aiLL 8/31/23 - The Best of AI: Creativity, Ethics, Readiness

Description

Chapters

Transcript

0:08 foreign
0:28 welcome everybody Welcome to the AI
0:31 learning lab
0:33 welcome to the Irregulars look what I
0:35 made
0:39 I'm not weird I'm on a regular I hate
0:43 Tobias what's happening
0:45 I even went over to printful to set up a
0:48 where is it
0:51 um
0:56 Where'd I put that is that it yeah
1:00 there you go
1:03 look at that
1:07 [Laughter]
1:16 you see what I'm saying you see how this
1:19 works
1:22 yeah
1:23 the colors better
1:25 there you go that's more what it looks
1:26 like
1:29 hey Winston what's shaking
1:31 what's happening
1:33 welcome to the AI learning lab hey Nancy
1:36 what's shaking love the shirt
1:39 Tobias is just now headed to the hot tub
1:41 you guys are living the life
1:44 hahaha
1:47 I just had the electrical system of my
1:49 cargo Haywire
1:51 so I don't know what's going on with
1:52 that some guy tried to come and jump it
1:54 and it just it acted like it was
1:56 possessed so I don't know what's gonna
1:58 go down with that tomorrow luckily I
2:00 don't have anywhere important to be
2:03 but that's a pain in the booty yeah that
2:05 shirt came up with that graphic today
2:08 and uh
2:11 yeah well I'll get it figured out I'm
2:13 trying to figure out how to
2:15 hook it up to tick tock if I can't do
2:17 that I'll just I'll just make a link on
2:18 my link tree something like that
2:21 um seen it on the Discord topic yeah
2:23 that's where I put it
2:24 um welcome everybody if you don't know
2:26 what that is we have uh
2:29 we have some folks here
2:31 that fancy themselves Irregulars
2:33 [Laughter]
2:36 they show up here a lot and they name
2:38 themselves
2:39 so
2:41 so uh I made this with generative AI
2:45 look at that
2:47 and then I made
2:49 a bunch of stuff
2:51 so I made that there's a new thing if
2:53 you haven't seen it called ideola or
2:55 ideogram.ai that can do text
2:59 and I've just been messing around with
3:01 it and popped that little bad boy up
3:03 today and then I put it in Discord and
3:05 someone's like that's gotta be a shirt
3:08 so so then I went over to printful and
3:12 set up a shirt
3:14 so we'll see it's none of it's none of
3:16 it's done yet so and I gotta deal with
3:19 my car tomorrow so it might take me I
3:21 might get it done over the weekend but
3:22 anyway
3:23 how can I use AI to kill people awesome
3:25 that's fantastic
3:27 um go get yourself an open source model
3:30 and
3:32 do what you need to do I suppose
3:35 ah uh welcome so if you're new here my
3:39 name is Kyle Shannon
3:41 this AI learning lab
3:43 um
3:45 big part of what we do here is talk
3:47 about okay so all this AI stuff's coming
3:50 what the hell am I supposed to do about
3:52 it
3:53 let me know if I'm back give me the
3:55 hearts
3:56 what did you use to generate those
3:58 images I'll show you in a second
4:01 um
4:02 so the images bus bus oh bus is act is
4:06 asking about is that image and then I
4:10 made all these images too with with
4:12 letters in them and you're like but mid
4:14 Journey can't do that and stable
4:16 diffusion can't do that what is this
4:19 magic you speak of so I'll get to that
4:21 in a second
4:22 um
4:24 so if you're brand new to this AI stuff
4:26 um go here go to chat.openai.com that's
4:29 the official chat GPT website feel free
4:31 to screenshot these
4:33 um
4:34 Claude Bing and uh chat GPT are kind of
4:38 the three that I use in conjunction with
4:40 one another because they all do
4:42 something slightly con slightly
4:43 different Claude you can put longer
4:45 prompts in Bing's connected to the
4:47 Internet is just kind of the the best
4:50 right now
4:52 um prompts.chat that thing below is
4:56 there's an echo that might just be on
4:58 your side but maybe not I don't know let
5:00 me if there's an echo let me know other
5:02 folks
5:03 Turtle kid is 26 now I know isn't that
5:06 crazy
5:07 um prompts.chat will teach you how to
5:09 talk to these things couple other things
5:12 um some other things to play with
5:14 bard.google.com that's Google's entry
5:17 into this it's not bad it's getting
5:19 better and apparently Gemini is coming
5:21 and there's there was a an article in
5:24 I don't know some news rag that was
5:26 talking about the uh the AI the AI Bros
5:30 are fighting uh because the The Bard
5:32 guys are saying that
5:34 um Gemini that's coming is going to be
5:36 five times better than gpt4 and I think
5:39 Sam Altman said yeah that's marketing
5:42 and it probably is but if it's anywhere
5:44 in the neighborhood of gpt4 that would
5:46 be amazing it will likely be multimodal
5:49 so it's worth just you know start paying
5:51 attention to what's going on there
5:53 um perplexity.ai is good if you go to
5:55 labs.perplexity you can play with metas
5:57 llama 2 models and then Pi dot AI is
6:01 chatty GPT it'll talk a lot and it's
6:04 connected to the internet it's very
6:05 interesting oh thank you for the Llama
6:07 Winston
6:08 uh and then
6:10 um Poe is uh another one where you can
6:12 play with multiple models and you can
6:14 have uh
6:15 Bots and you can make your own bot it's
6:18 a good way to practice prompting now
6:19 pictures pitches so mid journey is
6:23 really good these things
6:27 were all made with ideogram
6:30 so ideogram.ai
6:33 that one right there
6:35 it's like a it's like a week old
6:38 it's like like they're they're Discord
6:41 still has that new car smell
6:43 and my car doesn't because it has an
6:45 electrical problem so it's probably got
6:46 that fried electronic smell
6:49 um Leonardo and clip drop are both based
6:51 on stable diffusion so all of these are
6:53 good to play with promptbook openart.ai
6:56 promptbook we'll teach you how to talk
6:58 to these things and and kind of what's
7:00 possible with them and then
7:02 futurepedia.io is uh a big old directory
7:06 of stuff now
7:09 now let's go play
7:11 so there's the there's the uh printful
7:15 printful little mock-up I did
7:19 um what's going on here why is my
7:22 thing not working all right come on come
7:25 on come on
7:31 I do not know what's happening hang on
7:34 but we got you know
7:37 different little
7:38 different little mock-ups
7:41 so those are cool anyway
7:44 um I will get that up and running over
7:45 the weekend
7:46 um let's go play with
7:48 um
7:50 with ideogram
7:52 so people can see what it is
7:57 it's actually very very straightforward
8:00 so you get in and you can just scroll
8:04 down and see what other people have done
8:05 welcome to September
8:09 um
8:10 open a can of whoop ass that's pretty
8:12 funny uh you can do all this Bill Gates
8:15 saying I heart Linux
8:21 so good all right there's another can of
8:24 whoop ass
8:28 crazy I was once crazy all right so you
8:31 can go remix these so I can go grab this
8:33 and
8:35 grab the uh
8:38 so poster illustration typography where
8:41 the uh
8:43 where the attributes so I can go and I
8:46 can either remix this one or I can just
8:48 go make my own so I can copy that prompt
8:51 and I'm going to just paste it in here
8:56 and then I'm gonna go
8:58 Maniac screaming
9:04 I'm not
9:06 crazy
9:12 I'm irregular
9:15 [Laughter]
9:21 oh it's gonna be good I don't know but I
9:24 didn't put in the uh hang on let me put
9:26 in the it was poster typography and
9:32 what's the other one illustration I
9:33 think
9:43 is that what that said I'm not crazy oh
9:45 I'm not weird
9:47 hang on
9:50 I'm not weird crazy works too
9:57 w-e-i-r-day
9:59 I'm a regular generate let's make some
10:02 new ones
10:03 in more llm training data drama
10:06 Atlantics investigative report on pirate
10:08 books was interesting yeah caperna that
10:10 that was really interesting so there's a
10:12 there's a
10:14 um a guy created a data source of like
10:17 170 000 pirated books and he made it
10:21 available to like the open source
10:23 Community to be able to train their
10:25 stuff on these books but they're pretty
10:27 certain that most of the major llms also
10:31 trained on it so
10:33 so let's see I'm I mean reedlar I'm not
10:37 word I'm irre read lar I'm not weird I
10:41 not weird I regular
10:46 oh those are funny yeah so anyway yeah
10:50 so this is about
10:52 about how accurate these things are
10:54 where it doesn't quite get them right
10:56 but it gets them right enough that with
10:58 with two or three remixes you can get it
11:00 and sometimes
11:02 um flipping the order where the where
11:05 the words show up and the sentence can
11:07 make it make a cool make a difference
11:09 I'm not weird so so this kind of lost
11:12 the I'm a regular thing so let me try
11:14 moving
11:16 the thing to the front
11:19 I'm not I've got a regular Maniac
11:22 screaming and we'll say um
11:25 uh uh zombie
11:28 turtle
11:30 screaming
11:38 all right let's see how this does
11:41 poster
11:43 typography
11:47 um
11:51 illustration let's get rid of this here
11:54 all right let's see how this does
11:56 generate
11:57 so now we don't have man screaming now
11:59 we've got
12:01 um turtle
12:02 zombie turtle
12:04 I like turtles
12:11 I'm not weird I'm irregular oh I've
12:13 still got a remix on let me get rid of
12:16 that thing
12:18 let me let them finish going
12:27 generate all right
12:30 and here we go what tool are we using we
12:33 are using
12:34 um
12:35 ideogram.ai Ido
12:38 yeah what did Dr J said
12:41 ideogram.ai
12:44 it's a week old it can do words
12:48 um
12:49 that one's that one's pretty cute I'm
12:51 not weird it didn't quite get irregular
12:53 but we can probably
12:55 um do that I'm not weird I'm let's see
13:01 so it takes you a bit of time to get
13:03 them get them going right but we'll
13:06 remix this one see if we can get it to
13:08 see if we can get it to get the words
13:09 right and basically that's what I did
13:11 with that other one with this one
13:13 I uh I just went until I got it and it's
13:16 it's still a little odd but I took it
13:19 into Photoshop they've got a super
13:20 upsize in Photoshop that does some
13:23 generative AI upscaling
13:25 and uh yeah what's the fee amount right
13:28 now the fee on ideogram is free
13:33 free free free free free free free free
13:35 free free free free free free free free
13:39 how long will it stay free I don't know
13:43 a lot of people are using it a lot of
13:45 people are using it I'm not weird I'm
13:47 your regular
13:49 oh there we go I'm not weird I'm
13:51 irregular we got one
13:54 and you can have oh it's I'm not worried
13:56 but you can obviously go in and
13:58 Photoshop you know Photoshop the the
14:00 text in
14:02 so you could take that little dude
14:04 [Laughter]
14:11 comedy the body Cube score
14:15 Vibe check serenity
14:20 yeah so anyway
14:22 um
14:23 you know
14:26 what's exciting about this is you know a
14:29 that we can do it but B
14:31 this isn't mid-journer stable diffusion
14:34 right this is like a some company came
14:36 out of nowhere and figured this crap out
14:38 and we were assuming it would be
14:40 mid-journey or stable diffusion or maybe
14:42 Google would come out with something or
14:43 maybe Adobe would come out with
14:45 something nope some company out of left
14:47 field so what this is going to do is
14:50 it's going to force the hand
14:52 of any of those other Technologies if
14:55 they're sitting on
14:56 this capability they're probably going
14:58 to push it out faster
15:00 so
15:02 um
15:03 button is grayed out on this site button
15:05 might be grayed out simply because it's
15:08 uh it's overloaded I I logged in today
15:10 and at some point it said you know we're
15:12 just busy right now give us give us a
15:14 bit who are they I have no idea I have
15:16 no idea who they are I don't know if
15:18 they're funded I don't know if it's a
15:20 couple of kids in a case of Red Bull
15:21 that
15:23 got together over a weekend and popped
15:25 together some models I have no zero
15:27 ideas
15:29 I actually learned about open uh idea
15:32 ideogram.ai from someone in here
15:36 I think it was the night or the night
15:38 after it launched they said what do you
15:40 think of ideogram and I'm like I don't
15:42 know nothing about it and I went and
15:43 looked at it and I was like holy crap
15:48 all right so if you have questions about
15:51 AI pop them in the comments below I'll
15:52 do my best to answer them
15:54 um I'm a little distracted because my
15:58 car's electrical system just blew up I'm
16:00 trying to figure out what do I do I work
16:02 tomorrow what do I do about the hey
16:08 whatever just stuff it's just stuff it's
16:11 all stuff
16:13 and nonsense that's what our lives are
16:16 all about
16:18 um oh but you know what I did I went to
16:20 um I went to pie so I went to the pie
16:23 app
16:24 right there
16:26 and I asked pie I said hey can you so I
16:29 went to my car
16:31 um you know the dash lights came on and
16:33 when I hit the ignition button they just
16:35 went black
16:36 and I heard a click and everything went
16:38 black and then
16:40 I talked to her about it and she goes
16:42 well let me ask some clarifying
16:43 questions you know did the engine
16:45 attempt to turn over and then the scream
16:47 went black or like she has
16:49 troubleshooting questions she actually
16:50 had a decent clue what was going on she
16:53 said it's either the fuse box or a dead
16:55 battery
16:56 um and I think I think it might end up I
16:58 think it's probably a bad relay in the
17:01 fuse box there was a whole section of
17:02 the fuse box that all the fuses were
17:04 dead when the guy was trying to jump it
17:05 so uh so I think that's going on
17:08 um but anyway that's so so that's what's
17:11 going on but anyway pop questions in the
17:13 comments below I will do
17:16 um
17:17 I'll answer questions Nancy okay another
17:20 video of yours was still playing on my
17:23 phone sorry about that I'm everywhere
17:25 you can't lose me
17:30 my plan my plan is working
17:35 I also think speaking of plans working I
17:38 think we should have the little total
17:40 I'm not weird I'm a regular that should
17:42 be in our little window tonight don't we
17:44 think in honor of the Irregulars
17:47 I think it's only fair
17:49 day day day day day
17:55 yeah that's good make it nice and big
17:59 I'm not weird I'm a regular
18:04 raising a bunch of money don't hear
18:07 about them much we're going to hear
18:09 about them they're gonna they're they're
18:11 in a bit of stealth mode right now but I
18:14 think that they're gonna
18:15 they're gonna pop on the scene so right
18:17 now I've got seven things on my list
18:20 notably absent from the list is cohere
18:24 Apple and x x dot x dot a i
18:27 jelly old elon's thing and then there's
18:31 probably some others and like you know
18:32 the fact that ideogram just kind of
18:34 showed up out of nowhere I have a
18:36 feeling we're going to see something in
18:37 the large language model space that that
18:39 pops up where someone's figured out
18:41 something so
18:44 Stevo hey Kyle hey Irregulars what's
18:48 happening okay farmer saying hello Echo
18:50 mode offset Blaster Echo mode off
18:54 Blaster set to 10. Blaster set to 11.
