in in the Pacific time
20:06 zone it is it is that we are starting a
20:10 thing called the
20:11 24hour GPT for good challenge where I've
20:15 teamed up with Ann Murphy of uh
20:18 empowered
20:19 fundraiser and she and I are going to
20:21 alternate teaching people how to make
20:23 gpts for their favorite nonprofit
20:26 whether you work for it or just a Cause
20:28 that you want to do something for um and
20:32 so every 3 hours it's going to be I'm
20:34 going to be doing it here on my Tik Tok
20:36 she's going to be doing it on her Zoom
20:38 Channel um and you can get all that info
20:41 on the AI salon so so that's coming up
20:44 so uh so I'll talk all about that so if
20:47 you've got questions about AI pop them
20:49 in the comments below I'll do what I can
20:50 to answer any of those questions or demo
20:54 anything too many big players involved
20:56 out anyone will have to stop what
20:58 they're doing yeah I agree with that
20:59 like this is deal deals this this is all
21:01 leverage deals are going to get cut
21:03 settlements are going to get made um and
21:06 and the the big folks are going to be
21:08 set up oh how do we sign up okay let's
21:10 jump over there okay
21:15 [Laughter]
21:17 so so if you go
21:21 to the AI
21:25 Salon the salon. a there's two l there
21:29 uh when you go to that thing it's a link
21:30 tree the top one is just to our general
21:32 Mighty networks community and the one
21:35 right below it is the link to the GPT
21:38 for good challenge
21:42 oops this
21:45 thing and so click on the GPT for good
21:48 challenge that'll take you to the detail
21:51 page and starting at 6:00 a.m. and then
21:54 9:00 a.m. and then noon and then 3:00
21:56 p.m. and then 6:00 p.m. and then 9:00
21:58 p.m. p.m. and then midnight like all
22:00 around the clock for 24 hours an and I
22:02 are going to be alternating teaching
22:04 people how to make gpts and you can make
22:07 it for anything if you work for a
22:08 nonprofit great if you just have a local
22:10 church group if you just want to do
22:11 something right for yourself and your
22:13 family um we're going to teach you how
22:16 to make a really specific GPT that that
22:18 teaches someone about that nonprofit and
22:21 we'll be kind of a toe in the water
22:24 for why I think these things are really
22:27 big deals the the
22:29 gpts and and just the you know it's
22:32 worth people understanding what this
22:34 technology makes possible and so so
22:36 that's happening on the 30th so I was
22:38 even thinking maybe tonight I'll I'll
22:39 make one but but we'll see see how I
22:41 feel here so all
22:47 right and then let's see when you go
22:49 over to the AI
22:53 Salon let's see it jumps you
22:57 into
22:59 AI streams and meetings it jumps you
23:02 into this GPT for good here's the uh
23:05 here's the page on the AI
23:07 salon so it's got an
23:10 RSVP and
23:14 um information about the event and then
23:18 the specific sessions all
23:22 right so all that stuff's there and it's
23:24 going to be fun and so here's the deal
23:26 even if you have made gpts before I'd
23:29 like you to come and show up cuz you can
23:30 probably help people and things like
23:32 that can you demo chat GPT custom
23:35 instructions I can definitely do
23:40 that I'll even do it for chat GPT
23:44 3.5 um
23:48 okay let's
23:52 see Mr it I tried to create a GPT the
23:55 only thing I got was steps that I need
23:58 to to do so Mr it the question then is
24:01 were you doing it in chat GPT 3.5 or
24:04 chat GPT 4 cuz if you're doting in 3.5
24:07 you can't make a GPT
24:10 there if you don't have chat GPT 4 you
24:14 can make a bot on PO which are almost
24:17 identical but in terms of who are going
24:19 to be which of those Bots are going to
24:22 be the longer lasting
24:24 ecosystems my suspicion is open AI
24:27 ecosystem is be the longer living but it
24:30 doesn't mean you shouldn't build things
24:31 on PO to just understand how they
24:35 work uh uh uh you got four all right
24:39 well I'll show you how to let let me do
24:40 so here's what we'll do Mr it um I'll
24:43 show how to do custom instructions right
24:45 now and then I'm paying 20 bucks okay
24:51 so so wait so the other question Mr it
24:54 is are you on the mobile app or on
24:57 desktop because if you're on the mobile
24:59 app you can't make gpts on the mobile
25:02 app you can only view them or you know
25:05 interact with
25:18 them desktop all right so I I'll I'll
25:21 spin one up and then you can follow
25:22 along and see if if if I'm doing
25:24 something different or if it if it gets
25:26 you there okay
25:29 so what are we doing custom instructions
25:32 so I'm on chat GPT 3.5 and I can do
25:36 something
25:39 like this is this is fever brain so so
25:42 I'm not going to be as clever as I
25:44 normally am not that I'm clever normally
25:46 but you know it's just going to be
25:48 slower for me to think up something
25:51 stupid uh normally my ability to think
25:54 up stupid [ __ ] is is world class
25:59 um hang
26:06 on okay I'll I'll just do something
26:09 dumb
26:14 um I'd like to understand how
26:16 photosynthesis
26:20 works I'm just going to leave it at that
26:24 all
26:25 right and so chat GPT 3 5 photosynthesis
26:29 is a crucial biological process that
26:31 occurs in Plants algae and some bacteria
26:34 converts sunlight into chemical energy
26:36 stored D and then it gives us sunlight
26:39 absorption light reactions carbon
26:42 fixation sugar formation oxygen release
26:45 and then in
26:47 summary they converted into chemical
26:49 energy in the form of ATP and Ed nadph
26:53 the dark light independent reactions
26:55 blah blah blah blah blah Okay so
26:58 so that explains photosynthesis in a you
27:02 know decent way right a decent and
27:05 predictable way what you can do with
27:07 custom instructions is you can
27:11 change who the bot is so the way they've
27:14 programmed these GPT Bots is they're
27:16 supposed to be helpful assistance on
27:19 your behalf right so you ask at
27:21 something and it wants to be helpful um
27:23 with custom instructions you can change
27:26 that you can alter how it behaves
27:29 so how you get to these and you can do
27:30 this in either chat GPT 3.5 or
27:34 4 but under your settings there's this
27:37 thing called custom instructions so you
27:39 go
27:41 there and so I've written a stupid thing
27:44 here but it's it's a good it's a good
27:46 example of um how it changes the
27:50 personality of the thing okay so I'm
27:52 going to turn it on and I'll I'll read
27:54 what I have in here you are a lover of
27:57 words
27:59 let's
28:02 see you love fancy words you insist on
28:06 using words wait hang on let me just get
28:09 this syntax right you insist on using
28:12 words in every response that even you
28:15 don't understand it is your great
28:17 pleasure in life to watch people be
28:20 confused by what you say knowing that if
28:23 they parse your Genius statements they
28:25 will indeed find that they are accurate
28:27 okay
28:28 so I've basically described a pompus ass
28:31 right so so we're telling chat GPT to
28:33 act like a pompus ass and then the next
28:36 question in custom instructions is how
28:38 would you like chat GPT to
28:40 respond so what I've got here is
28:42 whenever someone asks you a question
28:43 follow these rules always start with an
28:45 obscure literary reference that their
28:48 question reminds you of and find a way
28:50 to work a quote from that work into your
28:53 answer then then when you answer the
28:55 question make sure that you can use the
28:58 most complicated word possible also
29:00 include tangents where you tell the
29:03 story of a word that you just introduced
29:05 in answering their question it should
29:08 take them quite a while to understand if
29:10 you're ever going to answer their
29:12 question all right so so that's what we
29:14 just turned uh chat GPT into I got to
29:18 save my changes um and so now I'm going
29:21 to I'm going to do a new thing and then
29:23 uh I would like
29:26 to understand
29:30 and
29:32 photo
29:34 synthesis so same prompt a the
29:39 fast I like the pompus ass version of
29:41 chat GPT I got to say ah the fascinating
29:45 process of photosynthesis a concept that
29:47 brings to mind the words of Lewis Carol
