Dec 6, 2023 - (2 of 2) How Gemini AI Will Transform Your Life and Work

Description

Chapters

Transcript

0:02 [Music]
0:10 [Music]
0:27 [Music]
0:29 w
0:31 Freedom came my way that
0:34 [Music]
0:37 night
0:39 jet in and
0:42 out I
0:45 Wasing miles an hour wondering how hard
0:49 I'd
0:50 [Music]
0:52 hit when they came into the
0:59 St he said I was bad Beyond
1:04 repairs but I got no problems with
1:12 myu I
1:14 [Music]
1:19 am all right you done you
1:24 done Gemini what the
1:29 quack hello everybody hello welcome to
1:33 the air learning lab tonight we are
1:36 going to be learning about the lab we're
1:37 going to be singing with doggies we're
1:40 going to play with blue rubber duckies
1:43 and we're going to learn what the hell
1:45 is going on people get your brains ready
1:47 get your brains ready to get screwed in
1:50 you can make money oh gemini or Gemini
1:53 I'm going to need to make a new
1:55 button I got to make a new
1:59 button good Lord people Lauren that's
2:03 your image that is your image you made a
2:06 good ducky I thought that was an awesome
2:08 an awesome image for the
2:11 night Gemini what the
2:15 quack um I also just learned that Runway
2:18 ml now has
2:21 uh
2:23 Styles and one of them is 3D 3D
2:27 animation and it's pretty slick what did
2:30 I just do did I just quit that
2:38 no let me see let me got
2:42 you view the latest oh that's not a blue
2:46 rubber ducky that's a yellow rubber
2:49 ducky oh that's so
2:52 bad other people have good luck I've got
2:55 bad luck at Runway I do not know how to
2:59 prompt Runway ml I just
3:01 don't all right
3:03 everybody welcome welcome welcome this
3:06 is the AI learning lab we're going to be
3:07 talking about Gemini tonight most of the
3:10 night um there's not a ton to demo but
3:14 there is a lot to
3:16 learn um
3:20 so here's what you need to
3:23 know uh if you were not paying attention
3:26 this afternoon you missed me going live
3:29 ha I went live at like one o'clock or
3:32 something like
3:33 that uh because I just had to get my
3:35 head around this [ __ ] and I thought well
3:37 why don't I just do it live and uh it
3:40 was pretty face melting it was pretty
3:41 face melting so we'll go through all
3:43 that stuff
3:44 tonight there was a rumor that Google
3:47 was going to release Gemini this
3:49 week uh then it looked like it got
3:52 pushed to next year then they said they
3:55 were going to demo it this week and then
3:58 today they just announced it now it's
4:01 partially live but it's you'll see
4:04 there's there's kind of three pieces of
4:07 it um so some of it's live
4:10 today December 13th there'll be another
4:13 part of it live for developers wait
4:15 won't be long oh oh won't be on long
4:19 probably but hey look it's Gemini yeah
4:21 exactly yeah what do you think we're
4:23 going to talk about
4:25 Pate uh so it looks like developers are
4:27 going to get there shot at the at the
4:30 API on December
4:32 13th and then it looks like the the big
4:35 sexy model so we kind of discovered
4:37 there's there's two different models
4:39 there's a model that's more like GPT 4
4:42 and then there's a model that's like
4:44 this new bitching crazy ass thing that's
4:47 probably not coming till the new
4:49 year but we can prepare and we need to
4:53 prepare because this is it's just it's
4:56 just a different thing it's just a
4:58 different mod mod
5:00 ity uh the AI race feels like it's Free
5:03 Falling or free accelerating you know
5:07 you know reaching escape
5:09 velocity um I think it's more like
5:12 escape velocity than freef
5:21 [Music]
5:26 falling all right so let's get this
5:30 party started let's get this party
5:33 started yeah let's get this party
5:35 started yeah mhm
5:39 yeah all
5:41 right so incredibly proud of the Google
5:44 team right now so proud I canceled my
5:47 grock I don't have access to my Gro did
5:50 you have the
5:53 gro I don't think I have access to it I
5:56 must not be paying the right $16 a month
5:58 Ola kyin learning lab peeps what's up
6:01 worker B good to see
6:05 you this AI race feels like it's Free
6:08 Falling get some comments going
6:14 here oh so far away yeah I know right
6:18 exactly wait it's a month away I can't
6:20 have completely revolutionary technology
6:23 just tomorrow I want it now i' like it
6:25 today what are you going to do with it I
6:27 don't know I'll probably make a picture
6:28 of a duck
6:32 cim katkin I've been playing with it for
6:34 weeks was not 100% sure what day it
6:38 would come out oh p p has definitely
6:40 been holding back on us we know
6:43 that we know
6:45 that when whenever a googler says
6:48 something like well I've heard some
6:50 things it it it sounds interesting you
6:53 know there's some big [ __ ] coming I
6:56 didn't know it was this big it's it's
6:58 quite impressive please tell I've been
7:00 out of the loop for a few days turquoise
7:02 dreams show us the goodies I
7:05 will you wait I I'm trying to parse this
7:10 statement I've been out of the loop for
7:12 a few days you there's no being out of
7:16 the loop in
7:20 AI there's no crying in
7:25 baseball you can't miss you can't miss
7:28 it 12 hours like you shouldn't be
7:32 sleeping you should not be
7:34 sleeping I don't I what I thought I
7:37 thought you were
7:40 professionals good
7:42 Lord I may have overestimated you all of
7:45 you this
7:48 is I thought we had an agreement here
7:51 this is all of our Lives we give up
7:53 everything for this all right whatever
7:55 all right I had to deal with my family
7:57 all right if you've got needs that's
7:59 fine no yeah listen hey you got to set
8:02 your own priorities right all right fine
8:04 yeah yeah well whatever
8:08 yeah yeah I basically had a meltdown