19:01 This Is Spinal Tap
19:03 uh
19:06 weird is more inclusive weird is just
19:09 the only way to be because you know what
19:11 every [ __ ] human being on Earth is
19:14 weird we're all weird we're all just
19:16 trying to figure it out no one's worse
19:19 than any well some people are worse than
19:20 other people but everyone's just trying
19:22 to
19:24 everyone's just trying to figure it out
19:27 and uh what tool is that that was
19:29 ideogram we talked about that the typos
19:31 work that's very Keel Kyle
19:35 hi Kylie Robert Rossi Rossi what's
19:38 happening
19:39 um
19:42 what tool are we using we got that
19:47 um what's the fee amount zero dollars
19:49 zero dollars and zero cents
19:52 buy one today get one free
20:00 do you ever remember the Saturday Night
20:02 Live skit First National change Bank
20:05 [Laughter]
20:09 you come in you give us a dollar we'll
20:12 give you 20 nickels
20:14 or a hundred pennies whatever you want
20:16 you got a five dollar bill we'll give
20:18 you a one four quarters
20:22 and then at the very end of the very end
20:25 of the commercial
20:27 you may wonder how we make our money
20:29 volume
20:30 [Laughter]
20:37 gen xers you remember that one don't you
20:39 first national change bank that must
20:41 have been late late 70s early 80s
20:44 something like that
20:46 um
20:47 why can't I generate on ideogram well
20:50 you you might have spelled it wrong bus
20:52 so it's idogram
20:55 idiogram not idea Graham
21:00 where's my
21:04 oh what what what what what
21:07 right there
21:09 i d e o graham.ai
21:13 love the grenade belly
21:16 oh yeah I know right
21:21 [Music]
21:23 oh on the wait it was on the other one
21:27 it was on wait for it
21:32 there grenade belly
21:36 all right
21:39 oh irregular button is grayed out
21:43 the [ __ ] Chronicles who are they
21:47 Kyle do you have a part-time job in Foot
21:49 Locker
21:51 solid troll Stevo thank you yeah I had a
21:55 I had a job of Foot Locker in the 80s
21:58 the stripes are a little faded
22:01 they may need to go change my shirt uh
22:04 been using gencraft I don't know what
22:06 gencraft is should I go check it out let
22:08 me go check it out
22:11 um
22:15 craft
22:20 what will you create
22:26 world's most powerfully iPhoto and video
22:28 are generation engine what will you
22:30 create
22:33 introducing video generation interesting
22:36 bring your ideas to life
22:39 yep
22:41 so that could be worth playing with
22:45 and it looks like they have an app
22:49 um you know I'm still I'm still in the
22:51 I'm still in the mid journey is the big
22:53 thing except ideogram
22:57 because it does text it's just a whole
22:59 different thing like I've
23:01 like if you just need something quick
23:03 for a presentation like a slide in a
23:04 presentation like I did a thing last
23:06 night for a bunch of first-time Founders
23:08 and I needed two different slides I just
23:10 hopped over to ideogram and instead of
23:13 doing like trying to find a stock photo
23:15 and then figuring out a headline I just
23:17 figured out the headline and told it
23:18 what I wanted and I just did it enough
23:20 time still I got something that worked
23:21 and popped it in there it didn't matter
23:23 like it was only going to be up for five
23:25 seconds so it didn't really [ __ ]
23:27 matter right I didn't need to do any
23:28 pixel peeping
23:30 so
23:32 um yeah these these tools are gonna get
23:34 these tools are gonna get quite powerful
23:36 quite quite powerful I mean they're
23:38 already pretty [ __ ] powerful uh even
23:41 to Troy drecker what's happening or
23:43 Decker sorry
23:44 [Laughter]
23:45 um
23:46 [Music]
23:51 find a mechanic on Labor Day weekend
23:52 actually I just I did roadside
23:55 assistance the guy was here in like 20
23:56 minutes and I'm like how the hell did
23:58 that happen
23:59 but then it was weird man his his
24:01 battery charger was going on and off and
24:04 there was clicking in the engine and
24:06 clicking in the under the dash if anyone
24:08 knows what's going on I think it must be
24:10 a dead relay or something because it was
24:12 a lot of that stuff uh that's the first
24:15 thing I do is ask pie pie is pretty cool
24:17 ask Pie about your car she may offer to
24:19 give you a ride yeah well she'll go
24:22 she'll go well before I give you a ride
24:24 let me just ask you one more clarifying
24:26 question
24:27 pie is definitely 20 questions
24:31 um how would you think AI takes over in
24:34 all Industries
24:37 [Applause]
24:39 um
24:43 I think the most
24:45 the more the a general answer that I can
24:49 give that sort of you can look at each
24:51 industry and figure out what it's going
24:52 to be for each of them is that
24:56 what generative AI is going to do first
25:00 is it's going to take all the repetitive
25:03 tactical execution work and it's going
25:05 to automate it
25:07 so just think about every industry so if
25:10 if you're talking about sales people a
25:13 big part of what sales people do after
25:15 they make the phone call is data entry
25:17 into Salesforce and if you've ever used
25:20 Salesforce it's an abject [ __ ]
25:22 nightmare
25:24 I'm sorry that was being too kind
25:28 um it is such a complicated nasty Haven
25:31 of
25:34 shitty user interface and too many
25:36 features added over too many decades but
25:39 you can't turn it off right because all
25:41 your shit's in there
25:43 and so so one of the first things that's
25:45 going to happen is complicated systems
25:47 like that and repetitive kind of
25:49 low-level tactical execution work that's
25:51 all going to get automated right so at
25:53 some point whether Salesforce does it or
25:55 someone else does it
25:56 salesperson does the call
26:00 the system will probably know what
26:02 number they called it'll probably record
26:04 the call transcribe the call create
26:06 notes for the call generate an account
26:09 in Salesforce if one doesn't already
26:11 exist generate a contact in Salesforce
26:13 based on the person that called if one
26:15 doesn't already exist
26:17 create a meeting
26:20 and put the notes for the meeting in and
26:23 then schedule a next meeting if the
26:25 salesman was any good right so that's
26:27 like one little teeny sliver think of
26:30 every single job that's computer
26:32 mediated in any way take the complicated
26:35 repetitive [ __ ] part of the job that
26:38 gets automated out
26:41 so that's the first thing that happens I
26:44 think well probably before that is just
26:46 personal productivity stuff employees
26:48 are figuring out ways to use these
26:50 things to write emails faster write a
26:53 blog post faster just just personal
26:55 productivity will happen then you'll get
26:57 into the automation stuff then you start
27:00 to get into all sorts of crazy [ __ ]
27:02 where
27:05 entire business models are going to be
27:07 changed
27:08 um
27:09 overnight so I think Consulting is in
27:12 for a rude awakening I think uh the
27:15 adding Industries in for a rude
27:17 awakening I think the creative arts are
27:18 in for a rude awakening they're in the
27:20 middle of striking right now to try to
27:22 prevent it
27:23 sorry I just noticed my hair is
27:26 some version of puffy
27:29 um
27:31 I'm really not well no I obviously am I
27:35 talk about my hair a lot
27:36 but but I never have although to be fair
27:40 to myself I've never gone live for three
27:42 hours a night
27:45 multiple nights seven days a week before
27:47 in my life so you know what are you
27:49 gonna do
27:51 um I don't know if that answers your
27:52 question bus but but it's it's uh it's
27:55 gonna be quite transformative I mean if
27:57 you just think about if you take the
27:59 repetitive shitty stuff and automate it
28:02 how many people
28:04 how many people's job is it to do that
28:07 repetitive you know low-level work a lot
28:10 so if that part of the job gets
28:12 automated out what do those people do I
28:14 don't know do they upskill and do other
28:16 things do they get laid off I don't know
28:20 um yes put that a regular turtle
28:22 everywhere Kyle I have chat GPT linked
28:25 to 11 labs to get voice
28:27 so how does The Voice show emotion
28:30 without
28:32 using punctuation
28:35 um
28:35 right now I don't think you can I don't
28:38 think 11 Labs has any kind of XML syntax
28:41 in their API but but they might so I
28:45 know within 11 Labs you can pick from a
28:47 pull down menu do you want angry whisper
28:49 happy sad
28:52 um
28:53 if in their API you can you can specify
28:59 emotion within text
29:02 then you could have chat GPT doing so
29:06 and this this is what I want to do with
29:08 11 Labs I just I don't think they do
29:10 this in their API yet but if they do
29:11 this is how you do it
29:13 you
29:15 um you generate text or you write text
29:17 for whatever you want 11 labs to say
29:21 and then in chat EPT you give it okay
29:23 here are the 11 Labs of motion and
29:26 here's how they structure their XML or
29:28 their Json whatever format they want it
29:30 in
29:30 and please analyze the text and figure
29:33 out what's the most appropriate emotion
29:36 and and basically tag
29:38 all of what the person's saying with the
29:41 XML or Json and then up left upload that
29:44 to 11 labs and then you'll get different
29:46 emotion within the thing there's two
29:48 things that that all of these voice
29:50 synthesizers are missing right now one
29:52 of them's emotion and and that I think
29:55 once you can tag it and have it do those
29:57 different things
29:59 um that's handled and then the other one
30:00 is Cadence right when do you slow down
30:03 when do you have a longer pause
30:05 when you talk fast you know and maybe
30:08 excited handles that
30:11 um maybe those emotional shifts handle
30:14 the Cadence
30:17 um
30:19 but yeah so so anyway that should I'm on
30:22 team vertical stripes thank you uh so
30:25 that that should help Robert uh check
30:27 out the the 11 Labs API if if they have
30:31 uh XML or Json that you can put in there
30:34 and send to the API
30:37 um then you could generate that with
30:39 Chachi PT
30:41 and you can make money with cha chippy
30:43 tea at 11 lab synthesizing voice and
30:46 talking to people as if they were human
30:48 just like this channel is he live he
30:51 doesn't seem to be answering my
30:53 questions I think he must be a recording
30:55 I'm sorry Dave I'm afraid I can't do
30:57 that
31:01 get hit
31:04 all right
31:05 you guys
31:08 you wait why why can't I get duet AI
31:12 um
31:13 I just because uh Google's rolling It
31:15 Out Slow you mean you mean the Google
31:18 adding it into Google Docs and [ __ ] like
31:21 that I just don't think they've launched
31:23 it yet
31:24 um it'll come it's it it'll come Kyle
31:27 workshops I know I know hey here's the
31:30 good news I I got caught up on I was
31:34 behind on AI Salon video uploads
31:37 um
31:38 IO Ronda and I are going to do a
31:40 workshop for her Channel and and so I
31:43 set up I've got that meeting tomorrow
31:45 um
31:48 I started thinking about the workshop
31:49 and I'm and how I can take the thing I
31:51 did for the big company and do so it's
31:53 it's in the works I I was actually
31:55 thinking about it today uh make me a mod
31:58 and I'll make an FAQ and list all the
32:00 tools du jour for folks okay great I'll
32:02 definitely do that corvino
32:04 um all right I'm gonna reach out to the
32:05 folks that have said they want to model
32:07 I'll reach out and we can set up
32:09 meetings that's a really good idea very
32:11 very good idea I like it I like it open
32:14 or closed Source this Tool uh is closed
32:18 source
32:19 they also don't have an API they're a
32:22 week old I asked them I'm like do you
32:25 have an API and they're like they're
32:28 like no
32:29 and they answered me like within 10
32:30 seconds I'm like oh my God you're brand
32:33 new
32:36 there's like 250 people in the Discord
32:38 when I joined it it was like holy [ __ ]
32:41 this thing just launched
32:43 they're gonna