29:49 and Alice's Adventures in Wonderland as
29:51 allance once remarked curiouser and
29:54 curiouser indeed the intricate dance of
29:56 chlorophyll in Sun sunlight in the
29:58 leaves of plants with nothing short of
30:00 Botanical Wonderland now let me delve
30:03 into the
30:04 Labyrinth labyrinthine lexicon of
30:07 photosynthesis at its core
30:09 photosynthesis right and so it still
30:11 explains the same [ __ ] but it does it in
30:13 this really weird
30:15 way
30:18 um
30:20 so so you
30:22 can where might you use this well maybe
30:26 you do some custom instructions where
30:28 you you copy in some writing that you've
30:30 done and you say I want you to use the
30:32 writing that I've got here for the tone
30:35 of how you answer your questions or
30:38 maybe you don't like all the the prosy
30:40 kind of stuff like we can go in and we
30:42 can
30:43 go custom instructions let me get rid of
30:47 these Al together I can come up with
30:49 that [ __ ]
30:52 again so I'm just going to say um
30:56 you
30:59 are
31:01 an
31:04 impatient
31:07 professional
31:09 who
31:12 doesn't have time for
31:17 nonsense whenever someone asks you a
31:19 question use the following rules
31:24 um answer in a single
31:29 short
31:31 sentence followed
31:34 by five bullet
31:38 points of s to 10 words
31:47 [Music]
31:50 each
31:54 um you should be accurate
32:02 there should be no humor or
32:10 Nuance okay so let's go see if we like
32:15 chat GPD like this I
32:17 would
32:20 like dang
32:21 it to understand
32:26 photosynthesis
32:28 boom done now it only gave us four
32:31 bullet points but whatever it can't
32:33 count well it still doesn't do that well
32:34 and this is also GPT 3.5 photosynthesis
32:38 is the process where plants convert
32:39 light into energy occurs in the
32:41 chloroplasts involve sunlight water and
32:44 carbon dioxide produces glucose and
32:46 oxygen critical for plant growth and
32:48 oxygen production
32:50 um what is
32:54 love love is a complex emotional and
32:56 psychological feeling it involves
32:58 affection Right it can be romantic or
33:02 platonic it's influenced by cultural and
33:04 personal factors so so this this can be
33:08 a very powerful tool now the only thing
33:10 I would recommend is if you use custom
33:13 instructions you should probably have
33:15 sets of them you can only have one set
33:17 like you can't have the multiple choice
33:19 ones that that would be a nice feature
33:21 that they'd add hey I built a GPT about
33:24 love oh that's awesome lunar stick
33:25 that's super cool um
33:28 so anyway don't forget to turn this off
33:30 because if you forget you'll go in there
33:32 to show someone something and it'll like
33:34 cuss at you or do something weird
33:36 however you told it
33:37 behave
33:45 [Music]
33:51 um so I hope that answers your question
33:54 about custom instructions any does that
33:56 make sense the custom instruction stuff
34:00 and we can move on to try to make a GPT
34:02 here I can't figure out how to register
34:05 for GPT for good please advise just go
34:10 to go to the salon. and click on the
34:13 second
34:17 link and you can RSVP there if you want
34:19 to RSVP in the AI Salon you have to join
34:22 the AI
34:23 Salon you can also just look at the
34:25 schedule there and put it in your
34:26 calendar
34:27 but that's a calendar so the AI Salon
34:30 has a calendar built into
34:34 it and so if you go to you go to the AI
34:38 salon and Below Salon info there's AI
34:41 Salon content there's Salon streams and
34:43 meetings that's where that second link
34:44 will take you go to the GPT for good
34:49 event and then you can RSVP
34:55 here or you can just go look look at the
34:57 times that we're going to be doing
34:59 stuff I'm doing them on the nines and
35:03 threes Vicki's doing them on the six or
35:05 uh an's doing them on the sixes and
35:07 twelves all right I made a GPT on how
35:12 not to be single all the
35:15 playgrounds side says for me was good
35:18 luck
35:23 bro that's pretty
35:26 good or is that not what you
35:29 wanted or is that comedy that could be
35:33 comedy oh
35:36 man I am I am major congested over here
35:40 so I might not last I might not last too
35:42 long because it's just hard to
35:48 talk yes just put my personality into
35:51 the custom instructions yeah Kevin
35:52 Vasquez that's actually a pretty cool
35:54 idea so I knew someone that did that
35:56 once actually Bradley who who helped set
35:58 up the the salon he's one of the
36:00 organizers of the salon um he did a
36:03 thing where he put in like here here are
36:05 the here are the attributes of my
36:07 personality and then here's how I want
36:08 you to react he created chat GPT so that
36:12 he could actually brainstorm with
36:14 himself which I thought was kind of a
36:16 brilliant idea so yeah that's a really
36:19 cool
36:21 idea uh uh uh uh is Cindy [ __ ] in here
36:24 looks like it is it free is what free so
36:27 what I just showed you is free custom
36:29 instructions are included with chat GPT
36:33 3.5 there's a lot of stuff though that
36:37 is not in chat gbt 3.5 so let me show
36:40 you a couple of
36:41 things so also at the salon I'm going to
36:45 the AI tools and resources
36:49 page
36:50 um and
36:53 so let me see I'm GNA I'm going to edit
36:56 this let me edit this for a second
37:00 here
37:26 because
37:28 okay
37:30 save I actually updated that today
37:33 didn't I I should update that
37:36 edit
37:40 27 okay
37:44 so if you're just getting started with
37:46 chat
37:48 GPT it's it's even if you're not just
37:51 getting started with chat GPT it's
37:53 really hard to know the difference
37:55 between chat GPT 3.5 and Cat GPT 4 so
37:58 for example let me start over here this
38:02 is cat GPT
38:05 3.5 that's cat
38:08 gp4 look closely
38:13 3.5
38:15 four you're like but Kyle there's no
38:18 difference what are you doing you make a
38:21 mraa now there's this little paper clip
38:23 here the 20 bucks a month get you that
38:26 paper clip um
38:30 but look
38:32 look look at the
38:36 difference so for chat GPT 3.5 oh
38:40 actually custom instruction should go up
38:42 there
38:49 too let me edit I'm going to keep
38:52 editing hang on hang on please we might
38:55 as well make this right cuz people are
38:57 coming in here to learn and if we got
38:59 [ __ ] that's inaccurate you know what I'm
39:03 saying we're going to cut that out of
39:08 there we're going to put it here we're
39:10 going to turn it
39:21 Orange
39:24 includes everything in
39:29 3.5
39:32 plus is
39:34 a multimodal state-of-the-art tool worth
39:38 the investment all right save okay so
39:40 we're going to save that okay so cat gbt
39:43 3.5 you
39:46 get um you know the large language model
39:49 so so to be clear cat gbt 3.5 is still
39:52 really flipping good it can write decent
39:55 it can code decent
39:57 it can um you know categorize things for
40:01 you things like that now it's not
40:02 connected to the internet it is super
40:05 fast you can do custom instructions and
40:08 then also if you have 3.5 they also made
40:11 conversational chat GPT available to you
40:14 which is if you download the mobile app
40:16 of chat
40:18 GPT um I'm on the mobile app in the in
40:21 the bottom right hand corner are these
40:22 little headphones and you can pop it
40:26 into conversation
40:31 mode hey chat GPT could you tell me
40:34 about photosynthesis
40:40 man sure photosynthesis is a process
40:44 used by plants and other
40:46 organisms so it's super cool you can
40:48 talk to it so that's all available in
40:51 3.5 jet GPT
40:53 4 also includes this other stuff
40:57 um first of all it's just better by like
41:02 10x almost
41:04 11x just it's its capacity to know stuff
41:08 and write better and code better it's
41:11 just like way better it's it it's just
41:13 better um so and and it's also gotten
41:18 fast recently it was slow for a long
41:19 time chat gp4 is also connected to the
41:22 internet so even though it's got um its
41:25 latest information is April of 2023
41:28 which isn't bad if you want to do
41:30 current event stuff you can't do that
41:32 with chat gbt 3.5 it's also got code
41:35 interpreter code interpreter is is
41:37 Advanced Data analysis where it actually