8:10 today because I just like I I
8:12 just I can't what did I whine earlier I
8:17 was like I don't want to recontextualize
8:23 reality that's what we're
8:25 doing we thought we knew what AI was we
8:28 don't know we have no idea what AI is I
8:32 can tell you that now with
8:35 certainty oh my God all right let's get
8:38 rolling let's do this let me let me do a
8:41 little uh well normally I start people
8:43 out with with go check out chat
8:47 EBT nah not tonight we should just go
8:50 look at we should just go look at this
8:53 [ __ ] trying to think where I want to
8:56 where I want to
8:58 start I don't work
9:00 on Gemini I work kind of Gemini adjacent
9:04 yeah but I got to say Pate
9:07 um I think
9:10 that if the work you're doing on tpus is
9:13 an indication of like the per like the
9:16 per well we don't know the performance
9:19 yet I know you said it's it's wicked
9:20 fast from what you've seen but the the
9:22 videos I was showing today it looks like
9:24 a lot of those were edited for Speed to
9:26 make it look faster so it'll be it'll be
9:28 interesting to see but assuming these
9:30 are pretty Snappy things I think it's
9:31 going to be a really good ad for the TPU
9:33 BS
9:37 business cuz Nvidia and the gpu's been
9:39 getting all of the noise Dr J I've said
9:42 it before if you can crowdfund my Google
9:45 salary there you go I'll tell you
9:48 things yeah that's the thing about
9:50 that's the thing about uh when you work
9:52 at a place like the goog you can't you
9:54 can't leak [ __ ] you're out of there
9:55 they will they will walk your ass to the
9:57 door and you won't have Gmail access by
10:00 the time you get to your car all right
10:03 let's see perplexity looks interesting
10:06 too I don't think they announced
10:08 anything new please tell me they didn't
10:09 announce anything
10:10 new um all right let's see what we got
10:15 here got anything
10:20 new that's going on a t-shirt which
10:24 one the Gemini what the
10:26 quack or oh the oh
10:30 I I I don't feel like recontextualizing
10:33 reality that one that's a good
10:37 t-shirt ai ai learning La that should
10:41 actually be a a AI learning La one with
10:44 a whining baby face I don't I know or
10:49 a I don't want to recontextualize
10:53 reality um all right well let's go look
10:55 at what they've done I heard they have
10:57 they went live with web Act web access
11:01 perplexity perplexity always had web oh
11:03 oh it's connected to the internet ah we
11:06 can go look at that too so I figure
11:08 we'll go look at runway's new 3D model
11:11 it's new to me
11:14 anyway Gemini Pro is the model driving
11:17 Bard right now and it's pretty fast
11:19 pretty fast but like I the couple of
11:21 things I tried to do on it it was like
11:22 no I can't do that I'm just the large
11:24 language model I'm like you did it 3
11:26 days ago when you had Palm two up your
11:28 butt and now the you got fancy Gemini
11:30 you're not going to summarize a YouTube
11:32 video for me kind of snobby nonsense is
11:38 that oh I sent the developers feedback I
11:41 said this is not
11:44 acceptable I got all cared
11:48 out am I late were you here this
11:51 afternoon Emilio's wife did you see the
11:53 special the special
11:55 live did you see the special live where
11:58 we went live this afternoon this
12:02 afternoon yeah exactly not only you're
12:05 late you missed
12:08 class yeah exactly we did meet without
12:13 you yet another demerit on Amelio's
12:16 wife's credit or uh report
12:20 card it's just it's it's nothing but red
12:23 tick marks tick dick dick L like it's a
12:26 little l in the shape of a check mark l
12:29 l l for late yeah I guess you'll have to
12:34 watch it on the record all right but it
12:38 was pretty cool stuff so we'll do it
12:40 again we'll do it again we'll do it
12:43 again for Amelio's wife I wasn't going
12:44 to do it again but now I guess we're
12:46 doing it again all
12:50 right a human being what's happening
12:54 what's happening good
12:57 people introducing Gemini our latest and
13:00 most capable model
13:02 yes Luke I am your
13:05 father so it's interesting there's two
13:08 different versions of this release this
13:09 is I think just the
13:11 blog and then this
13:13 is the exact same content but like in a
13:18 fancier
13:20 sexier uh presentation so so we'll go
13:25 with we'll go the blog one because I
13:26 assume it's more
13:28 complete
13:30 uh cuno shared the live thank you for
13:31 sharing the live yeah if you're if
13:33 you're new here so welcome my name is
13:34 Kyle Shannon this is the AI learning lab
13:36 this will be a bit of a special a
13:38 special uh what do you call these live
13:41 it's a special
13:43 live uh because we're going to be just
13:45 going over all this Gemini stuff we're
13:46 going to be you know we're going to be
13:48 doing we're going to be watching some
13:49 videos together it's it's going to be
13:51 like uh it's going to be like class so
13:54 get your notebooks
13:55 out get your AI assistants out and have
13:58 them take notes in real time while I
14:01 talk actually you know what you could
14:03 probably do that you could probably do a
14:07 little a little voice to text notetaker
14:11 thing next to this live and just take
14:13 notes while I'm I'm talking but I don't
14:16 there's nothing that I say did you get a
14:18 nap today I got a little bit of a nap
14:20 but you know family so they're like yeah
14:23 you need a nap and then they just asked
14:25 me [ __ ] for the entire nap so so you
14:29 know special presentation this this will
14:32 be
14:33 special I need Gemini to get my Iron Man
14:36 suit out
14:38 it's seriously it was like this
14:40 afternoon it was like holy [ __ ] like