32:45 if they haven't raised a pile of money
32:47 I guarantee you there are investors
32:51 sitting at the door sill of this company
32:57 I don't know anything about them so it's
32:59 possible that they're well funded and
33:00 they they have something good because
33:02 they have you know the money to do it
33:04 but it's also possible they figured some
33:06 [ __ ] out because they're clever and now
33:09 they're gonna go raise a ton of money
33:11 and they're gonna go after mid-journey
33:14 the image quality on this is kind of
33:16 stable diffusion 1.3 like it's not great
33:21 um and the typography is not great but
33:23 holy [ __ ] when it gets it right like
33:26 sometimes it's just awesome like you
33:28 know put famous person in the t-shirt
33:30 and and just put the text on the T-shirt
33:32 it does that really well
33:35 um is it context driven
33:38 um
33:39 well so what I'm suggesting Robert so
33:41 Robert's asking
33:43 for 11 Labs with that emotional stuff
33:45 he's asking is it context driven it's
33:47 not right now I think what 11 Labs does
33:49 right now is it basically just takes
33:50 their normal voice and that's what it
33:52 synthesizes
33:54 but if you can tag it with different
33:56 emotions within the thing you could use
33:58 chat GPT to determine the context
34:01 and say okay here's what the person
34:04 should whisper you know basically
34:07 tell chat GPT to act like an actor and
34:10 analyze the script and figure out which
34:12 emotions should be applied to which
34:14 sections of the script and it'll then it
34:17 will be context sensitive
34:19 that's that's how I plan to use that
34:21 that tool
34:22 um and and
34:23 quite frankly
34:25 um
34:26 the the reason I haven't really dug
34:29 dramatically into that is that
34:33 okay 11 Labs like
34:35 let's go let's make that happen and
34:37 maybe they have and I just didn't I just
34:39 missed it in the API
34:41 um but they'll make it happen at some
34:42 point
34:43 but these these synthesized voices
34:46 they're going to go from monotone to you
34:49 know having some kind of emotional
34:51 roller coaster in them um they're gonna
34:53 they're gonna get there sooner than
34:55 later and once that happens then because
34:57 they're already they're already good
34:59 enough that with certain voice types and
35:02 certain applications certain kinds of
35:04 video you can't really tell until like I
35:08 noticed one the other day I thought it
35:09 was just a guy talking and then it said
35:12 something like no no just
35:16 and you know it's supposed to be node.js
35:18 but it tried to pronounce the JS so it
35:20 said no
35:22 I'm like oh that's a synthesized voice
35:25 it's like so there's there's little
35:27 tells
35:29 um so true everyone is weird some others
35:31 somewhere than others of course and
35:33 that's why we collect here so far I've
35:35 only used chat GPT what's a better
35:37 option all right Heather so let me let
35:39 me let me give you my my two cents on
35:43 these tools
35:44 these tools they got they got some
35:47 action to them okay
35:53 ma okay chat GPT
35:57 this costs 20 bucks a month if you want
35:59 gpt4 but you also get what was formerly
36:02 known as code interpreter and it's now
36:04 known as advanced
36:06 data analysis
36:09 which is [ __ ] remarkable it's also
36:12 got plugins which are not [ __ ]
36:14 remarkable but some of them are useful
36:17 um and just gpt4 is really good and then
36:19 GPT 3.5 here is super fast so if you
36:22 want to just knock stuff out fast GPT
36:23 3.5 is really good
36:26 um also if you want to develop stuff
36:28 their API is really good their
36:30 playground's really good like they've
36:31 just they've been around for a while
36:32 they're the ones that kind of blew the
36:33 doors off this thing so so they they
36:35 kind of deserve the shakes right now
36:37 Bing is Microsoft's version of this
36:42 um tied to their search engine so Bing
36:44 is connected to the internet it's also
36:47 it's also a chatbot sitting on top of
36:50 the gpt4 model
36:52 it's not as good as gpt4 up here but
36:55 you're like but Kyle they're both gbt4
36:58 why wouldn't they be as good
37:00 because you have your your base large
37:02 language model down here and you have
37:03 your application layer sitting on top of
37:05 it so chat GPT is an application and
37:07 then Bing chat is an application they're
37:10 different
37:11 um tools and the fine tuning for how
37:13 they use that that base model is
37:15 different
37:16 uh Microsoft has made this very very
37:20 um what do you call it
37:21 um a lot of safety guard rails to the
37:24 point that it's really [ __ ] annoying
37:26 sometimes but it is connected to the
37:28 internet it can make images you can just
37:31 say make me an image of whatever it can
37:33 read images you can upload an image and
37:35 say write me 10 headlines for that image
37:38 um so it's it's worth having on the top
37:40 three Claude the reason it's on the top
37:43 three I like there's a lot of things
37:45 about Claude that I like it's from a
37:47 company called anthropic a bunch of
37:49 ex-open AI people they're using a thing
37:52 called the Constitutional learning model
37:53 which I think has a really good shot of
37:56 reducing bias and increasing alignment
37:58 basically let's make sure the machines
38:01 don't decide that we're in their way and
38:03 they're going to get rid of us and then
38:05 Claude also has a hundred thousand token
38:08 context window
38:10 on the other hand is a four thousand con
38:12 token context window what the hell does
38:14 that mean it means chat gbt after about
38:18 3 000 words like like if you put in
38:20 something longer than three thousand
38:22 words it just chokes out it can't do it
38:24 if you're having a conversation with
38:26 chat gbt and you go on for a while the
38:29 early part of the conversation it will
38:31 forget after three thousand words
38:33 so so it's got this kind of sliding 3000
38:36 word window where it'll it'll so if
38:38 you've ever noticed if you're working
38:39 with it for a while it's good it's great
38:41 it's awesome and then all of a sudden
38:42 it's just like it lost its mind it
38:44 literally lost its mind it forgot all
38:46 that Claude
38:48 that window is about 75 000 words so
38:51 like a a decent sized novels 50 000
38:54 words so you can literally copy and
38:56 paste a 50 000 word novel into Claude
38:59 and tell it to write an epilogue so
39:01 that's those three
39:03 um
39:04 Bard is also connected to the internet
39:07 so if Bing drives drives you crazy try
39:10 Bard
39:12 um
39:13 the model that Google's working on right
39:17 now that's going to replace the Palm 2
39:19 engine that's sitting underneath this is
39:21 called Gemini and it's supposed to be
39:24 remarkable according to google sources
39:27 uh but we'll see but this is not bad
39:29 right now uh and then lab stop
39:32 perplexity you can do research there you
39:34 can play with metas llama 2 models which
39:36 are open source thank you for the how to
39:39 use
39:40 um Pi dot AI is super chatty so I would
39:43 recommend installing the PI app so if
39:46 you go to the App Store search for pi
39:48 personal assistant
39:50 the company is inflection and uh put it
39:55 into talk mode in the lower right hand
39:57 corner you can sort of turn on the phone
39:59 icon and just talk to it it's really
40:01 good and then Poe you can make Bots you
40:03 can make your own Bots and practice your
40:05 prompting and use other people's Bots
40:07 and learn from them so those are the top
40:09 seven and and kind of what they do
40:11 different the different flavors it's a
40:14 lot of flavor a lot of flavor if you
40:16 know what I mean a lot of flavor a lot
40:18 of flea flavor yeah
40:24 you can save money with chechi pite
40:27 um
40:28 oh Tick Tock jump me to the bottom Tick
40:32 Tock jump into the bottom Tick Tock jump
40:35 me to the thanks Ted looks
40:38 to the bottom and I hate The Tick Tock
40:41 yeah yeah open a closed Source okay here
40:44 we are
40:45 there's while we were
40:48 um
40:49 I'm seeing
40:51 that it's like I got GPT angry and I
40:54 could tell by the voice but no
40:56 punctuation used
41:00 [Music]
41:02 I don't quite know what you mean Robert
41:04 Rossi
41:05 I'm saying that it's like I got chat I
41:08 got gbt angry
41:10 and I could tell by the voice but no
41:13 punctuation used
41:15 oh I see
41:17 um it might be let's listen Maybe 11
41:19 Labs has gotten their [ __ ] together and
41:20 they've added in they're they're
41:22 dynamically adding in that context I the
41:25 last time I used it they didn't it was
41:27 still very monotone
41:29 um but to be to be fair I haven't really
41:32 played so it's very possible they're
41:33 improving all these companies
41:36 you know they're getting better we're
41:38 you know it's still spitting duct tape
41:40 right now but uh we're getting less and
41:42 less spit and duct tape every day
41:46 um not in caps
41:48 all right the shirt thank you oh that
41:51 shirt or this shirt
41:53 that's a shirt
41:56 uh it freaked me out because I did not
41:59 use any punctuation at all oh
42:02 interesting
42:03 I don't know maybe did you go to 11 labs
42:07 to their interface or did you do it
42:08 through an API call
42:12 I mean I think I think it can probably
42:17 I don't think you need to put in
42:18 punctuation
42:20 um but if it but if it got angry at a
42:23 certain point maybe it's doing
42:24 contextual [ __ ]
42:26 difficult to pop up words in mid-journey
42:28 it's it's essentially impossible to do
42:30 words in mid-journey you can usually get
42:32 a letter or two but you can't get like
42:34 that was I don't know my sort of fourth
42:37 or fifth iteration
42:39 and it like the problem was it just kept
42:41 spelling irregular wrong because
42:43 irregular is a weird word and so it
42:46 probably doesn't have much in its in its
42:48 database but it sounded angry that's
42:50 interesting
42:52 um is this the holiday to buy a bed so
42:55 many choices this is a holiday I gotta
42:57 deal with my car uh after many attempts
43:00 I finally created a logo on ideogram for
43:02 my business today awesome Andrea that's
43:04 beautiful that's really cool
43:06 yeah it's it's listen the the the uh the
43:10 image generation part of ideogram is a
43:12 little it's a little sketch right now
43:14 it's it's more like Dolly than it is uh
43:17 stable diffusion or mid-journey
43:20 um but the text is good so people are
43:21 going to use it and a lot of people are
43:23 using it I think they did some crazy
43:25 amount of millions of images in their
43:27 first week thank you Robert for the Duke
43:30 the Dog nut I like the dog nut very
43:33 tasty I like a sugary snack I like a
43:37 sugar and snack Salesforce is a tool for
43:39 management not sales people big brother
43:41 yeah Salesforce is well I have a small
43:44 company we use Salesforce if we didn't
43:46 have Salesforce we couldn't remember who
43:48 the [ __ ] we talked to and like you know
43:49 we're an 11 year old company so
43:51 understanding that oh that person used
43:54 to work there and they used us and like
43:56 so Salesforce has its value but it's
44:00 just a you know talk about janky pieces
44:02 of [ __ ] their new interface is called
44:05 lightning and it's
44:06 slower than [ __ ] slower than molasses
44:12 but
44:14 sprinkle a little generative AI layer on
44:16 top of that it's going to be pretty good
44:19 all right uh productive like GPT play
44:22 hide and seek
44:23 with the kids for two hours
44:27 oh man Zach oh that's funny actually
44:31 yeah you can probably you could probably
44:33 do all sorts of cool games with pie like
44:36 there's something about something about
44:38 pie that I'm just I'm deeply intrigued
44:40 by because
44:43 she's gonna get well I have her set as a
44:45 she right now my pie is a she Australian
44:48 woman
44:49 she's going to
44:52 um
44:53 she's gonna start remembering more and
44:55 more and I don't know I there's
44:58 something really interesting going on