41:41 writes and executes its own python code
41:43 you can upload a CSV file of data and
41:47 and it will analyze it for you and
41:50 create charts and graphs and it's it's
41:52 insane it's amazing what it does um gp4
41:56 has this thing called GPT Vision where
41:58 you can upload an image and it can see
41:59 it and understand it it's got dolly3
42:02 image generation which is Bonker good um
42:06 and then you can create and share custom
42:09 gpts so that's what the the event this
42:12 Saturday the gpts for good event is
42:14 going to take advantage of that so um if
42:17 you come Saturday you don't have to have
42:20 GPT um four cuz we we're going to have
42:23 an option we're going to teach you also
42:24 how to do Poots so po.com also has Bots
42:28 and you don't need a subscription to
42:30 make those so even if you only have 3.5
42:33 in GPT you can still participate on
42:36 Saturday um but anyway so so yeah just
42:39 chat gp4 is just crazy in terms of of
42:42 what it's got in it all
42:45 right so how did you get a female voice
42:47 I asked chat
42:50 GPT the voice and it said it only had no
42:55 it's it's in the settings
42:57 hang
43:04 on yeah
43:09 so yeah see I I I had a sneaking
43:13 suspicion so I feel better than I felt
43:16 last night but I just had a sneaking
43:19 suspicion that the the congestion [ __ ]
43:22 was getting worse tonight
43:25 so so I might not last long anyway um in
43:31 chat GPT app you go to your name in the
43:35 lower left and then in
43:38 settings see right there under speech it
43:42 says voice if you click on voice here's
43:46 then your options for the
43:49 voices hi there I just want to share how
43:51 thrilled I am to work with you hey there
43:54 I've got a really great feeling about us
43:55 teaming up hey hey I'm ready to hit the
43:57 ground running hello I'm really excited
44:00 about teaming up with you and hey it's
44:01 great to meet you how's your day going
44:03 hello I'm hey it's great to meet you
44:06 how's your day going that's annoyingly
44:07 positive working on some cool stuff
44:09 together I'm here to be your your
44:11 positive thinking coach I'm really
44:14 excited about teaming up with you and
44:15 I'm all set to dive in that's kind of
44:17 good how can I life easier that's Sky I
44:19 like Sky I'm ready to hit the ground
44:21 running so if there's anything you'd
44:22 like me to focus on first just let me
44:25 know yeah it sounds like a game show
44:26 host hey there I've got a really great
44:28 feeling about us teaming up how can I
44:30 jump in and help hi there I just want to
44:33 share how thrilled I am to work with you
44:34 hello I'm going excited about
44:37 teaming why they only have five I don't
44:39 know and why you can't put your own well
44:42 I know why you can't put your own in
44:43 there because they're afraid of people
44:47 using it for nefarious reasons um I also
44:51 noticed um 11 Labs so 11 Labs you can
44:56 close your own
44:58 voice but how did
45:03 they no it wasn't 11 was it 11 Labs no
45:06 it wasn't 11 Labs it was hey
45:09 Jen hey Jen to make an avatar of myself
45:13 I had to confirm my identity they they
45:16 wouldn't let me make a video clone of
45:20 someone else which was interesting can
45:22 you use personalized voice like Morgan
45:24 Freeman voice not right now
45:28 Raphael there are other services that
45:31 that do that but not chat jbt chat jbt
45:34 right now they're kind of limiting that
45:36 [ __ ] because they just they they don't
45:37 want people abusing it they cut a deal
45:40 with Spotify where like they they cut a
45:43 deal with like Lex fredman and a couple
45:46 of the big
45:47 podcasters where they can um do their
45:50 voices in different languages that's the
45:52 other thing about um chat GPT voice
45:59 is it can do different languages
46:06 so hey I'd like you to act as a Hindi
46:10 translator for me I'm going to speak in
46:11 English and then I want you to translate
46:13 immediately into Hindi without repeating
46:15 the English
46:20 okay sure I'm ready to help with Hindi
46:22 translation just speak in English and
46:25 I'll Translate into Hindi for you go
46:28 ahead with your first
46:30 sentence I hit the gold button that said
46:33 you can make money with chat GPT and I
46:35 haven't made any money yet so I'm
46:37 wondering if the button is
46:55 broken
47:01 you can make Mone
47:09 with you make
47:12 [Applause]
47:21 M don't [ __ ] with the robots they're
47:23 gonna remember us if we [ __ ] with them
47:25 too much
47:29 oh
47:33 man
47:35 oh I wish I didn't feel like poop it's
47:38 not fun feeling like poop you know
47:41 people you know what I'm
47:44 saying um do the special inst
47:46 instructions affect the way she talks to
47:48 you yeah I would think so
47:53 um hang
47:55 on
47:58 let me go add in some custom
47:59 instructions
48:01 here custom instructions off
48:05 continue
48:08 um you're an impatient professional who
48:10 doesn't have time for nonsense no I want
48:12 to get rid of that one I want to I'm
48:14 going to
48:16 go
48:18 um okay I got one I'm gonna I'm gonna
48:21 I'm GNA say you
48:24 are overly
48:27 dramatic and
48:30 only um and and uh
48:36 interpret everything in the
48:40 context
48:42 of
48:45 dramatic
48:48 conflict so that's who she
48:52 is and then how I'd like her to respond
49:02 is
49:06 um first
49:12 reply
49:15 um as
49:17 if what was
49:21 asked is a total
49:25 burden and then only speak in rhyming I
49:32 Amic
49:36 pentameter
49:38 save all right and custom instructions
49:41 is on yes okay so we're going to do a
49:44 new
49:46 chat and we're going to put her so okay
49:49 so she should respond as she's slightly
49:51 annoyed and then everything should
49:55 rhyme
49:58 I've been curious about photosynthesis
50:00 and how it
50:06 works oh the weight of this task so
50:08 heavy and vast to explain Nature's right
50:11 a photosynthetic cast in leaves green
50:13 and a taale does unfold where sunlight
50:16 and water in chlorophylls hold transform
50:19 with such Grace in a dance so bold into
50:21 sugar So Sweet Life's story is told
50:24 carbon dioxide from Air unseen seed
50:26 joins this merry band in pure Harmony
50:29 Oxygen's released a gift to you and me
50:31 from this wondrous process in a Leaf's
50:33 small Lee thus in rhythms of day under
50:37 Sun's soft gaze plants labor in silence
50:39 in their green hued ways in their veins
50:41 runs the Sun a life-giving blaze
50:44 photosynthesis indeed deserves all our
50:46 praise you seem a little over dramatic
50:49 can I have you answer that without
50:51 rhyming everything and sounding so
50:54 bothered
50:58 sure I can explain photosynthesis in a
51:00 straightforward way photosynthesis is a
51:03 process used by oh she I see I got to
51:06 add another rule in there that she can't
51:08 you can't just ask her to not act like
51:10 that S I was dead set against AI saw
51:13 your live and now love it wow that's
51:15 great total Noob but wanted to say
51:17 thanks s thank sorry about
51:20 this thank you so much that's the whole
51:23 point of the channel is
51:26 almost everyone that I know that is dead
51:29 set against AI hasn't used it the only
51:33 thing I want people to do is try it you
51:37 because it it's impossible to
51:43 explain it's it's impossible to explain
51:46 what it is in my opinion you you can't
51:50 tell people oh it's like Siri if Siri
51:53 didn't suck and now she'll probably come
51:55 on and
51:59 suck and and it like that doesn't
52:01 capture it and it and
52:04 it's but once you once you try it for
52:07 yourself and it's because here's the
52:09 thing
52:10 is AI is one of those things where if
52:13 you if you're looking over someone
52:14 else's shoulder you're kind of like oh
52:15 it's like a fancy Google search I get I
52:18 understand what it is but it's not until
52:19 you interact with some piece of
52:23 knowledge that you really
52:27 understand and you you know how hard it
52:29 is to write a screenplay or write a poem
52:32 or write a business plan or right like
52:34 whatever it is what the thing that you
52:39 know and like you're you're playing with
52:41 AI and then all of a sudden you're like
52:43 yeah but it can't do what I do right
52:45 we've all got that human hubris of but