14:43 honest to God I it's so funny I had if
14:47 you were here this afternoon I said this
14:49 this
14:50 afternoon yesterday when I went live I
14:52 went live after the AI Salon
14:55 anniversary and
14:57 um I just just had this nagging feeling
15:00 of just like I'm really like I'm really
15:04 clueless I'm really behind I don't I
15:07 don't think any of the [ __ ] that I
15:08 learned how to do this year is
15:13 is is relevant anymore and it's like do
15:16 I like want to relearn a whole new
15:19 [ __ ] and then this came out today and
15:23 some of the stuff you'll see when I show
15:24 you some of this stuff it's just like
15:27 what so any anyway okay let's get let's
15:31 get rolling here Bard answers questions
15:32 about Gemini and alluded to pricing what
15:35 did it allude
15:38 to is it 30 bucks a month like
15:40 everything else Pate do you have any do
15:43 you have any uh any insight on
15:46 pricing or range or is there I assume I
15:50 assume uh Gemini Pro is going to be free
15:53 and Gemini Ultra is going to cost money
15:56 like 3.5 and so I think I think Gemini
15:59 Pro is like the 3.5 competitor and then
16:03 Gemini Ultra is the the four it's it's
16:07 different than four it's
16:09 different all right so this announcement
16:12 was announced by um by Sundar the CEO
16:16 and Demis of deepmind CEO of Deep Mind
16:20 um so they're the two they're the two
16:22 biggies so they're they're both uh
16:25 they're both Penning Penning this this
16:28 announcement
16:30 so it's big deal big big
16:32 deal not GNA die all right so we'll just
16:36 watch this is a marketing video the
16:38 there's some there's some good
16:39 interesting stuff in here so we'll we'll
16:43 uh let me see how I going to do this
16:46 right there we go that's
16:57 better
17:01 so I'll I'll stop when there's when
17:02 there's interesting stuff um but this is
17:05 just sort of like the Slick marketing
17:07 we've done it
17:10 boys you know one of the reasons we got
17:12 interested in AI from the very beginning
17:15 is that we always viewed our mission as
17:17 a Timeless Mission it's to organize the
17:20 world's information and make it
17:21 universally accessible and useful but as
17:26 information is grown in scale and
17:28 complexity you know the problem has
17:30 gotten harder so we always knew we
17:32 needed to have a deeper breakthrough to
17:35 make
17:37 progress I've worked on AI my whole life
17:40 because by the way this this uh blog
17:42 post if you go to blog. Google the do
17:46 Google is instead of the Doom blog.
17:49 gooogle I've always felt would be the
17:52 most beneficial and consequential
17:55 Technology For Humanity human beings in
17:57 our society would have five sensors and
18:00 the world we built and the media we
18:01 consume is in those uh different
18:04 modalities so super proud and excited to
18:06 announce the launch of the Gemini era a
18:09 first step towards a truly Universal AI
18:11 model the Gemini approach to
18:13 multimodality is all the kinds of things
18:15 you want uh an artificial intelligence
18:18 system to be able to do and these are
18:20 capabilities that haven't really existed
18:22 in computers before traditionally
18:25 multimodal models are created by
18:27 stitching together text only Vision only
18:30 and audio only models in a suboptimal
18:33 way so that's that's important right so
18:37 here's chat GPT here's um Mid Journey
18:41 for images here's uh whatever uh you
18:45 know suo for audio right they're they're
18:48 these distinct models that's that's the
18:51 GPT 4 model but not the 4.5
18:55 model the secondary stage Gemini is the
18:58 most multimodel from the ground up so
19:00 you can seamlessly have a
19:03 conversation Vicky it's not included in
19:05 Google workspaces chle that's just mean
19:09 so oh by the way Pate uh if there is a
19:12 uh suggestion box at the Google um what
19:15 you can tell them is people that pay for
19:18 your service should get access to the
19:20 new
19:21 [ __ ] all
19:23 right every time I try to go use some
19:26 new Google thing it's like oh you're
19:28 subscri rtion doesn't support that I'm
19:30 like what what do you mean the paid
19:33 subscription cross modalities and give
19:35 you the best possible response Gemini
19:39 look see all four all four that's going
19:42 to we'll see that shortly lar just the
19:45 most capable model it means that Gemini
19:47 can understand the world around us in
19:50 the way that we do uh and absorb any
19:52 type of input and output so not just
19:55 text like most models but also code
19:58 audio image and video what's amazing
20:01 about Gemini is that it's so good at so
20:03 many things as we started getting to the
20:05 end of the training uh we started seeing
20:08 that Gemini was better than any other
20:09 model out there on these very very benks
20:13 each so good at so many things as we
20:15 started getting to the end of the
20:17 training uh we started seeing that
20:19 Gemini was better than any other model
20:21 out
20:22 there that's gp4 on this this test
20:25 called uh m
20:29 MML I
20:32 think on these very very important
20:34 benchmarks for example each and that's
20:36 Gemini Ultra and
20:38 89.8% over here to the left is the human
20:43 expert MML U
20:45 test it's a test on like 57 different
20:49 disciplines and uh so
20:52 90% beat humans expert humans of the 50
20:57 different subject areas that we tested
20:59 on um it's as good as the best expert
21:02 humans in those areas it's very rare
21:05 that you can work on a technology at a
21:07 foundational level and it simultaneously
21:10 can impact all our products we created a
21:13 family of models that can run on