44:59 with that company
45:01 you have gorgeous hair I have three
45:04 thank you yeah my forehead has gotten
45:06 much taller but uh but yeah I still got
45:09 some some functional follicles
45:12 and they need cut this is a this is some
45:15 definite 19 this is like late 70s early
45:19 80s Sears Roebuck car carpet salesman
45:22 [ __ ] going on here
45:23 huh yeah it's bad tonight whatever
45:27 cyber security question mark
45:31 yes we should have it
45:34 um is your question
45:37 cyber security with these tools they're
45:40 janky pieces of [ __ ] right now uh open
45:43 AI just
45:45 four days ago just announced
45:49 hang on
45:52 um
46:01 foreign
46:07 T Enterprise so there is now
46:12 a version of chat GPT
46:15 that
46:17 has entered prize grade security and
46:20 privacy
46:21 customer prompts and Company data are
46:24 not used to train open AI models okay
46:26 good thumbs up
46:28 because now now here's what I don't know
46:32 their user their uh end user license
46:35 agreement
46:36 contradicted what they used to say
46:38 publicly oh we're not going to train on
46:40 the stuff but what their terms and
46:42 service says is we might
46:44 so I assume with this product they
46:48 actually don't they got data encryption
46:50 at rest they got certif they're
46:52 certified stock 2 compliant
46:54 they got an admin console single sign-on
46:57 domain verification analytics dashboards
47:00 for insight
47:02 um the admin console for bulk member
47:05 management that's actually really
47:07 important oh sorry anything bigger so
47:10 you can see them you're like Kyle you
47:12 know where irregular is we can't see
47:14 this crap
47:16 so there's all that crap
47:19 um
47:20 the permissions management is going to
47:23 be critical for an Enterprise product
47:25 right because you don't want an
47:28 administrative assistant chat gpting
47:30 into the same data that the CFO has
47:33 access to right that's super critical so
47:37 how do you take all of your data you
47:40 know of your organization and put it
47:43 into these magical mythical large
47:46 language models but do it in a way that
47:49 you don't reveal the wrong stuff to the
47:51 wrong people
47:53 um so yeah oh and then what else do they
47:56 have uh they've also got unlimited
47:58 access to gpt4 higher speed performance
48:01 two times faster unlimited access to an
48:04 advanced data analysis which is
48:07 um old code interpreter so so you get
48:10 that
48:11 32 000 token context window shareable
48:15 chat templates so you're going to be
48:16 able to have prompt libraries and then
48:18 free credits to use their API to build
48:20 [ __ ] so it looks pretty good what's
48:22 missing from this and what what Sam
48:24 Altman said they're going to add you
48:26 know as soon as they can is training
48:29 chat GPT on your company's data right so
48:33 this is everything but that
48:35 and this is you know this is welcome
48:37 they're starting to grow up our little
48:39 children growing up it's so cute they're
48:43 so cute you get an Enterprise security
48:46 version now oh they're so cute
48:50 yeah this was necessary this was needed
48:52 because you know everyone you talk to
48:54 they're like yeah but my company can't
48:56 use it well now your company can
48:59 now the question is will your company or
49:02 are they going to remain a company full
49:04 of secret cyborgs I don't know if you've
49:06 heard that term but it's from Ethan
49:09 Malik who's A Wharton Professor that
49:11 talks about the fact that
49:13 individuals right now are using chat gbt
49:16 in large language models to dramatically
49:17 increase their productivity but they're
49:19 not telling their companies about it
49:21 because the companies might fire them or
49:23 chop their heads off or whatever the
49:25 company's going to do
49:27 um and it's creating this disparity
49:30 where you've got an AI literate you know
49:32 subclass in a company that that might be
49:36 doing things like getting 40 40 hours
49:38 worth of work done say in five hours or
49:41 10 hours or two hours
49:45 um and what his point is you gotta
49:48 management has to First Embrace this
49:51 [ __ ] and then has to like
49:54 um
49:55 aggressively reach out to people that
49:58 are AI literate and bring them into the
50:00 fold and make it safe for them to do
50:02 that
50:03 your first sponsored Pros post will be a
50:06 hair care product
50:07 hey it's Kyle Shannon with the AI
50:09 learning lab you know some nights I'm in
50:12 here and I've got the the flyaways and
50:14 dandruff sprinkling off and smacking
50:17 into the camera it was embarrassing
50:20 yeah I it could be it could be you know
50:24 then I got hair Gunk look at it now
50:27 that's fantastic is that a toupee or is
50:30 that real oh it's real it's hair Gunk
50:33 get yours today click down here
50:38 could do that we could do that we can
50:40 turn this into that we don't need to be
50:42 about education screw that this is
50:45 America where capitalists we're gonna
50:47 start selling [ __ ] we're gonna start
50:49 selling T-shirts uh buttons buttons
50:54 you don't know what this is do you
50:55 people you don't know about the button
50:59 hmm
51:00 [Applause]
51:03 it's gonna be nuts
51:05 got me nuts
51:07 it's gonna be like Walmart in here we're
51:09 gonna have fishing equipment
51:11 boxes of crackers you can get everything
51:14 I'm gonna be they'll be fantastic
51:17 [Laughter]
51:19 all right
51:22 I'm saying it got angry without me
51:23 programming it that's very strange
51:25 Robert Rossi maybe they're experimenting
51:27 or you know again back to all of these
51:30 tools all of these tools all of these
51:33 tools
51:34 these all of these tools janky pieces of
51:38 [ __ ] right now
51:39 not because the engineers behind them
51:42 are bad programmers just because it's
51:45 early it's we're really early
51:49 opened a Floodgate
51:51 where companies all over the land
51:55 all over the world are saying oh my God
51:57 we gotta catch up and so they're all
52:00 just throwing [ __ ] out into the public
52:02 that probably hasn't been tested they're
52:04 like I need this I need this launched
52:06 yesterday
52:08 uh but we said it was going to take six
52:11 months I don't care it hasn't been
52:13 tested launch it
52:15 [Laughter]
52:18 and I'm pretty sure that's how Senior
52:20 Management all talks
52:23 uh I'm kind of miserably mesmerized by
52:26 your shirt
52:27 what if I do this I'll just do like him
52:30 on a boat
52:31 hi Kyle Shannon on a boat
52:37 I'll investigate it more yeah if you
52:39 find out anything Robert I'd be curious
52:42 to know
52:43 um yeah that's very strange it's very
52:45 strange unless you inadvertently like if
52:47 you were in their playground and and you
52:49 sort of picked angry but if if you
52:51 didn't
52:52 if you didn't do that and if it was a if
52:54 it was a uh an API call maybe there was
52:57 just some weird glitch where your thing
52:59 got tagged inadvertently or they had a
53:00 setting they changed their default to
53:03 pissed off they changed their defaults
53:07 AI learning lab you know
53:10 and another thing Johnson
53:12 get your dick off my desk
53:14 [Laughter]
53:16 oh Claude for someone else thank you
53:19 Robert for summarizing earnings call
53:21 transcript very good can query the
53:23 transcript too thank you kavuna
53:26 uh
53:32 uh I think his here wait if we know what
53:35 we mean I think he's here
53:38 I don't quite know what that means togas
53:40 Kyle I'm so sorry but I can't handle the
53:42 shirt it's breaking my brain good night
53:44 bye Jen
53:47 sorry
53:48 you win some you lose some with the
53:50 fashion choices you know what I'm saying
53:53 some people enjoy the shirt other people
53:55 are having seizures
53:58 it's gonna it's gonna happen all right
54:00 which good cartoon animation program do
54:03 you recommend that is similar to Dudley
54:05 I don't know Dudley let me go look at it
54:09 dude Lee so when it comes to these super
54:12 tactical tools I tend to be fairly
54:15 clueless about them oh I see this kind
54:17 of stuff
54:24 I don't know
54:30 my recommendation would be to go to
54:35 not there go to their futurepedia.io
54:41 and look up animation and look up
54:45 video animation cartoon
54:51 I would be surprised if there weren't
54:53 some tools that are specifically coded
54:56 to do that
54:57 how I would think about it right now is
54:59 I would say okay I can go to Mid Journey
55:02 I can figure out I can get it into a
55:04 style that that I want
55:06 and then generate little cartoon dudes
55:09 and then maybe take that to Runway or
55:12 yeah to Runway ml with uh with Gen 2
55:15 where you can upload an image and turn
55:16 an image into a video so I might do it
55:18 like that
55:20 um
55:22 but I would imagine that there are
55:23 people who have done that and probably
55:25 some of the companies like Dudley
55:27 are figuring out how to ai ai
55:31 power what they used to do so I would
55:34 just I would just pay attention to all
55:36 those companies and see see who's
55:38 talking about AI or or has made
55:40 announcements and just pay attention to
55:42 those but I I just don't know it's a
55:44 really specific thing
55:46 um but to be clear like
55:49 probably within a year you're going to
55:52 be able to do a fully animated doodly
55:55 kind of thing with voice over and the
55:59 text written and the animation and the
56:01 background music it'll just [ __ ] make
56:03 it
56:04 so without you having to like you know
56:07 do all the [ __ ] you have to do now
56:09 with all these storytelling programs
56:13 um he heard it then on irregular
56:18 ideogram yes ideogram dot a i is the
56:22 thing that made
56:24 that
56:26 weird I'm a regular
56:32 um
56:35 I can go change my shirt if this shirt
56:37 is bugging people I can go change it I'm
56:39 happy to do that oh ideogram is raised
56:41 22 million dollars so far including
56:44 andreasen based in Toronto interesting
56:46 there you go yeah that sounds about
56:48 right like that's a that's a really
56:50 solid it's probably a series A
56:53 um it might even be a seed round in this
56:54 Insanity around Ai and it's it's good
56:57 right so
56:58 yeah they're gonna they're gonna do some
57:00 good things they'll they'll their next
57:02 raise because they crack this first
57:04 people have been asking about this for a
57:06 year and a half
57:07 why can't I do text why the hands look
57:09 like [ __ ] what's wrong with your feet
57:11 why can't it do text can it do text how
57:13 do you make a sign how do you do text
57:15 it's like it's a lot of what I hear in
57:18 this channel is how do you make text and
57:20 they cracked it
57:21 the fact that they cracked it before the
57:23 the majors is is staggering to me so I
57:26 have a feeling their next round is gonna
57:27 be
57:28 significantly north of 22 million
57:31 dollars and their valuation is going to
57:33 be
57:33 stinging
57:35 what's what's what's also interesting
57:38 about these guys what what might hold
57:40 their valuation down is what they
57:42 trained their image model on
57:45 it's a crappy enough image model that
57:47 they might have done what Adobe did with
57:49 firefly where they trained it on public
57:51 domain and you know images that they had
57:54 the rights to that's why fireflies not
57:56 as good as mid-journey or stable
57:57 diffusion but it's not as problematic
57:59 because it wasn't trained on copywritten
58:01 data so depending on their approach to
58:04 copyright
58:05 that's going to be the Achilles heels of
58:08 these these video and image generating
58:10 tools is what they were trained on
58:13 um as as the lawyers get their [ __ ]
58:15 together and the lawyers I'll tell you
58:17 what intellectual property lawyers right
58:19 now they're like doing Conga lines
58:22 [Music]
58:24 hey Marge yeah
58:28 would you bring me in that yacht catalog
58:31 again
58:35 yeah
58:37 I'm an IP attorney oh you want to sue
58:40 everyone awesome let me write that up
58:42 for you