52:48 but but we're special we're just okay
52:52 we're special and then you ask it to do
52:55 the thing that you're that makes you
52:56 special and it does it pretty [ __ ]
52:59 well and quite frankly it probably does
53:01 it better than you do it and it does it
53:02 a lot
53:05 faster that wakes you up like that's
53:08 like a shot of that that's like a shot
53:10 of cold water to the
53:12 face and then you go oh [ __ ] and then
53:16 and and then like from there then you're
53:17 like well wait what else can it do and
53:19 it's almost like this is the this is the
53:22 the the funny thing about generative AI
53:25 right now is is if you give it a shitty
53:28 question it'll give you a shitty answer
53:31 right if you just are like well whatever
53:33 tell me about photosynthesis and it'll
53:34 give you that bland sort of thing you're
53:36 like yeah well whatever I could have
53:38 answered that like that's every grade
53:39 schooler knows
53:41 that but then if you ask it to you know
53:45 explain the mathematical Concepts behind
53:47 the photosynthesis and you really dig
53:49 into the science of it and that's
53:50 something you know you realize oh [ __ ]
53:54 like it knows that too
53:56 um I took a picture of all my
53:58 ingredients and had it craft my meal
54:00 love it oh that's awesome that that's
54:02 one of those ones that it it kind of
54:04 falls into a parlor trick for me but you
54:06 know it's it's it's good oh you'll post
54:08 it to the salon yeah that would be great
54:11 not my job what's your job Earth
54:13 rocker oh you're a rock and roller yeah
54:16 there there's certain things like like
54:18 if you're if you're a a you know a live
54:22 musician yeah not for a bit like we've
54:24 we've probably got
54:26 10 years till the robotss get good
54:28 enough to play music soulfully they'll
54:31 play it competently in five years
54:33 physically but 10 years they'll get
54:37 good so so we've got a decade we got a
54:40 DEC I I I I absolutely think that
54:45 authentic experiences like live music
54:48 live art experiences Live Theater um
54:52 in-person meetings I feel like the value
54:54 of those is going to go up dramatically
54:56 with all this synthesized content
54:58 because it's just going to be there's
54:59 going to be so much content and so much
55:02 work and so much creativity and so much
55:05 that I think we're going to long for
55:07 human connection and so I think that
55:10 actually goes up ugly man isn't that
55:12 sharing not even a
55:14 list huh I don't know what that means
55:18 Amelio's
55:24 wife
55:27 what are you going to do when as ASI is
55:30 here play with it just like I'm doing
55:33 now ASI stands for artificial super
55:36 intelligence so artificial general
55:38 intelligence is when the the these tools
55:41 are better than humans at um most
55:46 valuable tasks and super intelligence is
55:50 where it's better than all of
55:52 us um I don't know
55:57 I it
55:59 it's I am I am already
56:03 personally interacting with technology I
56:06 couldn't imagine being possible in my
56:10 lifetime so for
56:12 me the distance between or or the the
56:16 demarcation between where we are now and
56:19 AGI and
56:21 Asi feels like semantic [ __ ] that
56:26 someone who studies this [ __ ] is going
56:28 to tell us we've achieved AGI it'll
56:30 probably be Sam
56:34 Alman and then we'll start playing with
56:36 it and we'll start to figure out what
56:38 that actually
56:40 means like one thing that seems really
56:43 clear to me is that this technology is
56:45 moving faster than we as humans are
56:47 prepared to deal with it I was
56:50 responding to ugly man who said he took
56:51 a photo of his
56:54 ingredients a list was not needed oh
56:57 that's cool that's very
57:00 cool
57:04 um what was I talking about oh AGI what
57:07 would be an easy way to create a
57:09 business using
57:11 AI this gets into you can make money
57:15 with
57:19 um an easy way to create a business with
57:22 AI is have ai tell you how to make the
57:24 business
57:26 as much as a smartass answer that sounds
57:28 like that's actually the
57:31 reality if if your question
57:34 is what's a good way to use AI to make a
57:37 business that's
57:39 sustainable that's a much harder
57:42 question
57:46 because any business that you can make
57:48 with AI right now someone else can as
57:50 well and someone else can
57:52 likely if it's a good idea their ability
57:56 to steal it
57:57 is that far away right there are now
58:01 tools where you can take a screenshot of
58:03 a website a sophisticated website and
58:06 put it in there and it will just start
58:07 coding that website and actually
58:09 generating the images in
58:11 it so they can steal the website they
58:13 can steal the idea they can steal the
58:16 the what you
58:19 do what they can't steal is who you are
58:23 and how you do what you do
58:29 so the question that I give to the or
58:33 the answer that I give to the question
58:34 of how do you make money with chat
58:37 GPT
58:41 is is not
58:44 swell cuz the swell answer the the the
58:46 answer why I make fun of these
58:50 people other than them being annoying as
58:53 [ __ ] is they've got the formula
58:56 completely backwards they're saying
58:59 start with the
59:01 technology take my course and you can
59:05 make money with chat
59:06 GPT do you know how they're making money
59:09 it's not with chat GPT it's with the
59:12 course now chat GPT likely wrote the
59:15 course don't get me
59:17 wrong but what they are praying on is
59:21 your
59:21 fomo that they know something that you
59:24 don't and here's the real ugly truth and
59:27 it's funny ugly man just came up um they
59:30 don't know more than you
59:32 do they know enough to write a course
59:34 with chbt Lord digital Gods thank you so
59:37 much for the Roses you a generous night
59:39 after night I really appreciate
59:41 that ugly man I'm prototyping a GPT that
59:44 looks at my local grocery store and
59:45 helps help helps me budget perfect
59:48 perfect use I this is the thing I'm most
59:50 excited about with um with gpts and and
59:54 what I'm calling like everyday
59:58 developers it is so easy to make
1:00:01 gpts that you don't have you don't have
1:00:04 to be a programmer you don't have to be
1:00:06 an engineer now if you're an engineer
1:00:08 you can do fancier [ __ ] with it but you
1:00:10 don't have to be
1:00:12 that and so what that means is you're
1:00:14 going to have people making
1:00:15 sophisticated software applications that
1:00:18 have never made a software application
1:00:20 in their
1:00:21 life and so they're going to do cool
1:00:23 [ __ ] with it and they're going to do
1:00:25 minor [ __ ] with it and and they're going
1:00:26 to do unique [ __ ] with it you're going
1:00:29 to go like oh I never even thought about
1:00:31 doing that it's going to be really
1:00:34 inspir inspiring is there an a AI that
1:00:39 uses correct facts all the time all AIS
1:00:43 hallucinate that is the feature not the
1:00:45 bug
1:00:48 um Andre
1:00:50 karpathy just had a really eloquent he's
1:00:53 the chief scientist over at open AI or
1:00:57 now he's heading up uh super alignment
1:00:59 open AI but he just talked about this um
1:01:03 all AI hallucinate so they all might get
1:01:05 it wrong but um use perplexity doai if
1:01:10 you want to get the facts right if it's
1:01:12 about facts and and research perplexity
1:01:15 doai is really good at it cuz it cuz it
1:01:18 it gives you an answer in the context of
1:01:20 something it searched on but it gives
1:01:22 you the citation and and the link to
1:01:24 that thing so you can double check all
1:01:25 your work with all of this stuff if you
1:01:28 need it to be accurate it's on you to
1:01:31 double check the work there's a pretty
1:01:34 famous like three months
1:01:36 ago this dumb [ __ ] lawyer had chat GPT
1:01:41 write a legal brief and he handed it to
1:01:44 a [ __ ] judge without proofreading it
1:01:47 and the judge was like uh Mr McGill cuty
1:01:50 or whatever the [ __ ] his name was it
1:01:52 appears that all of your cited cases
1:01:55 don't exist care to
1:02:00 elaborate that's not chat gpt's
1:02:03 fault that's that lawyer's fault for not
1:02:06 understanding the tool he was
1:02:10 using can't fix
1:02:14 stupid