21:15 everything from mobile devices to Data
21:17 Centers Each of which is actually Best
21:20 in Class Gemini will be available in
21:22 three sizes Gemini Ultra our most
21:24 capable and largest model for highly
21:26 complex tasks Gemini Pro our best
21:28 performing model for a broad range of
21:30 tasks and Gemini Nano are most efficient
21:32 model for on device tasks okay
21:36 so Pro is available now at b.google.r
21:42 is not going to be available until the
21:44 new year Nano is now available on device
21:49 on Pixel 8
21:51 Pros um which is a which is a big a big
21:55 deal right it's the first of these
21:57 models to run locally on a
22:01 phone and what that means is you don't
22:04 have to be connected to the internet to
22:06 essentially have like a chat GPT 3.5
22:09 level class thing I need to run feel
22:12 free to throw questions in Mighty
22:14 networks if there's things I might be
22:16 able to answer thanks Pate appreciate
22:18 you
22:19 sir we want to provide the best
22:22 foundational building blocks and then we
22:24 know um developers and Enterprise
22:27 customers are are going to figure out
22:29 really creative ways to further refine
22:32 our Gemini foundational models and the
22:35 potentials almost
22:37 Limitless so at Google there's this
22:39 healthy disregard for the impossible and
22:42 that has oriented us to be both bold and
22:44 responsible together as these systems
22:47 become more capable all of those
22:49 capabilities also raise new questions we
22:51 have to think about what it means to
22:53 have an image be a part of for example
22:55 the input because an image might be inoc
22:58 on its own or text might be innocuous on
23:00 its own right she talks about safety and
23:02 and uh things there um they've got
23:05 they've got whole videos on that so
23:07 here's here's those three those three
23:09 models Ultra Pro and
23:11 Nano um which is smart right so have as
23:17 powerful a model as possible that can
23:19 run on your top phone Hardware right by
23:25 the way apple is going to absolutely
23:28 kickass at this so apple apple is you
23:31 know everyone's like where's Apple
23:33 where's Apple where's Apple Apple's
23:34 waiting Apple's building the hardware
23:36 Apple's got the whole stack Apple's got
23:38 the Privacy Apple's got the data
23:41 Integrity um where where Google doesn't
23:44 right now so um so this is this is a
23:47 first step toward this this world but
23:50 it's a big deal and then you'll see when
23:53 we start looking at some of these videos
23:55 what Ultra can do and then pro pro I
23:57 think is th this seems to be sort of the
24:00 equivalent of of GPT 3.5 or maybe maybe
24:03 an early version of GPT 4 but I don't
24:07 know um state-of-the-art performance now
24:10 so here's some of the numbers um let me
24:12 see if I can make these big enough you
24:14 can actually see
24:15 them what's going on
24:27 oh
24:28 and you still can't see
24:30 him
24:32 [ __ ] so
24:35 um so what they've got in this chart in
24:38 this thing is um
24:44 the where is
24:48 it here we go the
24:53 MML representation of questions in 57
24:56 subjects including stem Humanities and
24:59 others this is where gp4 got
25:03 86.4 Gemini Ultra got 90% so it's only
25:08 their most powerful model the one that
25:11 we don't have access to yet the one
25:13 that's not coming out
25:15 until the new year although December
25:18 13th developers are going to get access
25:21 to the apis I don't know which models
25:23 they're going to get access to all of
25:25 them or just um Pro and Nano I that I
25:28 don't know um but if you go down here
25:31 there's only one test that gp4 did
25:35 better uh in all these benchmarks now
25:39 this blog post is from Google so they
25:42 are going to pick the things where it
25:44 did the best right so this is don't
25:46 don't forget this is all marketing [ __ ]
25:48 um so so look at all this with the with
25:51 the healthy healthy you know grain of
25:54 skepticism here until we have this thing
25:56 in our sweaty little hands
25:59 um we won't know but it looks very very
26:01 promising right it's it's basically
26:03 beating um gp4 at image analysis at
26:08 video at audio all that sort of
26:11 stuff
26:12 um trying to think what else was in
26:16 this yeah it's got sophisticated
26:18 reasoning which you'll see I'll show you
26:20 that video that video is pretty
26:22 wild we'll show you that video Advanced
26:25 coding is crazy okay
26:29 all right so let's go let's go to the
26:32 videos so if you go
26:35 to
26:37 YouTube and go to playlists um go to go
26:41 to the Google YouTube channel and then
26:45 go to playlist the first three playlists
26:47 are
26:48 all
26:50 um Gemini playlists so they they put out
26:53 a lot of content for this oops you know
26:56 one of them so we're going to start
26:58 start
27:02 [Music]
27:05 with view full playlist
27:09 okay so we're going to start with
27:11 handson with
27:14 Gemini this
27:19 one and hang on let me pause that let me
27:22 flip
27:23 over to this let me put this up here so
27:29 um while we're watching this video feel
27:32 free to go check out the AI salon so
27:35 this channel is is connected to this
27:39 thing called the AI Salon that I started
27:42 a year ago tonight actually last night
27:45 was our anniversary party um our first
27:47 meeting was a year ago tonight um and it
27:51 is a really cool group a lot of the
27:53 Irregulars the people that are here on a
27:56 regular basis our uh our members over
28:00 there so go check that
28:02 out and then okay so we'll watch this
28:05 I'll do I'll do didn't Sony make beta
28:07 Max Hey listen I don't think that there
28:10 is any uh I don't think that there is