58:43 oh you want to counter sue the other
58:46 people awesome let's do that too all
58:48 right
58:49 y'all kind of think of it
58:51 I need to fly a lot could you bring in
58:53 the jet catalog too
58:56 yeah that's IP attorneys right now
58:59 [Laughter]
59:02 next 20 years next 20 years it's just
59:06 gonna be like
59:08 just gold coins falling from the ceiling
59:11 of law firms oh my God
59:16 it's gonna be crazy it's gonna be crazy
59:19 it's going so crazy you don't even know
59:21 how crazy it's gonna be I mean nuts I
59:23 mean nuts
59:25 uh thank you for that update on ideogram
59:27 and what they raised that that feels
59:28 like like
59:30 yeah that feels right that feels about
59:32 that feels right Harvard releases
59:35 guidelines for AI use in classrooms
59:37 thank [ __ ] god did they that's
59:39 awesome
59:42 they also wasn't it was it Harvard
59:47 they're they're comp PSI class that that
59:49 one the the like the compsi 101 class I
59:52 think it was Harvard or their
59:54 programming class that the teacher's
59:56 assistant was going to be a chatbot a
59:58 like a chat GPD chatbot so you'd have 24
1:00:01 7 access and it would check your work
1:00:03 and stuff like that
1:00:05 um
1:00:06 Harvard
1:00:10 Harvard AI guidelines
1:00:19 released its initial guidelines
1:00:31 generative artificial intelligence
1:00:34 welcome to Harvard yes
1:00:36 my name is Winchester
1:00:39 McKinsey III uh at Harvard as you know
1:00:44 welcome I knew your grandfather well
1:00:47 um
1:00:48 as you know at Harvard we take
1:00:51 the institution quite seriously general
1:00:54 of AI is clearly a tool of the future
1:00:57 and
1:00:58 we would like you
1:01:00 all right
1:01:01 so let's see what we got
1:01:04 generated AI is a type of artificial
1:01:06 intelligence that can learn and mimic
1:01:08 large amounts of data
1:01:10 learn from and mimic large amounts of
1:01:12 data
1:01:14 to create content such as text images
1:01:16 music videos code and more based on
1:01:18 inputs or prompts
1:01:21 the university supports responsible
1:01:23 experimentation with generative AI tools
1:01:26 [ __ ] yes
1:01:31 yes
1:01:34 you don't have to love them
1:01:36 you don't have to do everything with
1:01:37 them
1:01:38 responsible experimentation that's the
1:01:41 stage we're all in right now every
1:01:44 University should take the lead here
1:01:46 what's interesting is Boston was the
1:01:49 first city that had generative AI
1:01:51 guidelines and I I know that Harvard's
1:01:54 not the first to do this but Harvard
1:01:55 doing this is a big big [ __ ] deal
1:01:57 when did they do this
1:02:00 do we have a date
1:02:04 July 18th so this is this been out for a
1:02:07 bit
1:02:11 protect confidential data yes review
1:02:13 content before publication it lies make
1:02:15 sure that it doesn't live before you
1:02:17 turn before you make stuff public or
1:02:19 turn it in adhere to existing academic
1:02:21 policy good don't cheat
1:02:23 be alert for fishing yeah
1:02:26 oh generative AI has made it easier for
1:02:28 malicious actors to create sophisticated
1:02:30 phishing emails and deep fakes so you're
1:02:32 going to get fished
1:02:35 news and tools
1:02:37 here's tools
1:02:40 open AI API access available now through
1:02:44 the Harvard University I.T Department
1:02:47 AI sandbox pilot coming soon Friday
1:02:50 August 11th so this is like a whole this
1:02:52 is like a living document where they're
1:02:54 updating this this is great this is
1:02:56 [ __ ] great
1:02:57 it's great
1:02:58 awesome
1:03:03 awesome awesome awesome awesome
1:03:07 I would have gone to Harvard I would
1:03:09 have done it I would have done it
1:03:13 you know
1:03:14 if I was accepted or applied you know
1:03:17 you know what I'm saying you know what
1:03:18 I'm saying I definitely Penn State was
1:03:21 the University for me right
1:03:24 because I lived in Pennsylvania
1:03:25 Go nindy Lions yay
1:03:28 [Laughter]
1:03:33 All Right Moving On
1:03:36 uh the a an AI quadcopter has beaten
1:03:39 human Champions at drone racing no [ __ ]
1:03:41 really
1:03:43 those human
1:03:44 if you watch those drone races
1:03:47 like I can't fathom that humans have
1:03:51 reactions that are that fast like it
1:03:53 it's astounding to me what what those
1:03:56 drone Racers do it makes sense that an
1:03:58 AI
1:03:59 could get good at that but holy crap
1:04:02 that's crazy that's crazy
1:04:05 I wonder I wonder if it beat them on a
1:04:08 course that it knew a lot like I wonder
1:04:09 how fast I can learn a course like if
1:04:12 you take a professional drone racer and
1:04:14 an AI
1:04:14 I wonder I wonder who learns it faster
1:04:17 because like I know once you train an AI
1:04:19 on a course I bet it you know it just
1:04:21 does it [ __ ] brilliantly every time
1:04:24 wow that's crazy that's nuts that's very
1:04:27 cool
1:04:28 Gannett has oh
1:04:33 I was reading a comment tick tock
1:04:37 well I didn't do this to me and thanks
1:04:40 for the gummies uh I like the gummies
1:04:42 the gummies is tasty I like a taste of
1:04:45 gummies uh I like to taste of gummies
1:04:47 that they got the good uh the the tasty
1:04:50 the tasty fruit I like the the fruit the
1:04:52 fruit gummies the the turkey gravy gummy
1:04:55 not so much I don't like the turkey
1:04:57 gravy gummy Uh some people made the
1:04:59 turkey gravy gummy and they said how
1:05:01 would you like I know you like gummies
1:05:02 would you like a turkey gravy gummy and
1:05:04 I thought oh that could be tasty and
1:05:06 then I tasted it and I was like no no
1:05:07 not this is wrong that's a I like the
1:05:10 fruity gummy I like the fruit of gummy
1:05:13 thank you sugar
1:05:15 all right where the [ __ ] was I
1:05:22 connect to pause AI experiment after
1:05:26 botched High School sports articles
1:05:29 by Claire Duffy CNN
1:05:31 um doesn't surprise me that gets back to
1:05:34 Harvard's rule number two
1:05:37 verify your work before you publish it
1:05:39 yeah so so so so there's there's gonna
1:05:44 be a a just a [ __ ] parade of media
1:05:49 articles of Senior Management going AI
1:05:53 writes articles fire all the reporters
1:05:56 this is awesome
1:05:58 oh what do you mean it lied oh we're
1:06:00 going to be sued for libel don't use AI
1:06:02 anymore
1:06:04 um
1:06:05 that's just going to continue and that's
1:06:07 just frankly like I have no sympathy for
1:06:10 them like
1:06:12 part of part of the mission of this
1:06:14 channel is I want people
1:06:17 individuals and businesses to experiment
1:06:20 with these tools enough to understand a
1:06:23 what they do B how they do it see what
1:06:27 they're good at and D but they're not
1:06:30 good at
1:06:32 and then don't be an idiot
1:06:34 have strategy lead with strategy
1:06:38 um
1:06:39 what's that thinky thing we do oh right
1:06:41 critical thinking
1:06:43 like actually think about these things
1:06:45 and how you use them and if you
1:06:47 understand what they're good at and what
1:06:49 they're not good at then you can put
1:06:50 processes in place that say sure have
1:06:53 these machines write a bunch of first
1:06:54 draft articles and then let's beef up
1:06:56 our editorial our editors
1:06:59 to like make sure this stuff's actually
1:07:01 good
1:07:02 infactual and solid
1:07:05 that's a process problem not a
1:07:07 technology problem that that was someone
1:07:09 going ah AI is fantastic launch it
1:07:14 we're gonna make a fortune
1:07:18 that level of the ability to just have
1:07:23 machines do [ __ ] like that is coming
1:07:25 this is not there now so everything's an
1:07:29 experiment right now and should be
1:07:30 treated as such everything
1:07:33 um even the even the point
1:07:35 um what Robert was saying about 11 Labs
1:07:38 just all of a sudden got angry
1:07:42 one of the one of the amazing things
1:07:46 about being alive right now is we've got
1:07:47 all these systems where you've got API
1:07:50 calls to all these different systems
1:07:51 right
1:07:52 so as a developer it's really easy for
1:07:55 you to go oh I can put together a
1:07:57 remarkable tool well if a month from now
1:08:01 after you've built your remarkable tool
1:08:03 the service on the other end of that API
1:08:05 if they change something so like the
1:08:07 default is angry or or or they change
1:08:11 the default from don't cuss to cuss and
1:08:14 you're not paying attention to it you're
1:08:15 going to be in trouble right so that's
1:08:17 the downside of all this to us and
1:08:20 you're not paying attention to it you're
1:08:22 going to be in trouble right so that's
1:08:23 the downside of all this you know you
1:08:26 know API accessible Services is when
1:08:29 they just you know change [ __ ]
1:08:32 so so gotta be careful with this stuff
1:08:36 um forgot to use your button I don't
1:08:38 know what button are you referring to Oh
1:08:41 you mean this bathroom
1:08:44 ah would you lie to my mother you could
1:08:47 do it you can do money with the chat
1:08:53 is um how do we say
1:08:56 um
1:08:57 is um the money in the gold ah yes the
1:09:02 like the the bigger brain make it rain
1:09:05 Make It Rain uh
1:09:08 Goldmine uh you Gold Miner you can
1:09:15 yeah yeah
1:09:18 visual artists fight back against AI
1:09:21 companies for repurposing their work
1:09:22 they do
1:09:25 um thank you very much Winston
1:09:26 for making a rain I like the rain
1:09:31 um
1:09:39 yeah
1:09:41 artists are pushing back against it and
1:09:45 it's not going away
1:09:48 it's you know what's weird right now is
1:09:50 like I feel like the framing of this
1:09:52 conversation is
1:09:55 big corporations are [ __ ] that are
1:09:57 going to steal everything which you know
1:09:59 I understand the default right
1:10:02 um
1:10:03 a lot of those big corporations are
1:10:05 going to have their business models
1:10:06 upended too
1:10:09 right
1:10:12 so so just as as the artist's work may
1:10:15 be repurposed the
1:10:23 the
1:10:25 the ability for anyone to
1:10:28 easily make world-class content without
1:10:31 needing
1:10:32 Sony Entertainment
1:10:34 to be the sponsoring entity is gonna
1:10:37 [ __ ] up Sony Entertainment too like this
1:10:39 isn't just [ __ ] up the actors and the
1:10:42 artists and the writers and the voice
1:10:43 over people and the
1:10:45 CGI people and the effects people
1:10:49 it's gonna [ __ ] up business models like
1:10:54 business models
1:10:56 right so Spotify like here's a here's a
1:10:59 good example
1:11:01 Spotify when the Drake and weekend song
1:11:04 came out and it got all those views they
1:11:07 yanked the song right so it's a fan
1:11:09 generated
1:11:11 piece of work
1:11:13 they synthesized Drake in the weekend
1:11:15 gets all these views on Spotify Spotify
1:11:18 pulls it down
1:11:21 and so
1:11:23 because Spotify has to not piss off all
1:11:25 the artists because they want the
1:11:27 artist's work on their platform
1:11:30 they're gonna have to
1:11:32 reject AI generated [ __ ]
1:11:36 well what does that do
1:11:38 it creates a giant [ __ ] vacuum
1:11:42 where some snot-nosed little kid
1:11:45 or some pissed off gen xer like me is
1:11:49 gonna go hey
1:11:50 what if we had a Spotify that was
1:11:53 nothing but AI generated synthesized
1:11:55 songs
1:11:58 and it's not going to be Sony doing it
1:12:00 and it's not going to be Spotify doing
1:12:02 it it's going to be a couple of kids in
1:12:03 a
1:12:04 college dorm room they're like oh that's
1:12:06 [ __ ] cool
1:12:08 and they'll start out with all the grime
1:12:10 stuff because Grime said do anything you
1:12:12 want with my likeness just give me 50 of
1:12:14 it so they'll cut a deal with her
1:12:17 and she'll be the front and center star
1:12:19 and then there'll be all these other
1:12:20 ones and then they'll just have like the
1:12:21 dark corner and in the dark corner you
1:12:24 can listen to the [ __ ] you're not
1:12:25 supposed to because Sony Music doesn't
1:12:27 like it and Spotify won't play it and
1:12:29 guess what
1:12:30 that service
1:12:32 will obliterate Spotify at some point