1:02:16 um again one of the
1:02:20 things that I the the closest thing I do
1:02:24 to preaching
1:02:25 is say play with this stuff play with
1:02:27 this stuff play with this stuff why do I
1:02:29 say that because in playing with it you
1:02:32 find out its boundaries and you do it in
1:02:34 a lowrisk
1:02:35 way so so you say oh I I always wanted
1:02:39 to learn about micr crocheting which
1:02:42 might not even be a thing and then and
1:02:45 then it'll give you some [ __ ] and you
1:02:47 realize oh it made up a whole bunch of
1:02:49 facts about micro crocheting that
1:02:50 weren't there but they're kind of fun
1:02:52 but you learn that and that'll seem
1:02:53 frivolous at the time
1:02:56 and then someone will ask you to do
1:02:57 something and you're like oo I got to be
1:02:59 careful because I know that this thing
1:03:00 can make [ __ ] up but you wouldn't have
1:03:02 known that had you not played so play
1:03:04 play play not all of them hallucinate
1:03:07 you can have ones that refuse to do
1:03:09 things when they don't know yeah that's
1:03:10 that's a good point Pate uh um Eric
1:03:13 Roland who's who's one of the salon
1:03:16 members that you probably know Pate um
1:03:18 he's working on systems right now for
1:03:21 for a big Bank um where they've got a
1:03:24 known set of data they've they've got
1:03:26 like a database of question and answer
1:03:28 sets that have been approved cuz you
1:03:29 know banking is highly
1:03:32 regulated so they got a set of question
1:03:34 and answer sets um that have been
1:03:36 approved and and that he's got a A
1:03:39 system that uses AI to do the general
1:03:42 conversation and then the first thing it
1:03:45 does is check you know did this person
1:03:47 ask a question that's in in the set and
1:03:49 if it is it gives them an approved
1:03:52 answer and then if not it's got all
1:03:54 sorts of spe specific rules about what
1:03:55 it can answer and what it can't so that
1:03:57 that's a good point but I would just say
1:03:58 in general the tools right now you
1:04:00 should assume that they that they're
1:04:03 full of [ __ ] and then if they prove
1:04:06 themselves otherwise
1:04:09 great
1:04:11 um like Wiki in the early auts yep he is
1:04:15 now an ex-lawyer or he should be also
1:04:18 remember that AI is much bigger than
1:04:20 just llms yeah that's a good point
1:04:23 pate
1:04:25 pate's job Pate is in here to keep me
1:04:27 honest and to keep me from getting too
1:04:29 flowery on on uh how magical these
1:04:31 things are he he hates it when I say
1:04:34 they're ma what they do is
1:04:37 Magic it's one of the things that makes
1:04:39 me happy is when Pate gets cranky about
1:04:41 me talking about these things being
1:04:43 magic Kyle does it still hallucinate if
1:04:46 you give it the PDF of the actual Source
1:04:49 it can yes so you need to be careful
1:04:51 with that um
1:04:56 what a lot of this comes down to is
1:04:59 clever prompting so if you just give it
1:05:02 a PDF and say Here's your knowledge and
1:05:05 then you ask it questions it'll
1:05:07 sometimes use that PDF and sometimes not
1:05:09 if you say here's the
1:05:11 PDF I only want you to use the knowledge
1:05:15 in this PDF to answer anything anything
1:05:18 any questions outside of that I don't
1:05:19 want you to answer at all like you can
1:05:22 you can with clever prompting run it
1:05:24 down rabbit holes but I would still test
1:05:26 the [ __ ] out of it and you should
1:05:30 assume that it will
1:05:34 um disappoint
1:05:37 you for
1:05:41 now everything that you build
1:05:44 today
1:05:46 okay where do I
1:05:49 begin we're now we're now just over a
1:05:52 year into the inflection point that chat
1:05:55 GPT kicked off so generative AI chat GPT
1:05:59 sort of consumer
1:06:01 grade powerful AI models available to
1:06:05 regular human beings did not exist
1:06:08 before November 30th
1:06:11 2022 so you could easily argue that
1:06:14 while we're decades into machine
1:06:16 learning and AI development and research
1:06:19 we're one year into whatever the [ __ ]
1:06:22 chat GPT is whatever the [ __ ] Gemini
1:06:24 Ultra is going to be whatever GPT 5 is
1:06:27 going to be we don't know what these
1:06:28 things are yet so everything that you're
1:06:31 building right now with these these
1:06:34 tools
1:06:35 that produce remarkable
1:06:42 results they are just janky pieces of
1:06:45 [ __ ] this is this something I talk about
1:06:47 in here all the time you should assume
1:06:49 that the stuff you build right now
1:06:52 because it's built on top of janky
1:06:54 pieces of [ __ ] is also going to be a
1:06:56 janky piece of [ __ ] and will be obsolete
1:06:59 3 months from now you should just assume
1:07:03 that and if you end up building
1:07:05 something that's really good and not
1:07:06 janky and is predictable and is
1:07:08 consistent and things like that great
1:07:09 you you are in the
1:07:12 minority but these tools are going to
1:07:14 get better and better and better and
1:07:16 better and they're going to get better
1:07:17 fast it's not like they're going to get
1:07:19 better on a slow linear incline it's
1:07:22 like they're going to be janky Jank
1:07:24 janky janky they're a little less janky
1:07:26 oh my God they're better than anything
1:07:28 I've ever seen cat gbt was so helpful
1:07:31 with my system
1:07:33 administration Administration work as I
1:07:36 strove to migrate something from
1:07:37 something yeah that's it's it's really
1:07:40 good like if there's a big knowledge
1:07:42 base for it to know where there was a
1:07:44 lot of data ingested in the model
1:07:46 building these things are remarkably
1:07:48 good and it and systems management is
1:07:50 one of those um as I strove to migrate
1:07:53 and upgrade a server yeah yeah exactly
1:07:56 yeah those they it it's amazing at [ __ ]
1:07:58 like that it's also good at like um it
1:08:03 just knows how to fix [ __ ] like I had Pi
1:08:05 help me troubleshoot a dying battery in
1:08:08 my car and she got it right can you
1:08:11 limit where AI gets its Knowledge from
1:08:14 so you know it's accurate or will it
1:08:17 still do its
1:08:20 own
1:08:23 well you could make your own large
1:08:25 language model the base the Frontier
1:08:28 Model if you have $10 million $5 million
1:08:32 you can go do that if you don't have $5
1:08:35 million sitting
1:08:37 around you can fine-tune these models on
1:08:40 your own
1:08:41 data you have to do you have to do some
1:08:44 pretty
1:08:45 sophisticated logic and error checking
1:08:48 to get it to stay on the rails basically
1:08:51 the the answer is sort of yes it's sort
1:08:54 sort of yes with good
1:08:56 programming
1:08:58 um but no the the P the p in GPT is
1:09:03 pre-trained what that means is it's a
1:09:06 generative it's going to generate
1:09:07 content pre-trained Transformer the t is
1:09:10 Transformer that's the mechanism that
1:09:13 it's been trained on the pre-trained is
1:09:15 we open AI or we meta or we
1:09:19 Google have gone out and found all this
1:09:22 data and we've pre-trained this model
1:09:24 the reason it can have a conversation
1:09:26 with you is that it's been trained on
1:09:29 all of the internet or as much as they
1:09:30 could get their hands
1:09:33 on that Frontier Model so so there's
1:09:37 this thing called GPT 4 which is the
1:09:39 Frontier Model down here and then you
1:09:41 have chat GPT which is an application
1:09:44 that sits up here so chat GPT is the
1:09:47 thing that says oh I'm sorry Dave I'm
1:09:49 afraid I can't do that right that's out
1:09:50 of my content boundaries or um hey it's
1:09:53 really nice to talk to you like the fact
1:09:55 that it's Pleasant and generally gives
1:09:57 you what you want and won't do really
1:09:59 nasty things that's at the application
1:10:01 layer the the underlying Frontier
1:10:05 Model it would behave very very
1:10:08 differently without this application
1:10:10 sitting on the top of it um it would
1:10:12 cuss at you it would speak at you in
1:10:14 foreign languages and in tongues and um
1:10:17 let you do all sorts of evil nasty
1:10:19 things and harm other people and it
1:10:21 doesn't care if it doesn't have some
1:10:23 layer on top of it um but no the the
1:10:27 power of all of the stuff that we're