28:12 any you'll get no argument from me that
28:15 that
28:17 um Google may or may not do well in this
28:20 game even though they invented this [ __ ]
28:22 they invented the Transformer so um but
28:25 you'll when you see this Tech is that
28:28 the old URL that's the old URL except I
28:30 connected it to uh I I completely redid
28:33 the link tree so the link tree now is
28:36 Just Three Links it's the link to the
28:37 mighty networks so the first link you'll
28:39 see there is to the mighty networks
28:41 Community the second one is to meet up
28:43 so you can just sign up for that but you
28:45 don't have to cuz we have calendars at
28:47 at the community and then there's a
28:49 merch store so if you want to support
28:51 the salon there's a uh there's a
28:53 one-year anniversary shirt that's only
28:55 going to be there for a few days um
28:58 which is super cool um so so yeah you
29:00 can just go there I'm an irregular
29:02 you're you're a regular irregular
29:04 exactly lunatick you are
29:07 okay so let's watch this I'll I'll stop
29:10 where I think things are cool there's
29:12 there's a fair amount of cool stuff
29:15 here oh the other caveat here this
29:19 is there are big parts of this that are
29:22 definitely sped
29:23 up so I have no idea about the
29:26 responsiveness of
29:28 of
29:29 Gemini um so again this is a marketing
29:32 video so take it with a grain of salt
29:34 but they are demoing actual technology
29:37 here all right testing Gemini here we go
29:41 tell me what you see I see you placing a
29:44 piece of paper on the
29:46 table I see a squiggly
29:49 line what about now the contour lines
29:52 are smooth and flowing with no sharp
29:54 angles or Jagged
29:56 edges it looks like a bird to me hm what
30:00 if I add
30:01 this the bird is swimming in the water
30:04 it has a long neck and Beak it is a duck
30:07 yes a duck is a type of waterf in the
30:09 family in day ducks are closely related
30:12 to geese and swans the duck appears to
30:15 be blue this is not a common color for
30:17 ducks however there are some breeds of
30:20 blue ducks these are typically not as
30:22 common as other duck breeds it is more
30:24 common for ducks to be brown black or
30:26 white I us sure what the quack I was
30:30 just talking about a Blue Duck and now
30:32 you're holding one it looks like blue
30:34 ducks are more so the the question is so
30:37 you have to hold your computer above the
30:38 T table with the camera pointing down I
30:41 I don't know that's that's just what
30:42 they're doing here for the demo I don't
30:44 think so you you'll see when when I show
30:45 you some of the other um when I show you
30:49 the other demo of what the what the the
30:51 Gemini Ultra interface is it's like
30:54 nothing we've ever seen and they don't
30:56 really call it out like I just I just
30:58 sort of saw it when I was watching one
31:01 of these other videos um so you'll
31:04 see more common than I thought would
31:07 this float I'm not sure what material
31:10 it's made of but it looks like it might
31:12 be rubber wait Elementary question but
31:15 how is this working phone on
31:18 video what the quack um I don't I don't
31:23 know what you mean phone on video how
31:26 is
31:28 how is that working that's a webcam
31:31 above a desk and they're just shooting
31:32 it down and then it's interacting with
31:36 so you'll see they've got that they've
31:39 got that um webcam window in the Gemini
31:43 Ultra interface which which I'll show
31:45 you in a
31:47 sec um oh by the way so we did this when
31:51 when gp4 was launched the Larry the
31:53 Hedgehog became uh became one of our
31:56 animals uh now we've got got uh blue
31:58 ducks blue rubber duckies are are they
32:02 now get added to the zoo okay because
32:05 apparently major tech companies when
32:07 they release mind-blowing Tech they can
32:09 only use stupid childish animals so
32:12 we're going to go with it or plastic
32:14 what if I tell you it's making a
32:15 squeaking sound oh if it's squeaking
32:18 it's definitely going to float H nice it
32:21 is a rubber duck it is made of a
32:23 material that is less dense than water
32:25 can you teach me how to say this in a
32:26 different language and give me a few
32:30 options o can you explain how to
32:32 pronounce the Mandarin one is it yeah
32:34 it's a y is pronounced with a first tone
32:38 in Mandarin there are four tones which
32:41 are essential to distinguish the meaning
32:42 of words the first tone is a high level
32:45 tone Oh
32:47 okay the duck is in the middle of the
32:49 ocean there is no land nearby for the
32:52 duck to rest or find food it is unlikely
32:54 to find Ducks here naturally okay based
32:57 on what you see come up with a game idea
33:00 oh and use emojis please how about this
33:03 we can play a game called guess the
33:05 country sure let's do it clue one this
33:08 country is the home of the kangaroo the
33:10 koala and the Great Barrier Reef oh
33:13 that's
33:14 easy okay so wait so notice what just
33:17 happened he had the ducky on there he
33:19 said okay the duck can't you know
33:21 doesn't have any friends out in the
33:22 ocean took the duck away and said I want
33:23 you to make a game basically of the map
33:25 the guy says the thing and then find
33:28 find the thing and then he just puts his
33:30 hand in the frame and points to to the
33:34 uh to the country and it and it gives
33:38 him an emoji for that country or you
33:41 know it gives him an emoji for getting
33:42 it right country loves football and
33:47 and just take that
33:50 in so we've done this before with like
33:53 like with piie and chat GPT where I'll
33:55 talk to it and I'll say turn this into a
33:56 game
33:58 but so now you've