1:12:40 and it'll [ __ ] up Sony Music so so even
1:12:42 though the the artists right now are
1:12:45 pissed off and they're like you're
1:12:46 you're [ __ ] for using AI everyone's
1:12:49 gonna use AI the companies the
1:12:52 individuals the artists the upstarts
1:12:56 so
1:12:58 yeah
1:12:59 your button is far higher I know I gotta
1:13:02 I gotta re-record it I gotta figure out
1:13:03 exactly what I want to do with the
1:13:05 recording and then I'll get it dialed in
1:13:06 I'll get it there
1:13:08 I think so it has to be context based it
1:13:10 freaks me out API oh okay Robert Robert
1:13:14 Rossi okay so Robert Rossi
1:13:17 with the 11 Labs voice thing he called
1:13:20 via the API and what came back was it
1:13:24 was pissed off
1:13:25 it talked it talked it talked angry it
1:13:28 was angry talk oh
1:13:30 11 laps you talk so angry hmm
1:13:35 um
1:13:36 little spit grit a whole lot of duct
1:13:39 tape yeah exactly I like totals
1:13:42 I like turtles too I was just talking to
1:13:45 Pie about the irregular she loves that
1:13:48 we exist
1:13:51 I'm not weird
1:13:53 I'm a regular
1:13:55 I like turtles oh
1:14:02 mine too
1:14:04 we are King Kyle's Army I am no King ah
1:14:09 life's hard it's harder if you're stupid
1:14:12 um
1:14:13 yeah I I am uh I am not a king I am new
1:14:17 king I am new king oh look
1:14:21 look how young I look when I put tape
1:14:23 over the lens ooh stuck
1:14:25 sorry 80d moment
1:14:33 tape's fun
1:14:38 hair giving off Monty science Morty
1:14:41 Seinfeld Vibes it really is
1:14:44 good lord it's just I just I just can't
1:14:47 I just can't be bothered my electrical
1:14:49 system on my car lost its [ __ ] mind
1:14:51 maybe Chachi BT did it
1:14:53 jet GPT expands custom instructions
1:14:57 for free users yep they yeah they did
1:14:59 that a while ago
1:15:01 if you don't know what custom
1:15:02 instructions are let me just show you
1:15:04 that that's that's worth taking a quick
1:15:05 look at
1:15:06 so when you use chat GPT
1:15:11 um
1:15:14 you know you just tell it [ __ ] like uh
1:15:17 let's see like um what happened there
1:15:24 oh I clicked on it I clicked on one of
1:15:26 their suggested
1:15:27 one of their suggested things let me do
1:15:29 a new chat so if I go to church EPT and
1:15:32 go
1:15:32 um
1:15:34 um
1:15:36 write me uh let's see me a tweet
1:15:43 about irregularity
1:15:51 all right
1:15:53 go all right
1:15:56 embracing life's irregularity sometimes
1:15:58 in those beautiful patterns emerge when
1:15:59 we let go of perfection and let the
1:16:02 unique twists and turns create our story
1:16:04 okay
1:16:06 the fact that it didn't immediately go
1:16:08 to a poop joke is very disappointing but
1:16:10 fine it it
1:16:12 took irregularity in context and it did
1:16:16 something appropriate fine I'll take
1:16:17 that but
1:16:19 with custom instructions so you go to
1:16:22 your
1:16:23 um settings
1:16:24 and there's this thing called custom
1:16:26 instructions
1:16:27 and so you can tell it the kind of
1:16:30 person you want it to be so I want you
1:16:32 to be in in my case I wrote here you're
1:16:34 a master gardener who specializes in
1:16:36 giving gardening advice
1:16:39 um and then how would you like chat GPT
1:16:40 to respond when anyone asks you about
1:16:42 gardening before you ask them before you
1:16:45 answer them ask to clarifying questions
1:16:48 so you really understand their goals
1:16:49 then provide them with a short answer
1:16:52 I'm going to say after they answer wait
1:16:54 I gotta turn this on wait enable for new
1:16:57 chats
1:16:59 um
1:17:00 after they answer
1:17:04 provide them with a with a short answer
1:17:07 and a bulleted list of next steps so I'm
1:17:11 going to turn this on
1:17:13 I'm going to say new chat
1:17:15 I'm gonna say
1:17:17 um I want to make
1:17:21 a cucumber Garden
1:17:25 uh like cucumber Garden
1:17:32 in
1:17:34 Denver
1:17:37 um
1:17:39 absolutely I'd be happy to help you
1:17:40 create a successful cucumber Garden in
1:17:42 Denver could you please clarify a bit
1:17:45 about your goals we told it to ask a
1:17:46 question and then ask two specific
1:17:48 questions do you have a specific area in
1:17:50 mind for the garden are you looking to
1:17:52 grow cucumbers for personal use or on a
1:17:54 larger scale
1:17:56 uh let's just say
1:17:58 um in my yard
1:18:01 and personal
1:18:06 um
1:18:08 great
1:18:09 and so now give me the answer thanks for
1:18:12 the clarification here's your concise
1:18:13 guide to getting started with a cucumber
1:18:15 Garden in Denver and there's my little
1:18:16 bulleted list choose a Sunny Spot
1:18:19 well-draining soil planted cucumber
1:18:22 seeds or seedlings provide a trellis
1:18:24 let's say
1:18:26 um
1:18:29 it's August is it still okay to plant
1:18:32 seeds
1:18:41 just a couple more questions what's the
1:18:43 average temperature range in your area
1:18:44 during August are you planning on to
1:18:46 grow
1:18:47 cucumbers from seeds or seedlings they
1:18:49 kind of screwed that up but you see it's
1:18:51 now behaving as we want it to behave so
1:18:53 custom instructions let you
1:18:55 fine-tune chat gbt to do stuff that you
1:18:59 like how else could you use it you could
1:19:00 do things like
1:19:02 um
1:19:03 you could say you know I want you to be
1:19:06 my assistant and
1:19:08 um
1:19:10 write emails for me when I ask you to
1:19:12 and here's an example email and how I
1:19:15 write and I want you to follow the tone
1:19:16 and structure of that email and it will
1:19:20 just do that every time you use it
1:19:22 so that's Q in that queue in that Q
1:19:25 that's pretty cute so that's available
1:19:27 to everyone that's in cheap gbt 3.5 for
1:19:30 free
1:19:33 and uh yeah open AI deploys web crawler
1:19:36 in preparation for gpt5 yes correct
1:19:39 Thelonious bring in the news tonight
1:19:41 thank you for this this is awesome
1:19:44 um
1:19:44 yeah they they they launched a web
1:19:47 crawler to to go
1:19:49 um
1:19:51 uh train Chachi bt5 so they could get
1:19:53 more of the internet
1:19:54 and a lot of the big media companies are
1:19:57 like oh we can just disable the web
1:20:00 crawler and [ __ ] them until they pay us
1:20:03 hundreds of millions of dollars and
1:20:04 that's what all of them are doing right
1:20:06 now and Google
1:20:08 is taking their giant piles of cash that
1:20:10 they've had for they've been compiling
1:20:12 for a bunch of decades and they're
1:20:14 cutting really good deals with all the
1:20:16 media companies and I bet they're gonna
1:20:18 do their best to cut out uh open AI from
1:20:22 the good data and then open AI gonna
1:20:24 struggle it won't be it gonna be very
1:20:28 salty it's gonna be very Saucy they
1:20:30 won't they're gonna leg wrestle using
1:20:32 Elon Musk and and Mr Zuckerberg you
1:20:35 think that fight's gonna be a good one
1:20:38 Google and open AI That's gonna be a
1:20:41 tussle
1:20:42 they're gonna be a tussle SEC turns its
1:20:45 gaze from crypto to AI interesting
1:20:49 latest AI news and that's the way it was
1:20:52 I actually met Walter Cronkite I was
1:20:55 selling yo-yos at FAO shorts it was my
1:20:58 first job in New York City selling it
1:21:00 yoyos at FAO shorts and I remember
1:21:02 Walter Cronkite came up the stairs with
1:21:04 this with his son and I I did a trick
1:21:07 and Walter Cronkite said whoop walking
1:21:11 the dog
1:21:12 laughs
1:21:18 I got good stories that's a good one
1:21:20 love humor thank you I love humor too
1:21:24 pleasure kind of
1:21:28 how was the most recent Meetup it was it
1:21:30 was quite good we had um oh thank you
1:21:32 for the beaver
1:21:34 I do rather enjoy a beaver that one was
1:21:37 lovely that was a lovely little fuzzy
1:21:39 beaver and beaver's wonderful wonderful
1:21:43 little creatures
1:21:44 um
1:21:45 yeah the uh the AI Salon was was really
1:21:48 good we had five different people
1:21:52 um share their projects sander uh who we
1:21:55 met here
1:21:56 presented his his new uh sort of ai's
1:22:00 tool directory kind of thing uh from
1:22:04 from the Netherlands at like two in the
1:22:07 morning for him so that was pretty cool
1:22:09 uh Bondo for hairdos I need Bondo my son
1:22:13 designed a bot that allowed him to clock
1:22:15 into school while he slept in
1:22:18 solid solid I like the uh I like the
1:22:22 initiative I like it
1:22:25 and like what's wrong with that he's
1:22:26 learning how to do that he's learning
1:22:27 how to use the tools
1:22:30 how do you get your page to look like HD
1:22:34 I'm using an iPhone 14 maybe I don't
1:22:37 know I don't know
1:22:39 it could just be that you know
1:22:41 This is So Glorious it enhances the
1:22:44 pixels
1:22:47 no
1:22:48 okay all right five
1:22:52 the Kyle irregular Island Captain
1:22:53 that'll that'll take I'll be your tour
1:22:56 guide I'm your regular tour guide uh
1:22:59 what did GPT say about what uh
1:23:04 um
1:23:06 what was the thing
1:23:15 um I don't know what that the answer to
1:23:17 that question was
1:23:22 there a spot for us on Discord just
1:23:25 posted in the salon we need an irregular
1:23:27 channel uh that's a good point I'll make
1:23:29 in a regular channel I was also thinking
1:23:32 um I think I think what I might do is is
1:23:34 create an irregular
1:23:37 um role
1:23:39 and then
1:23:41 um only the Irregulars can have access
1:23:43 to that channel
1:23:44 um I think that would be good so yeah I
1:23:46 was thinking about that the other day so
1:23:48 I think I will do that that's a very
1:23:50 good idea jrc now here's the deal here's
1:23:52 what I don't quite know in that little
1:23:54 scenario
1:23:55 like I don't really care if someone says
1:23:57 they're an irregular and they don't
1:24:00 really watch my lives but but um but how
1:24:03 do we
1:24:04 let people like I don't know how to know
1:24:07 who the joined
1:24:10 Discord is the person that's in here
1:24:13 because sometimes the names don't match
1:24:15 so that I need to figure out
1:24:18 what was picture to video site the
1:24:20 picture to video site is is runwayml
1:24:22 I'll go show you how to do that because
1:24:24 it's super [ __ ] cool
1:24:27 so we're gonna go Runway
1:24:36 [Music]
1:24:38 log in with Google because I want them
1:24:41 to have all of my data all of it you can
1:24:43 see everything I do Google I like how
1:24:46 you hug me
1:24:48 um you go text to video
1:24:50 and then within text to video you have
1:24:53 an image upload
1:24:56 and so like here's a whole bunch of
1:24:59 images from my camera roll I guess yeah
1:25:03 so there's a bunch of photography I did
1:25:07 um
1:25:08 why can I not scroll the page up can I
1:25:14 that's weird
1:25:17 maybe do a command minus
1:25:22 this is strange it's trash
1:25:26 but let me I'll get a cool
1:25:29 let me get a cool uh
1:25:33 Cloud image here's a good one
1:25:37 all right so there's the little image
1:25:39 we're going to turn into a video
1:25:43 oh cool so they've added some new things
1:25:46 so this is motion so I can I can
1:25:49 increase or decrease the motion that's
1:25:51 really good because one of the
1:25:53 frustrations with with image to video is
1:25:56 sometimes you wanted it to move a lot of
1:25:57 things a lot and it wouldn't move them
1:25:59 barely at all so let's say we want this
1:26:02 to move a lot and then there's another
1:26:03 setting here
1:26:05 oh you can you can
1:26:07 use a seed so you can you can use an
1:26:10 image over and over again
1:26:12 you can interpolate and frame out smooth
1:26:14 out your frames you can upscale it
1:26:16 automatically you can remove a watermark
1:26:18 okay all right so
1:26:22 let's generate it I can't why can't I
1:26:24 generate it need help with prompting
1:26:27 um
1:26:34 205 seconds left
1:26:36 do I need to write something oh I do
1:26:39 um
1:26:40 um
1:26:42 Breezy sky
1:26:46 okay so now we're gonna turn that image
1:26:49 into
1:26:51 a breezy sky with an 8 out of 10 motion
1:26:54 setting
1:26:58 and we'll see how it does nine seconds
1:26:59 oh this is slower than it used to be but