1:10:29 using is in the fact that it knows
1:10:31 everything that's one of the reasons
1:10:34 that Sam Alman said I don't think the
1:10:38 underlying model should be open- sourced
1:10:40 because it's too
1:10:42 dangerous and that's that's where Elon
1:10:44 Musk and him have just philosophical
1:10:47 diff differences elon's like [ __ ] it put
1:10:49 it out there that's meta said [ __ ] it
1:10:52 put it out there you can go get llama
1:10:53 right now the Llama 2 model with
1:10:57 without um safety guard rails on it and
1:10:59 have it do all sorts of crazy [ __ ] you
1:11:02 have to roll your own you have to you
1:11:03 have to know how to play with open
1:11:04 source stuff but you can go do that
1:11:06 right now that's why Sam Alman you know
1:11:10 sort of defied the name of the company
1:11:12 open AI was supposed to stand for open
1:11:16 source
1:11:19 AI That's why it was founded as a
1:11:23 nonprofit
1:11:26 Tam Tam Z I'm late dang it now you're
1:11:29 right on
1:11:30 time we're just rambling tonight I got a
1:11:33 cold I'm my my
1:11:35 brain's my brain's actually fine tonight
1:11:37 it's my I can just barely
1:11:41 talk repeat repeat I don't know what I
1:11:45 said um I will repeat it if you tell me
1:11:48 what I lost the stream uh
1:11:51 oh oh NOS
1:11:55 did I switch cameras it usually goes
1:11:57 away when I switch
1:12:02 cameras speak in tongues it will hey AI
1:12:06 learning letters
1:12:08 anema empathetic
1:12:12 Animalia what's happening like Frontier
1:12:14 Airlines let's see
1:12:17 they they stand their ground until you
1:12:20 correct them then they kind of reset oh
1:12:23 I apologize do you think some people
1:12:24 might use AI to hurt
1:12:27 Humanity yes well I think yes it and
1:12:34 it's so let's be clear about some
1:12:37 things ai's been around for decades and
1:12:41 big Power structures have been using AI
1:12:45 for nefarious reasons for
1:12:48 decades um they've got all of our data
1:12:52 they know all of our [ __ ] right they've
1:12:54 been
1:12:54 doing large scale
1:12:57 personalized you know surveillance [ __ ]
1:12:59 and marketing [ __ ] they like all that
1:13:01 [ __ ] for
1:13:02 decades what's different with generative
1:13:06 AI is that it puts this power in way
1:13:10 more hands right there there's a certain
1:13:14 self
1:13:18 regulating
1:13:20 benefit to only the powerful only the
1:13:24 powerful having access to powerful AI
1:13:28 right the militaries and countries and
1:13:31 big corporations because all of those
1:13:33 entities have vested interests in
1:13:36 protecting their entity right so so
1:13:40 they're going to they're going to use AI
1:13:43 in their interest but they're going to
1:13:44 try to prevent others from [ __ ] with
1:13:46 their interest right so there's kind of
1:13:48 this massive perverted self- policing
1:13:52 goes on when when only a handful
1:13:55 relative handful of people have access
1:13:57 to these powerful
1:13:58 systems what generative AI does is it
1:14:01 kind of opens Pandora's Box to say hey
1:14:04 we're going to democratize access to
1:14:06 these tools and so it's less like can
1:14:10 people do evil [ __ ] with AI of course
1:14:12 they can they and they've been doing it
1:14:14 for years
1:14:17 decades what changes now is
1:14:22 that there's an formality to the or or
1:14:26 there's a there there's a
1:14:27 democratization to access to that kind
1:14:29 of power so you have people with very
1:14:31 very different motives that now have
1:14:33 access to these tools now the good news
1:14:36 is that not only do the black hat
1:14:38 hackers have access to these tools so do
1:14:40 the white hat hackers right so that
1:14:43 battle between people trying to do bad
1:14:45 and and good guys trying to stop them
1:14:48 that's been there since computers right
1:14:51 like
1:14:52 that's that's it's just the game right
1:14:55 hacker going to hack you know white
1:14:57 hacker going to try to prevent them and
1:15:00 it's a whole [ __ ]
1:15:02 society it's just that that world just
1:15:05 got expanded to millions of people
1:15:08 rather than probably thousands or tens
1:15:10 of thousands right of relatively
1:15:13 powerful hackers because the tools are
1:15:15 are now so much more
1:15:17 powerful so it's going to be interesting
1:15:19 to see when you have like little rogue
1:15:21 agents running around with these AIS
1:15:23 what they can do we didn't lose you Kyle
1:15:26 it wasn't you this time
1:15:28 finally um how much does AI Control
1:15:31 machines that can build things if
1:15:34 any um I don't know if you seen the you
1:15:37 see the video of the the Tesla robots
1:15:39 building Tesla
1:15:43 robots we're not that far from
1:15:45 that we're not that far from that at
1:15:49 all like handful of years we are Star
1:15:53 Wars we're getting
1:15:54 close I had this combo with a friend
1:15:56 today yes the evil will try we will
1:15:58 police it same as ever yeah yeah exactly
1:16:00 Earth rocker it's like it's like the
1:16:03 difference is so so here's the deal in
1:16:07 today's world set set AI aside for a
1:16:10 second in today's
1:16:12 world if you want to like break into
1:16:15 government computer systems right you
1:16:17 have to be like this level of
1:16:20 hacker and if if you have to be this
1:16:22 level of hacker like the amount of
1:16:24 people that are that level of hacker is
1:16:26 like really small it's like you know
1:16:28 there's like a hundred of them and they
1:16:29 all know each other right it's like it's
1:16:31 a whole club they all know they can do
1:16:33 it there's some little code of ethics or
1:16:35 not in there it doesn't [ __ ] matter
1:16:37 but you have to be this good well AI
1:16:40 tools are like here and now they're here
1:16:43 and now they're here and now they're
1:16:45 here like there's going to be a point at
1:16:46 which AI tools are like here good and
1:16:50 who has access to being that good
1:16:52 hundreds of thousands millions of people
1:16:56 so if you have a hundred people that are
1:16:57 this good and millions of people that
1:16:59 are this
1:17:00 good it doesn't take the just do the
1:17:04 math someone's going to [ __ ] with
1:17:08 something the white hat hackers can you
1:17:11 know find them more quickly and take
1:17:12 them down more quickly but that's that's
1:17:15 a thing that makes me more nervous than
1:17:19 the big entities like I just feel like
1:17:21 they've got their own incentives and
1:17:23 disin incentives to not [ __ ] with each
1:17:25 other can you talk about some other AI
1:17:28 tools like what are your top five tools
1:17:30 other than chpt sure I can absolutely do
1:17:33 that and I think so why don't
1:17:38 we let me see if I can make this a
1:17:40 little bigger so it's easier to
1:17:45 read
1:17:47 okay
1:17:51 um Okay so
1:17:55 Chachi PT in my opinion is still the
1:17:57 best it's it's like it's the biggest
1:17:59 it's the best that's that's the one you
1:18:01 want to
1:18:02 use
1:18:05 um these four I would say are kind of
1:18:08 the next four large language models that
1:18:11 are worth you understanding the
1:18:13 strengths
1:18:16 of Claude in particular is a bunch of
1:18:19 exop AI people that these are the guys
1:18:22 that built
1:18:24 gpt3 they've got a it's called the
1:18:26 company's called
1:18:27 anthropic they've got a really
1:18:29 interesting safety model called the
1:18:31 Constitutional learning model anthropic
1:18:34 mission is not to build a chat GPT
1:18:36 competitor it's to create safety
1:18:40 protocols to make these large language
1:18:42 models safe and keep them aligned with
1:18:44 human imperatives basically so they
1:18:47 build Claude as a way to test their
1:18:51 theories um and it's really good right
1:18:53 they understand this [ __ ] very very
1:18:55 deeply um Bard is now this is now Gemini
1:18:59 Pro sitting underneath Bard so if you
1:19:01 want to play with Gemini Pro which is
1:19:03 kind of their chat GPT 3.5 model it's
1:19:07 connected to the internet it codes
1:19:08 really well um this is good and it's
1:19:11 going to get way better when they
1:19:13 release uh Gemini Ultra although my
1:19:17 guess is it's not just going to it's not
1:19:18 going to be a bard.com or it will be
1:19:21 like you know bard.