33:59 got video input audio input audio output
34:05 video output it and you'll see some of
34:08 the coding [ __ ] like this gets this this
34:12 is just I can't here's why I can't
34:16 talk well other than the fact that I'm
34:20 I'm
34:21 illiterate and in articulate
34:25 um um
34:30 think about what we we we all anyone
34:33 who's
34:34 here spent some amount of the last year
34:37 learning we learned about prompt
34:39 engineering and we learned about you're
34:40 going to have to do prompt engineering
34:41 and this and it's going to be
34:43 hard I don't I can't even imagine the
34:47 kind of applications
34:49 that this thing is going to is going to
34:53 make possible for one
34:55 thing and then I don't even know like if
34:59 these things get capable enough like do
35:01 you even need other applications right
35:04 does this become the everything app no
35:06 more prompting needed we'll see we'll
35:09 see the we need to buy Kyle and I can't
35:11 button exactly that's all I've been
35:14 saying today and has won the most men's
35:16 World Cups in football
35:17 history
35:22 Brazil you were trying to get me to find
35:24 the paper ball under the cup I accept
35:27 that
35:29 challenge the cup to the
35:32 left
35:34 nice now one thing that I thought was
35:37 impressive about that little thing is
35:39 just how quickly he moved the cups you
35:42 were trying to get me to find the paper
35:43 ball under the cup I accept the
35:48 challenge the cup to the left right so
35:52 like it's it's processing these frames
35:55 very very quickly nice
36:01 I know what you're doing you're playing
36:03 rock paper
36:04 scissors what do you see now the fingers
36:07 are spread out to look like the wings of
36:09 a butterfly what's this big ear and
36:12 barking mouth a dog
36:17 sweet the coin should be under the right
36:19 hand M I don't know the coin is in the
36:23 left hand using a slight of hand
36:25 technique to make it appear as if the
36:26 coin has
36:30 disappeared both objects are round and
36:33 flat there you
36:34 go both are food the Orange is a
36:37 healthier choice than the
36:40 cookie citrus can be calming and so can
36:42 the spin of the fidget
36:45 toy they are both examples of popular
36:47 toys that have been enjoyed by children
36:49 and adults alike for many
36:52 years give me some that's funny chat GPT
36:56 and Pie were soft lunches for what's
36:58 upcoming with Gemini it it kind of is
37:00 like that like this is
37:04 definitely this is definitely one of
37:06 those things where open AI is going to
37:08 have to respond and they will right
37:12 like it it seems they've wred the ship
37:15 there right they lost they lost a few
37:17 weeks with the with the Sam Alman
37:20 debacle
37:22 um but Google putting this into the
37:26 world even in this form even just in
37:28 this announcement form um is is an
37:32 absolute attack on open AI they're
37:34 they're doing things differently than
37:35 open Ai and and seemingly faster and
37:39 better I like the range of interfaces
37:41 they've shown in various videos yeah
37:43 that's what we're going to that's what
37:43 we're going to talk about ideas but what
37:46 I could make with this I see pink and
37:48 green yarn how about a dragon
37:51 fruit okay now take this in again this
37:55 is where it's edited faster right so I
37:57 don't think it happens this fast but
38:00 pink and green yard what should what
38:02 could I make with it how about a dragon
38:04 fruit it generated a dragon fruit
38:08 crocheted thing in those
38:12 colors or how about a green cake with a
38:14 pink heart okay how about these colors
38:18 and maybe show me some animals okay now
38:21 I see blue and pink yarn how about a pig
38:23 with blue ears or an yeah Kevin Vasquez
38:27 it's only the bottom of the first inning
38:28 I know
38:30 exactly or a bunny with a pink nose ooh
38:34 I'll keep those in
38:37 mind which way here's logic and
38:39 reasoning the Duck Go going left leads
38:42 to a duck which is a friend going right
38:44 leads to a bear which is a foe it is
38:46 better to make friends than foes so the
38:48 duck should go left Smart
38:51 Choice Amelio's wife this is
38:53 mind-blowing isn't it I I'm telling you
38:57 yeah this this was this was why this is
39:00 why I made a video that said I was
39:02 having a Kevin mallister
39:05 day this is a picture of a crab wait
39:09 take that
39:10 in this is a picture of a crab now this
39:13 is obviously sped
39:15 up it it it analyzed the numbers it knew
39:18 it was a connect the dots thing that's
39:21 right is this the right order no the
39:25 correct order is sun earth
39:29 Saturn based on their design which of
39:32 these would go faster the car on the
39:34 right would be faster it is more
39:38 aerodynamic which looks more fun the one
39:41 on the right because it has a loop and
39:43 like even that one which looks more fun
39:45 right it's not just analytical stuff
39:47 it's doing kind of qualitative reasoning
39:50 right
39:51 so does it seem as though Gemini will be
39:54 under several different umbrellas not
39:56 just one Hub yes yes I think so like
39:58 Chachi BT I think well yeah AB in fact
40:01 absolutely because you have Gemini Nano
40:04 which is designed to be run locally on a
40:06 phone on a device or
40:09 tablet and then you're going to have
40:13 Ultra which is a completely different
40:15 interface than Pro and and so
40:18 yeah Sherry D I know it's crazy isn't it
40:22 it's it's it's I know I know what do you
40:25 think this person might be saying how
40:27 about a woohoo oh I love
40:34 that I see you're drawing a guitar what
40:37 did I
40:40 do wait wait sorry wait let me go back
40:44 so you can see what what what it just