1:27:02 you know what do you want it's turning a
1:27:04 picture into full-on video
1:27:06 shut up you whiny baby I want it now I
1:27:11 want it now
1:27:12 I can't believe that it takes me three
1:27:15 seconds to make fully functional video
1:27:17 this is annoying
1:27:22 ah
1:27:24 we're at 44 it's gonna be so good
1:27:28 do I have anything here where you can
1:27:30 see what I did before
1:27:33 I could come back to this but I'll just
1:27:36 I'll just deal oh my God what am I gonna
1:27:39 do with my car tomorrow oh it's annoying
1:27:46 I think I might just pull out I know
1:27:48 which relay I think is bad I think I
1:27:50 might just pull out the relay and go to
1:27:52 the automotive store and see how much
1:27:54 that relay is
1:27:58 oh come on this is actually much slower
1:28:01 than it was but my My Hope Is that what
1:28:03 it's going to generate is much better
1:28:04 than it was because it used to be a
1:28:07 little a little soft and weird
1:28:11 oh I think I had upscaling turned on so
1:28:13 it's probably upscaling as it's
1:28:14 generating this
1:28:17 all right we're at 80 it's gonna be good
1:28:19 it's gonna be okay I'm gonna be okay
1:28:22 so micro Runway ml it's the one
1:28:29 um
1:28:30 I know there's some others out there
1:28:32 okay here we go all right you ready
1:28:38 okay that was weird
1:28:41 that was nice that was not
1:28:45 nice not like I don't know why I just I
1:28:49 don't know why it made that cut
1:28:52 but it's still kind of cool
1:28:55 it's just not what I want to help us
1:28:57 improve the result uh I'm gonna give
1:29:00 that a two
1:29:02 render quality didn't match prompt lack
1:29:04 of movement composition human autonomy
1:29:06 didn't match preview uh not that
1:29:11 why can't I give you my what I I don't
1:29:14 none of your things match my feedback
1:29:19 she
1:29:21 didn't match prompt
1:29:24 all right anyway that's it go play with
1:29:27 that you know similar to all these tools
1:29:29 you can get some really good stuff you
1:29:31 can get
1:29:33 um
1:29:34 some really crappy stuff like that
1:29:36 started out really nice and then it just
1:29:38 did that jump cut to something
1:29:39 completely different so I don't know
1:29:40 that was weird
1:29:42 all right I like the hail Graham shirt
1:29:44 thank you very much
1:29:46 what app is that I don't know what I was
1:29:48 using but it was probably
1:29:53 uh ideogram
1:29:56 ideogram makes very scary chunky looking
1:29:59 kids though I found
1:30:01 yeah
1:30:02 because it's not a very sophisticated
1:30:04 image the image generation part of this
1:30:06 is not super sophisticated it does some
1:30:09 really weird stuff sometimes like I I
1:30:11 had a guy walking in front of like a
1:30:13 really cool wall with graffiti letters
1:30:15 on it the letters were perfect but the
1:30:17 dude walking like his face was all
1:30:19 mangled his feet were weird and yeah
1:30:22 it's a little it's a little weird like
1:30:23 that I'm not on slack oh did I say slack
1:30:26 no I meant Discord
1:30:29 jrc I'm talking Discord
1:30:33 for the for the AI Salon we have a
1:30:35 Discord not a slack they're kind of the
1:30:37 same thing so I get them mixed up
1:30:39 um
1:30:40 they're super General Harvard releases
1:30:42 guidance for AI yep we talked about that
1:30:44 I'll DM here yeah just DM here but but
1:30:49 we can do Discord I need to get on it
1:30:52 anyway yeah car Bruno oh I'm gonna make
1:30:54 I'll make a uh an AI learning lab
1:30:58 channel on the Discord and then I'll
1:31:00 make a roll there and you can only you
1:31:02 can only get in a channel if you got the
1:31:04 roll and you can only get to roll if you
1:31:07 ask your tour guide
1:31:09 am I late to the party hey everybody
1:31:12 Share Bear well that was a while ago
1:31:14 you've been here for a while so I'm
1:31:16 gonna jump to the bottom I'm 95
1:31:17 questions behind so if I haven't been
1:31:19 answering your question that's because
1:31:21 I'm behind
1:31:22 um
1:31:23 all right it's 9 59. I can't stay too
1:31:26 late because I got to deal with my car
1:31:28 [ __ ] in the morning so you get up early
1:31:30 and get that [ __ ] going
1:31:32 um
1:31:38 all right
1:31:42 I think people want to know that their
1:31:44 art had a human involved I don't think
1:31:46 people will care
1:31:48 well it depends on how you define art
1:31:51 I think that I think that authentic
1:31:55 I think that authentic human engagement
1:31:57 the value of that's going to go up
1:31:59 dramatically so I think there's a lot of
1:32:01 stuff that
1:32:05 human involvement
1:32:07 is going to be critical really important
1:32:10 value goes up
1:32:12 if I need to make a shirt like this
1:32:15 I don't give two shits that a human
1:32:17 wasn't involved
1:32:20 um if I need to crank out a bunch of
1:32:21 just images that are going to go and
1:32:23 throw away social media anyway
1:32:26 I I don't really care right
1:32:29 so I think it's
1:32:31 I think it's very context specific
1:32:35 right if you want to do art if you want
1:32:37 to do Cinema Cinema's still going to be
1:32:41 Artisans and Craftsmen and things like
1:32:43 that but if I need to knock out a crappy
1:32:45 video for some corporate commercial
1:32:47 thing does that need to be that
1:32:50 handcrafted it probably doesn't
1:32:54 um
1:32:55 so where those boundaries are that's all
1:32:57 going to be personal choice [ __ ] there's
1:32:59 gonna be there's gonna be a massive
1:33:00 backlash to this generative AI stuff
1:33:04 where people are going to go full analog
1:33:06 they're gonna go [ __ ] computers [ __ ]
1:33:08 this nonsense I'm going full analog
1:33:12 we're we're making films with
1:33:14 hand-cranked film cameras we're gonna
1:33:17 put it in the chemicals and we're gonna
1:33:19 show movies on she tongue on the side of
1:33:22 buildings in local cities and those
1:33:24 things I bet those things are going to
1:33:26 take off in popularity
1:33:30 and then there's going to be all sorts
1:33:32 of super automated [ __ ] that might be
1:33:35 soulless might be brilliant who knows
1:33:38 I think people want to be engaged and
1:33:39 entertained and have an appropriate
1:33:42 experience when they need it if all I
1:33:45 need is to get some information about
1:33:48 something at the right time and I get it
1:33:50 I don't care
1:33:52 if a human made that or if the machine
1:33:54 made that but if I want to have a human
1:33:56 experience then that's
1:33:59 more important than ever right so it's a
1:34:02 spectrum this is not this is it's one of
1:34:05 the dangers right now that everyone is
1:34:07 like everyone's going to the binary
1:34:09 place like we live in a binary Society
1:34:12 right now you know I'm red you're blue
1:34:14 you're evil you're evil man
1:34:18 and now it's like AI good AI evil man
1:34:23 no AI gray lots of grays
1:34:28 Shades of Gray
1:34:30 many
1:34:33 in some context AI brilliant in some
1:34:35 context AI evil
1:34:40 in some context humans really important
1:34:42 in other contexts humans not so
1:34:44 important
1:34:46 and if this can do the shitty work great
1:34:50 let's free us up to do the the more
1:34:52 important work and reconnect with one
1:34:55 another and not be so
1:34:57 at odds
1:34:59 what if that what if AI
1:35:02 cleared a path for us not to suck as
1:35:05 humans
1:35:06 because we weren't over stressed and we
1:35:08 weren't over taxed
1:35:10 and we weren't doing soul-crushing
1:35:12 repetitive work that doesn't require any
1:35:15 of our creative input
1:35:18 that's a possibility
1:35:21 so all that low-level repetitive work
1:35:23 [ __ ] it let the machine do that
1:35:26 and maybe I'll go start that band I
1:35:28 always wanted to start
1:35:30 [ __ ] far out great
1:35:32 and it's going to be an acoustic band
1:35:34 and we're going to play in corn fields
1:35:36 awesome
1:35:37 great
1:35:42 you know that dude with the big red
1:35:43 beard that just had the [ __ ] massive
1:35:45 hit right the the rich rich people from
1:35:48 Rich or
1:35:50 rich people north of Richmond whatever
1:35:51 it was
1:35:53 um
1:35:55 not a classically trained singer not a
1:35:57 classically trained musician doesn't
1:35:59 follow the rules of country music
1:36:03 saying this really roughly recorded song
1:36:06 and it [ __ ] resonated right
1:36:09 became a number one hit out of literally
1:36:12 nowhere right like he's Blown Away by it
1:36:16 what's that about super authentic value
1:36:19 of authenticity I think it goes up
1:36:21 dramatically but I don't think it's a
1:36:23 requirement for art at all
1:36:26 just different contexts and different
1:36:28 definitions of Art
1:36:30 right creative output
1:36:34 some of it's art some of it's just we
1:36:36 need an image here
1:36:38 I don't care if a person does it someone
1:36:41 else does it but it needs to say I'm not
1:36:44 weird I'm irregular and it needs to have
1:36:46 a turtle in it
1:36:48 is that something I would have done no
1:36:51 well mine wouldn't have been that good
1:36:53 for one thing
1:36:54 is that necessarily My Style no if I
1:36:57 spend a bit more time I could get
1:36:58 different styles I probably will that
1:36:59 was just an experiment I got that to
1:37:01 that today and it's kind of fun and
1:37:03 people like it so there it is
1:37:06 but does it matter
1:37:07 so here's what matters
1:37:09 I prompted it
1:37:11 I [ __ ] around with it long enough to
1:37:13 have one that I'm like yeah that's kind
1:37:15 of fun
1:37:16 let me share that and I shared it people
1:37:18 were like oh this should be a t-shirt
1:37:19 let's do this
1:37:21 all right cool it resonated awesome
1:37:22 great far out
1:37:24 so the human intervention was the
1:37:26 curation
1:37:28 so I was involved
1:37:30 see
1:37:31 I made that
1:37:33 did I actually paint the pixels no does
1:37:36 it matter I don't think so
1:37:44 like people asking if I'm qualified
1:37:46 qualified to be an idiot on Tick Tock I
1:37:49 think so
1:37:50 [Laughter]
1:37:53 pretty sure I am
1:37:56 language engineering yes this is
1:37:58 language engineering we are in a we are
1:38:00 in a new phase where language is the new
1:38:03 programming
1:38:05 anyone can change some tensors together
1:38:07 and call it AI it's true they can
1:38:12 and a lot of people are
1:38:14 am I late you're a little late but you
1:38:17 know I'm still tooling around you've
1:38:19 been here for a while now
1:38:20 um but what exactly is the personal use
1:38:23 for the Cucumber
1:38:27 what but well whatever do you mean dear
1:38:31 Ted
1:38:33 you use a cucumber for salads
1:38:36 I think it is patently obvious why I
1:38:40 would personally use a cucumber
1:38:43 no I don't know what you're considering
1:38:46 how can they stop them though it's on
1:38:48 the web
1:38:51 well yeah so so that's that's in
1:38:54 reference to
1:38:56 um
1:38:58 the large language models not being
1:39:01 trained on media companies so open AI
1:39:05 part of the bot that they created has a
1:39:09 has an option that you can turn it off
1:39:10 you can turn it off for your domain
1:39:13 so
1:39:14 even though New York Times articles
1:39:16 likely live in places other than the New
1:39:19 York Times like
1:39:21 they're on some library site somewhere
1:39:23 and they're in places elsewhere
1:39:26 they're still the copywritten work of
1:39:28 the New York Times so if the New York
1:39:31 Times says open AI cannot train on New
1:39:33 York Times data
1:39:36 then
1:39:37 at the point in the future where the
1:39:39 these large language models can start to
1:39:42 do citations
1:39:44 um they won't be able to use New York
1:39:45 Times citations those will be out of the
1:39:47 mix right so so it's not
1:39:50 Canon Canon will it find something that
1:39:53 was in the New York Times I'm sure it
1:39:55 will crawl a bunch of [ __ ] where where
1:39:57 there's all sorts of blurry lines but
1:39:59 that but it won't be officially in there
1:40:03 and that's that's the difference because
1:40:05 it is going to be valuable to have a