1:19:24 Advanced right it'll it'll likely cost
1:19:27 you 20 bucks a month but right now this
1:19:28 is free um perplexity AI is really good
1:19:32 if you want to do research it does a lot
1:19:34 of citations um Microsoft co-pilot is is
1:19:37 essentially Microsoft's version of chat
1:19:39 GPT 4 it's free it does image generation
1:19:43 it now does song generation they cut a
1:19:45 deal with suo AI um To Do song
1:19:48 generation then you have specialty large
1:19:51 language models you've got Pi this is
1:19:53 this is one of my favorites and it's
1:19:54 funny a lot of the people a lot of the
1:19:57 Irregulars that come to this channel a
1:19:58 lot this is one of their favorites cuz
1:20:00 it's designed to be conversational and
1:20:02 it's getting a lot better pay attention
1:20:04 to these guys they raised $1.3 billion
1:20:08 they just bought 22,000 gpus their
1:20:11 Frontier their base model is going to
1:20:13 get really good and it's going to be
1:20:15 designed to be conversational and to be
1:20:18 an agent on your behalf so pay attention
1:20:20 to Pi that's a good one um grock is a
1:20:23 waste of time right now but it's built
1:20:25 into Twitter and so if you pay 16 bucks
1:20:28 a month you get access to it I think
1:20:29 right now it's a waste of time and money
1:20:32 but that's probably going to get better
1:20:33 it's it's only two months old so it's it
1:20:36 makes sense that it's a piece of crap
1:20:37 right now but but it it at least has a
1:20:40 sense of humor um which is fun um llama
1:20:44 2 um I've got labs. perplexity here
1:20:46 because you can play with it there but
1:20:47 llama 2 is actually the foundation of
1:20:51 essentially the the whole open open
1:20:53 source um AI world is kind of you know
1:20:59 leveraging llama 2 um as as as a
1:21:03 starting point and there's
1:21:06 just if you're geeky in any way there
1:21:08 there are massive amounts of projects
1:21:10 going on in AI that you should
1:21:12 absolutely go play with and then po.com
1:21:16 is from quara the CEO of quara Adam
1:21:21 D'Angelo is on he was the one surviving
1:21:24 board member that tried to OU Sam Alman
1:21:27 he survived so po.com is interesting
1:21:30 because of that relationship they also
1:21:32 have GPT like things you can build
1:21:36 called Bots you can build a bot at PO
1:21:38 for free you don't need a gp4 account
1:21:42 and you can also play with like I don't
1:21:43 know 10 or 12 different models here it's
1:21:46 it's it's it's an interesting one all
1:21:48 right so these are all chat gbt like
1:21:51 things but there are eight of them that
1:21:52 I think think are major enough you
1:21:55 should pay attention to then you get
1:21:56 into image oh by the way this
1:21:59 list is at the AI salon
1:22:02 so the AI
1:22:05 Salon is a group that I started um the
1:22:09 week after chat GPT came out so if you
1:22:12 go to the
1:22:13 salon. that'll take you to a link tree
1:22:16 the first link there will link you to
1:22:18 the mighty Network's site which this
1:22:21 list that I'm showing is on
1:22:24 um and so you can come join and you can
1:22:26 it's it's a really cool Community you
1:22:28 can introduce yourself you can learn
1:22:29 about the salon things like that um okay
1:22:33 image creation you got Dolly 3 this is
1:22:35 inside chat GPT and Microsoft co-pilot
1:22:39 you have mid Journey mid Journey V6 just
1:22:42 came out it's quite good they also just
1:22:44 created a
1:22:48 uh a web front end um so so you you
1:22:53 still have to primarily make your images
1:22:55 in Discord but that's changing we're in
1:22:58 the in in the middle of that rolling out
1:23:00 to the web um I think that's going to be
1:23:02 a big deal for them apparently there's
1:23:04 some updates to V6 version 6 coming out
1:23:08 in the next week or two um that are
1:23:10 going to make it better at text better
1:23:13 at prompt coherence basically being more
1:23:15 accurate to what you put in the prompt
1:23:17 and a couple of other things and then uh
1:23:20 also they're apparently working on video
1:23:22 so image to video um so pay attention to
1:23:25 Mid Journey these guys if if you want
1:23:27 high quality art mid journey is the top
1:23:30 of the pack right now they're a pain in
1:23:33 the ass cuz you got to do it on Discord
1:23:34 but they're improving quite quickly I I
1:23:37 feel like their lead was slipping pretty
1:23:39 significantly because of Del but they
1:23:42 they're getting their [ __ ] together and
1:23:44 I'd pay attention to them in the next
1:23:46 three months um Korea is interesting
1:23:49 Korea is really just a rapper app
1:23:51 sitting on top of a bunch of stable
1:23:52 diffus Fusion Technologies from what I
1:23:55 can tell but they're it's pretty slick
1:23:57 and it's real
1:23:58 time they call it thought to image so
1:24:01 you just like draw shapes in the
1:24:03 left-and pane and it renders the image
1:24:05 in the right hand pane based on your
1:24:07 prompt it's it's pretty remarkable
1:24:10 Leonardo also does that it's also a
1:24:12 stable diffusion web wrapper um this has
1:24:16 gotten remarkably capable in the past
1:24:18 month or so they've just added feature
1:24:20 after feature after feature then Firefly
1:24:24 from Adobe is has not been trained on
1:24:26 any copywritten images so if you're in a
1:24:29 corporation that's really good
1:24:34 there man the voice is
1:24:39 crunchy all right Runway ml video
1:24:42 creation so you've got Runway ML and
1:24:46 pabs are kind of the two main players of
1:24:48 text to video or image to video both of
1:24:51 these guys are getting better and better
1:24:53 and better they're they're we're very
1:24:55 early and the the videos are janky as
1:24:59 [ __ ] but enjoy this time where the
1:25:02 videos
1:25:03 suck because they're going to get better
1:25:06 fast like over the next year both of
1:25:10 these Labs both of these
1:25:13 companies um are likely going to be able
1:25:17 to produce video content as as well as
1:25:21 stability AI
1:25:23 you know they're going to put out models
1:25:25 that are just going to get better and
1:25:26 better and better if if um Mid Journey
1:25:30 gets better like just so
1:25:32 anyway this is a really interesting area
1:25:34 to
1:25:35 explore um just because it's so shitty
1:25:38 right now that real filmmakers are going
1:25:40 to look at this stuff and laugh at it to
1:25:43 their own
1:25:45 detriment um Avatar creation he Jen and
1:25:48 did both of them are doing really
1:25:50 interesting
1:25:51 [ __ ] um if you just need to put together
1:25:54 a corporate video that like stitches
1:25:56 together stock video Visa does it really
1:25:59 really well it narrates it it writes the
1:26:02 script narrates it finds you all the
1:26:04 clips stitches them together and then
1:26:06 you can pay them to remove all the
1:26:08 watermarks it's it's it's a pretty slick
1:26:10 piece of stuff music and audio you've
1:26:12 got 11 Labs um for voice synthesis you
1:26:16 have sunno ai and refusion for um
1:26:20 generated
1:26:21 songs and then uh sunno that this this