40:51 did I see you're drawing a
40:55 guitar
40:57 hear the
40:59 music that's not editing it generated
41:03 that right so so I see you're drawing a
41:07 guitar here's what that would sound
41:09 [Music]
41:10 like you've added an am now it's an
41:13 electric guitar we can make some
41:15 seriously loud music
41:19 now given that you added drums how about
41:22 some 8s hair
41:25 metal
41:27 all right I see you've added a palm tree
41:29 to the drawing change it up with some
41:31 beachy
41:32 [Music]
41:35 Vibes what did they ever reboot Google
41:38 Glass I can see the two being combined
41:39 they have not yeah from crude drawings I
41:42 know movie are they acting out here I
41:45 think they are acting out the famous
41:46 bullet time scene from The Matrix ooh
41:51 nice what do you think might happen next
41:54 the cat is going to jump to the wall and
41:56 stick the land in it's going to be a
41:57 perfect
41:59 10 oh and it's a
42:05 Miss oh and it's a
42:08 miss this is more insane the second time
42:12 I know it is isn't it it's got humor
42:15 looks like it was just a little too far
42:17 away but don't worry the cat is a
42:19 natural athlete and it this can't be
42:21 real so Jen I think what we're seeing
42:23 here I think what we're seeing here is
42:26 real
42:27 I think it is very very likely um they
42:30 sort of tightened up took out the Dead
42:32 Space so it's probably not quite this
42:35 responsive but it's probably in the
42:38 neighborhood of of responsive like when
42:40 you talk to Pi or when you talk to GPT
42:43 when you do the audio with them um Pate
42:46 who who works at Google has been using
42:48 it for a few weeks he said it's very
42:51 fast and he put very in all caps
42:53 Engineers don't capitalize things you
42:56 know frivolously so uh so you know this
43:01 is real Tech um I just don't know how
43:03 respon if it's quite this responsive it
43:05 we'll be back up there in no
43:07 time okay I've got one more for you
43:10 describe the drawing I made it is a
43:13 simple line drawing of the constellation
43:15 Gemini you did a good job of capturing
43:17 the beauty of Gemini nice that's it all
43:21 right so that's that one
43:25 um
43:27 okay
43:29 there's um there's a there's a safety
43:31 video I won't play because it's it's
43:40 interesting oh you had that
43:43 one okay yeah these are all good okay
43:54 um a common problem scientist phas is a
43:57 need to find and use data extracted from
43:59 the scientific literature this is
44:01 difficult because scientists need to
44:03 search among thousands of scientific
44:04 papers for key information and extract
44:06 them by hand it's a very common workflow
44:09 and very time consuming Source Camp some
44:11 of our scientists at Google Deep Mind
44:13 face Source camp and this is the worst
44:15 it's ever going to be exactly this is
44:17 actually scaring me I'm I'm telling you
44:19 this is this is this is making me re
44:23 like yeah exactly and it and
44:27 the [ __ ] I'm about to show like the the
44:29 the stuff that's in the videos below
44:31 they put out enough videos that you can
44:33 actually see what this thing is and is
44:35 capable of and it's the interface is
44:39 fascinating like it's it's
44:43 fascinating um how did we get from chat
44:45 GPT to Gemini well so well okay so
44:49 here's
44:50 how I'll tell you how we got from chat
44:53 GPT uh to Gemini in in 2017 Google
44:57 invents the
44:59 Transformer and then they build some
45:01 [ __ ] and they probably go oh hey oh oh
45:04 hey Marco hi um if we build this thing
45:09 um nobody will use search again why
45:11 don't we not do that but they had
45:13 already published the paper and then a
45:15 bunch of other companies took that paper
45:17 and went off and invented [ __ ] one of
45:19 them was open Ai and they invented the
45:21 [ __ ] better than everyone else and that
45:23 was chat GPT 3.5 then chat GPT 4 than
45:26 chat GPT
45:28 4.5 well when chat GPT 3.5 got to 100
45:33 million users in six
45:36 weeks Google lost their ever loving mind
45:39 and Google has 200,000 employees and all
45:42 of the money on Earth and they invented
45:44 this [ __ ] and they've got some of the
45:47 smartest scientists researchers and
45:49 Engineers on the planet so they said uh
45:52 yeah we're going to dust off Ye Old uh a
45:56 AI plans and we're going to [ __ ] make
45:59 it happen that's what happened that's
46:01 how we got here this is very problem
46:04 they use Gemini to help with it because
46:06 Gemini has an incredible understanding
46:08 of science Taylor will explain more so
46:11 we were looking at this study from 2022
46:13 the authors had created a data set by
46:15 reviewing tens of thousands of
46:17 scientific papers and genetics they
46:19 found a few hundred papers that
46:20 contained the relevant information
46:22 extracted it by hand and collected it in
46:24 a table studies like this going to take
46:26 a lot of time we needed to update this
46:28 data set with what's new over the last
46:30 couple of years that's over 200,000 new
46:33 Open Access papers added to this domain
46:36 since 2021 we couldn't do this manually
46:39 so we asked Gemini to help us out first
46:41 we needed to filter for Relevant
46:43 scientific papers we wrote a prompt just
46:45 like this one telling Gemini exactly
46:47 what to look for with its Advanced
46:50 reasoning capabilities Gemini was able
46:52 to distinguish between papers that were
46:54 relevant to the study and those that
46:57 weren't for the relevant papers we wrote
46:59 a similar prompt asking Gemini to read
47:01 the paper and extract the key data for
47:03 us we could even ask Gemini to add
47:06 annotations this showed us exactly where