1:40:08 large language model that you can say
1:40:11 I want you know I want to write uh an
1:40:15 article about whatever the news topic is
1:40:17 and I want my sources to be trusted news
1:40:20 sources from the following list
1:40:23 and if Google can do that and open AI
1:40:26 can't
1:40:27 more people are going to use Google for
1:40:29 that particular task right so so this
1:40:32 starts to become very important there
1:40:33 was a in the in the early days when
1:40:36 Edison invented the cylindrical
1:40:40 record player this you know cylindrical
1:40:43 wax record player and the other dude did
1:40:45 the platter
1:40:47 the other dude
1:40:48 cut deals with the most famous
1:40:51 performers of the day
1:40:53 and Edison was like this is technically
1:40:56 Superior and I'm just gonna get you know
1:40:59 an opera singer not the opera singer
1:41:04 the opera singer was the one everyone
1:41:07 wanted to listen to so they listened to
1:41:09 the flat one more than the cylindrical
1:41:12 one even though technically the
1:41:13 cylindrical one was better
1:41:17 I would also argue that the cylindrical
1:41:19 one was a pain in the ass form factor
1:41:21 but
1:41:23 you know a giant platter ain't exactly
1:41:26 convenient either
1:41:28 uh uh uh I used to think that Walter
1:41:31 Cronkite I used to think Walter Cronkite
1:41:33 Sergeant shorts and Chef Boyardee were
1:41:36 one human I was confused that's awesome
1:41:40 can I be an irregular in The Irregular
1:41:43 Channel yes Firefly just when you join
1:41:45 the Discord over there
1:41:47 um so
1:41:51 I don't know how we're gonna do this
1:41:54 maybe just put so if you go to the AI
1:41:57 Salon Discord just put like a i l l in
1:42:01 parentheses in your name or just use
1:42:02 make sure you use the same name you use
1:42:04 here on tick tock and if I recognize
1:42:06 people I'll assign you the role
1:42:11 um that's why I'm always Vicki very nice
1:42:13 uh I like the image you generated of
1:42:16 Elton and Zuckerberg boxing and kissing
1:42:18 I didn't make that one but that one was
1:42:19 awesome it was hilarious we need an
1:42:21 irregular secret handshake and passcode
1:42:23 we do we do all your data are belonged
1:42:27 to Google it really is I mean
1:42:30 you know I think one of the reasons that
1:42:33 um
1:42:34 the the rumor of why Gemini is going to
1:42:37 be so good is that it was trained on
1:42:40 YouTube data
1:42:41 they own all the YouTube data they got
1:42:44 all the videos so not only do they have
1:42:47 internet data they have all the YouTube
1:42:48 data plus they have all of our metadata
1:42:50 from all the [ __ ] we've ever done and
1:42:52 everything we've ever searched
1:42:54 so and everything we've ever clicked on
1:42:56 so like the the
1:43:01 it is not beyond the realm of
1:43:02 possibility that that Google's next
1:43:06 version of their large language model is
1:43:08 going to destroy uh gpt4
1:43:13 but it's also very possible it comes out
1:43:15 and it's a janky piece of [ __ ] because
1:43:17 that's the stage we're in you
1:43:20 know
1:43:22 all right let's see pirate Irregulars
1:43:25 irregular laughs oh matey ah yeah Runway
1:43:29 ml it's great it's great Italians great
1:43:32 I tell ya like today's youth will never
1:43:34 know the agony of dial up internet
1:43:36 indeed they will not know they do not
1:43:41 appreciate the pain that we've gone
1:43:44 through for them
1:43:46 they do not know the pain
1:43:49 of listening to this
1:43:58 [Music]
1:44:03 if you're too young to know what that
1:44:05 sound is [ __ ] you
1:44:10 that was our internet
1:44:13 our internet had to call up other in
1:44:15 other computers on the internet
1:44:18 and that's what it sounded like
1:44:20 and then the text with like type across
1:44:23 the screen so slow that you could watch
1:44:24 it kind of like Chachi PT is now right
1:44:28 you type in a thing and you're like yeah
1:44:29 it's remarkable but couldn't it be
1:44:31 faster yeah it's going to be faster it's
1:44:34 going to be like poof
1:44:37 yeah those were the days those were the
1:44:40 Glory Days when the internet was a
1:44:43 struggle
1:44:45 you couldn't just click on [ __ ]
1:44:49 first word problems yeah I know we're in
1:44:52 the middle of many first world problems
1:44:54 right now turn down your movement
1:44:55 setting on Runway
1:44:57 oh
1:44:59 that's a good idea I could I could have
1:45:02 thought of that duh everybody knows that
1:45:05 all right I'm gonna go
1:45:08 motion
1:45:15 I don't want to fill out the survey I
1:45:17 want to change my motion setting all
1:45:19 right I'll put it to four
1:45:22 five four all right four
1:45:25 we'll say generate
1:45:27 all right so that's gonna do its thing
1:45:29 I'll come back to that that's a good
1:45:31 idea thank you toast post
1:45:33 and Toes post 8.5 out of 10 on the name
1:45:35 I like it it's solid thanks for walking
1:45:37 us through that absolutely uh Kyle
1:45:40 and Silver Fox oh thank you Kyle and
1:45:43 Silver Fox you're welcome thank you
1:45:45 that was Runway yes don't mix Turtles my
1:45:48 dad worked at Bell Labs I saw the early
1:45:50 dial-ups when they had two cups to
1:45:51 connect the phone exactly
1:45:54 yeah the the two cups and the and the
1:45:57 phone and the exactly
1:45:59 exactly exactly Lunatic Fringe PBS I was
1:46:03 on Echo Echo in New York was how I got
1:46:06 my start on the internet You've Got Mail
1:46:10 uh my lights are stuck on 36k what the
1:46:13 hell
1:46:15 oh
1:46:17 that's weird
1:46:21 yeah I don't um this account is weirdly
1:46:25 Shadow banned right now like like I'm my
1:46:29 um my videos that I've been putting up
1:46:30 they used to get anywhere between two
1:46:33 and six thousand views and now they hit
1:46:35 like two or three hundred and they're
1:46:37 like they're going really fast they get
1:46:39 to two or three hundred and then they
1:46:40 just stop and it's like 10 views an hour
1:46:44 just really slowly so something's going
1:46:48 on I don't know maybe it had something
1:46:49 to do with the fact that there was a
1:46:50 scammer account and we're in some sort
1:46:51 of seven day Shadow ban [ __ ] thing or
1:46:56 they're like until he gets his hair
1:46:58 worked out we're not gonna promote this
1:47:00 this is this is disturbing to the
1:47:02 children
1:47:09 oh dear that sound makes me crying my
1:47:12 teeth
1:47:14 hahaha
1:47:17 You've Got Mail 56k yeah exactly
1:47:20 remember getting a 56k when you had a
1:47:23 14-4 you were like I'm baller baby
1:47:28 that was cool I remember when we got our
1:47:31 first T1 line at the agency.com offices
1:47:34 that was
1:47:35 when we took our internet company from
1:47:38 dial-up modems throughout the office to
1:47:41 uh
1:47:42 T1 line an Ethernet run to every desk
1:47:46 is like the future
1:47:48 it's like the future in those little
1:47:50 wires you just put them little wires
1:47:53 there it's like a foam wire but it was
1:47:55 magic
1:47:56 ring internet everywhere always on
1:48:01 you youngsters are like well yeah the
1:48:02 internet's always been on no it wasn't
1:48:06 it wasn't
1:48:07 you had to call it
1:48:10 hello Internet could you give me some
1:48:12 data uh yes I can it's going to be
1:48:14 boring does that work for you uh yeah
1:48:17 could you give me a lot of it yes I will
1:48:19 it'll be confusing and unformatted okay
1:48:24 that was the internet
1:48:27 none of this fancy as videos and
1:48:30 pictures and chats and FaceTime and
1:48:34 goggles and AR
1:48:36 no
1:48:38 we we had an underlying blue word and
1:48:40 you clicked on it and it took you to
1:48:42 another page and we liked it we thought
1:48:44 it was remarkable yeah
1:48:47 you kids with your multi and media put
1:48:51 together as a single entity
1:48:55 just nothing like that
1:48:58 trivial what you do now so easy
1:49:02 you didn't have to work for your
1:49:03 knowledge
1:49:11 and you are educating everyone in here
1:49:15 in different as to what is current and
1:49:18 AI
1:49:19 yes yes I am trying to educate everyone
1:49:23 I don't know I you know part of its
1:49:26 education part of it's a little soapbox
1:49:28 part of it's I think a lot of what I
1:49:31 talk about hopefully is common sense
1:49:32 right
1:49:34 what I'm trying not to do is pass
1:49:37 judgment on technology because the
1:49:39 technology is just going to do what the
1:49:40 technology is going to do we're the ones
1:49:42 that [ __ ] everything up because we're
1:49:45 afraid of change and and all that sort
1:49:47 of stuff the technology is going to keep
1:49:49 marching so where I try to live is okay
1:49:51 five years from now if this keeps
1:49:53 marching at this pace what does the
1:49:55 world look like and then the only thing
1:49:57 you can think is
1:49:59 holy [ __ ]
1:50:01 the world's gonna be completely
1:50:03 different
1:50:06 we can't imagine it
1:50:09 how do I prepare get curious go
1:50:13 to
1:50:16 here
1:50:17 if you haven't gone to Chachi PT go
1:50:19 there right now go go go go go go go go
1:50:21 there right now and if you don't know
1:50:24 what to say it say to it go there
1:50:27 prompts.chat that'll tell you how to
1:50:29 talk to that
1:50:34 eh
1:50:35 eh
1:50:36 um
1:50:39 you know what I mean
1:50:43 listen Jackie
1:50:45 you're not gonna have to have a word
1:50:48 about this internet thing
1:50:51 I like my modem
1:50:53 I like dialing up to the internet I
1:50:55 don't like it always to be on Jackie
1:50:59 okay
1:51:06 a guy just has an electrical meltdown
1:51:09 all right folks I'm gonna do a few more
1:51:11 here and then I'm gonna go just turn 59
1:51:14 12 minutes ago Ted looks at life happy
1:51:17 birthday sir
1:51:18 I'll be 59 in May so congratulations
1:51:22 that's a big well it's not a big one at
1:51:25 our age like the decades are good at our
1:51:27 age you know we'll have a party on the
1:51:29 decade so 59 you're getting there but
1:51:32 next year will be the big one but
1:51:34 congratulations happy birthday uh my
1:51:36 legs are stuck and I'm 36k but she share
1:51:39 bear it's up to 40K so I don't know if
1:51:41 you saw that
1:51:44 lost you
1:51:46 you lost me
1:51:48 my ex-boyfriend did the same 400 AOL
1:51:51 bill in one month
1:51:52 my son ran up a 400 AOL bill in one
1:51:56 month and we were poor back then I
1:51:58 almost killed him I bet good Lord
1:52:01 yeah oh right because it was per minute
1:52:05 you get the little CD-ROM with like 5
1:52:08 000 minutes and then you'd burn through
1:52:10 you didn't realize five thousand minutes
1:52:12 was was like three sessions and you're
1:52:15 like oh I gotta conserve my minutes oh
1:52:17 let me just talk to this person oh [ __ ]
1:52:19 it's four hours later just like these
1:52:21 lives
1:52:23 oh my God all right everyone uh listen
1:52:27 had a good time tonight uh nice hanging
1:52:30 out with you sorry for the hair it's
1:52:31 just at some point I'll get it cut here
1:52:34 um and sorry for the shirt for those
1:52:36 that uh made you cross-eyed what are you
1:52:37 gonna do uh what date May May 7th
1:52:41 um
1:52:41 I still have no G AOL address oh if you
1:52:44 still have an AOL email address that's
1:52:47 it's kind of cool now like not
1:52:50 necessarily but it's rare and uh it's
1:52:55 kind of cool it's kind of cool it's very
1:52:57 nostalgic so uh I would hang on to that
1:52:59 could be worse than something I'll sell
1:53:02 you my AOL account for five thousand
1:53:04 dollars
1:53:07 oh good lord great stream fun thank you
1:53:10 thank you very much everybody
1:53:12 um Hardy went from Bell labs to the
1:53:14 Pentagon my dad that is oh nice uh
1:53:17 sencom
1:53:18 designed
1:53:20 badge of honor
1:53:23 peace
1:53:24 see dude see a Stevo felonious thank you
1:53:27 to bias Ted happy birthday cabruno thank
1:53:31 you thank you thank you
1:53:33 Share Bear great shirt peace out I
1:53:35 appreciate it vertical stripes vertical
1:53:38 stripes oh wait and let's let's go out
1:53:41 on a bang
1:53:43 I'm not weird
1:53:45 I'm irregular
1:53:48 not everybody
1:53:50 thank you for your time