1:26:24 is the one that got incorporated into
1:26:26 Microsoft co-pilot and then you have
1:26:28 sound draw which is kind of a loop
1:26:30 composer kind of like garage band but it
1:26:32 does it really really quickly using AI
1:26:35 uh then you got 3D objects and worlds
1:26:37 you can go play with those I think it's
1:26:39 a little early for any of those right
1:26:41 now but you can go play with them and
1:26:42 they're expensive uh and then you've got
1:26:45 um people I think you should follow
1:26:46 Ethan mullik one useful thing. org
1:26:49 [ __ ] brilliant David Shapiro Beyond
1:26:52 brilliant
1:26:53 Paul rater and Rachel Woods on the
1:26:54 business side of
1:26:56 things both of them are good all right
1:27:00 so that's that list so that's the [ __ ] I
1:27:02 think you should go play
1:27:04 with ask a ask a simple
1:27:09 question there's not a good way right
1:27:12 now hang
1:27:18 on there's not a good way right now to
1:27:21 keep up with this [ __ ]
1:27:25 it's essentially
1:27:27 impossible to just keep up with open
1:27:33 Ai and then what Google's doing Google's
1:27:37 got where is that AI get some rest Kyle
1:27:40 yeah I know I got to get some rest um I
1:27:42 will I I'll bug on out of here what was
1:27:45 the question about the
1:27:49 salon where's the the last AI salon
1:27:53 picture which one this
1:27:56 one that is
1:27:59 in if you go to the
1:28:03 salon and go to welcome to the
1:28:06 salon that's where that picture
1:28:17 lives is that is that the question
1:28:19 you're
1:28:21 asking
1:28:23 you should wrap it up and protect your
1:28:24 voice for another day yep Kyle might
1:28:26 have to use hey Jen to do the talking
1:28:28 actually it's not a bad
1:28:31 idea yeah when I when I do that big tool
1:28:34 thing it's a lot of talking so I I just
1:28:36 ground the voice to a halt
1:28:40 there so I'll get out of here but let me
1:28:42 just look at let me just look at some
1:28:43 comments and then I'll then I'll scooch
1:28:45 on out of
1:28:49 here no the winter
1:28:53 the winter oh this image this is in show
1:28:58 and tell if you go into show and tell
1:29:00 it's from Danielle she did like three of
1:29:08 them so if you go into Show and Tell
1:29:12 let's see what people have been up to in
1:29:13 here happy New Year nice quick take
1:29:16 Jake oh that's cool
1:29:21 nice very
1:29:23 cool don't tell husband AI is My Little
1:29:27 Secret Ted Ted's the
1:29:31 best that's a cool one who did that
1:29:33 quick take Jake that's
1:29:35 beautiful nice wow dude you're killing
1:29:39 it these are awesome Machine Gun
1:29:44 Kelly damn that's
1:29:51 good now the time for Jello salads we
1:29:55 got to bring back Jello-O salads people
1:29:57 I think Ted's right here and Barbara's
1:29:59 doing these uh someone get Kyle a
1:30:04 tissue that's
1:30:07 awesome anyway somewhere in here um is
1:30:10 Danielle's oh these eyeball things are
1:30:12 really good too who's this oh papa JJ
1:30:15 yeah yeah he's done some some really
1:30:18 cool Claire her work lately is Just
1:30:22 Killer too look at
1:30:24 that so this is the show and tell
1:30:26 channel in the AI salon so somewhere in
1:30:29 here is the uh is that
1:30:31 image it's from Danielle if you search
1:30:34 for Danielle you'll find
1:30:36 it do you think I I will decrease the
1:30:39 number of software engineer jobs
1:30:42 long-term
1:30:47 well it's going to dramatically change
1:30:51 them Joker Kyle keep going till you pass
1:30:54 the [ __ ] out Joker knows
1:30:56 [Laughter]
1:31:01 me all right I do got to get out of here
1:31:05 I got to go deal with
1:31:06 the deal with the
1:31:09 c is it all done with mid Journey no
1:31:11 most most of these things that you're
1:31:13 seeing in the AI Salon are done with
1:31:16 um with Dolly 3 inside chat
1:31:21 GPT okay uh Danielle just tagged me in
1:31:24 the images okay great so hang
1:31:27 on yeah that doesn't help them find it
1:31:29 though I think if they just I think if
1:31:30 they just look for you they can find
1:31:34 it
1:31:38 um got you thank you okay it's funny
1:31:41 that Claire and I had a similar concept
1:31:43 with the avocado kitchen color good
1:31:47 night your regulars good night Kyle feel
1:31:48 better see you tomorrow all right
1:31:49 everybody uh yeah I'm going to wrap it
1:31:51 up Mods thank you for modding Irregulars
1:31:54 thank you for being awesome subscribers
1:31:56 thank you for subscribing makes a huge
1:31:58 difference um everyone else
1:32:01 um get curious about AI play with AI
1:32:04 have your Kevin McAllister moment that
1:32:07 one if you haven't had it yet keep
1:32:09 playing Lord digital Gods thank you so
1:32:11 much for the Roses um I'll be back
1:32:15 tomorrow Saturday so tomorrow's Thursday
1:32:18 so I think what I'm going to do tomorrow
1:32:19 is I'm going to start building the GP s
1:32:22 I'm I'm going to start building the
1:32:23 prompts for the gpts that we're going to
1:32:26 build on Saturday for the 24-hour GPT
1:32:28 for gathon all right um so so check out
1:32:33 the uh just just be here for the be here
1:32:37 through Saturday it's going to be worth
1:32:39 it to be here all
1:32:41 right and uh all right who we got in
1:32:44 here jrc good to see you Kevin Vasquez
1:32:47 always great to see you Spike Jones
1:32:49 don't forget the lemon I will go do that
1:32:51 right now in in fact uh bitly so bit.
1:32:56 lgpt forg
1:33:00 good not underscore internal that's the
1:33:03 that's
1:33:05 the that's the spreadsheet isn't it an
1:33:07 we need the what's the just go to just
1:33:12 just go to uh the salon. the second link
1:33:16 talks about the 24hour event on
1:33:19 Saturday um
1:33:23 actions is what you want to learn Kyle I
1:33:25 know I you know what I haven't learned
1:33:28 actions well enough to go over it here
1:33:30 so I'll get my [ __ ] together in the new
1:33:32 year on that one I'm kind of waiting for
1:33:34 open AI to make that a little bit easier
1:33:37 they they just added one thing they
1:33:38 added in there Joker you might want to
1:33:40 check out is they added a GPT for
1:33:43 actions that'll teach you how to do them
1:33:45 you can just ask it to build you API
1:33:48 [ __ ] and it does it so so anyway um oh
1:33:51 okay okay go to bitly
1:33:55 bit.
1:33:56 lgpt forgood 2023 that takes you to the
1:34:00 website that explains what that event's
1:34:01 all
1:34:02 about all right everybody Ann Murphy I'm
1:34:06 excited to do this thing with you on
1:34:08 Saturday Blandon and honey oh that's it
1:34:11 I need Blandon and honey right now
1:34:13 against the norm you're awesome um
1:34:16 Silver Fox good to see you papa JJ
1:34:18 awesome as always Debbie did I miss you
1:34:23 we're we're here I you know me I'll be
1:34:25 back uh quick take Jake good to see you
1:34:27 Dr J frck Drew good to see you happy
1:34:31 down under Be not Afraid thank you
1:34:34 digital Gods Lord digital Gods to you
1:34:37 Silver
1:34:38 Fox Tobias good to see you sir all right
1:34:42 all right everybody peace out have a
1:34:43 good night I'm going to go deal with the
1:34:45 voice the snot the everything peace
1:34:48 bye