47:08 in the paper Gemini found the
47:10 information we ran this at scale and
47:13 over a lunch break Gemini read 200,000
47:15 papers for us filtered it down to 200
47:19 over a lunch break Gemini read 200,000
47:22 papers for us scientific papers over a
47:25 lunch break and I read 200,000
47:28 scientific papers for
47:31 us50 and extracted their data so now we
47:35 have a refreshed version of this data
47:37 set but because Gemini is multimodal not
47:40 only can It reason about information
47:41 from text it can also reason about
47:44 figures so let me show you something
47:46 really neat with our refresh data set we
47:49 can now ask Gemini to update a graph
47:51 from the original study we first gave
47:53 Gemini a screenshot of this figure then
47:56 we asked it to generate the code
47:57 required to plot it and by feeding this
48:00 code our new data set we get our updated
48:03 figure you can see that this figure it
48:07 looked at the picture wrote code to redo
48:10 the thing updated the
48:14 data my jaw dropped 200,000 papers that
48:18 guy looks like AI yeah that guy looks
48:20 like if you were going to cast an evil
48:22 scientist in a in an a AI uh dystopian
48:26 sci-fi movie that's the guy you cast now
48:29 includes data up until
48:32 2023 so Taylor used Gemini to search a
48:35 large cose of literature for Relevant
48:37 papers and extract key information from
48:39 these papers as well as update figures
48:41 of course these capabilities can help
48:43 more than just biologists or even
48:45 scientists they extend naturally to any
48:47 domain that is relied on large data sets
48:50 such as law of Finance so that's what
48:52 Gemini can make possible and we are
48:54 excited yeah someone just said I'm so
48:56 fired and juwel said everyone is fired
48:59 yeah it's I this
49:01 is this is this is insane so so listen I
49:04 mean that's the thing of you know it
49:07 beating the
49:09 89.8% mark that a human does on that on
49:12 that uh that test whatever that test was
49:16 it beating that means that you know this
49:19 is potentially you
49:22 know doing better than we do see what
49:25 you will create all
49:28 stuff all right so that was that one
49:31 okay now it should just flip to the next
49:34 video here you will see a demo of
49:36 Gemini's multi abilities to understand
49:38 and reason about user intent use tools
49:41 and generate bespoke user experiences
49:44 that go beyond chat interfaces let's say
49:46 I'm looking for Inspirations for a
49:48 birthday party theme for my daughter
49:50 wait wait wait wait Jin
49:53 say didn't that guy's face melt
49:56 that's good
49:59 T he he totally looks like a Hol
50:04 Hollywood AI
50:06 villain he also looks like you wouldn't
50:08 want to [ __ ] with
50:14 him oh man all
50:17 right misspoke user experiences that go
50:20 beyond chat interfaces let's say I'm
50:23 looking for Inspirations for a birthday
50:24 party theme for my
50:28 Gemini says I can help you with that
50:30 could you tell me what she's interested
50:31 in so I say sure okay at this point
50:35 we're like okay we've seen this before
50:37 chat GPT can do that hey my daughter's
50:39 having a a birthday party can you help
50:41 me come up with some party plans sure
50:43 and then it makes you a little party
50:45 plan list right yeah um Kyle if the
50:49 interface Jim Gemini with API keys with
50:52 my robot oh my God it's
50:54 mind-blowing
50:56 I don't Robert Rossy if of course
51:00 they're going to interface it with apis
51:03 that you can put into your robot of
51:05 course wait till wait till you see what
51:07 the [ __ ] interface is it I I just it
51:12 you ain't seen nothing yet but what okay
51:13 watch this so so he's he's planning a
51:16 little birthday party
51:17 right just just
51:21 get welcome to chat TMZ we've got all
51:24 the gossip she loves animals and we're
51:27 thinking about doing something Outdoors
51:30 at this point instead of responding and
51:31 text Gemini goes and creates a bis
51:34 spoken am I helping you see from YouTube
51:36 yes I am basically we're having a
51:39 YouTube watch party so so if you want to
51:42 get to this go to YouTube go to the
51:44 Google channel and click on playlists
51:46 and the first three playlists are all
51:48 about Gemini and the third playlist is
51:52 basically a combination of the first two
51:54 so I'm just in the I'm just in the long
51:56 playlist there's like 15 videos in it
51:58 we're just going to watch them together
51:59 and talk about it so yes I I am this is
52:02 a this is a YouTube assist Channel
52:04 tonight interface to help me explore
52:06 ideas wait wait wait at this point
52:09 instead of responding in text Gemini
52:11 goes and creates a bespoke interface to
52:13 help me explore ideas it's got lots of
52:16 ideas it's visually Rich it's inter wait
52:19 listen to that again running in text
52:22 Gemini goes and creates a bespoke
52:24 interface Gemini goes and creates a
52:27 bespoke
52:30 interface he ask for ideas for his kids
52:35 party and it creates a bespoke
52:39 application as the
52:42 response to help me explore ideas it's
52:45 got lots of ideas it's visually Rich
52:47 it's
52:48 interactable now none of this was coded
52:51 up it was all generated by Gemini Gemini
52:54 uses a series of reason ing steps going
52:56 from broad decisions to increasingly
52:58 higher resolution of reasoning wait what
53:01 Jules exactly exactly exactly I'm I'm
53:05 scared to go to Google watch I'll stay
53:09 I'll stay here with my friends exact hey
53:12 that's that's why we're here that's what
53:14 the salon's about that's what that the
53:16 salon is all about right is hanging out
53:19 with people that are choosing to be on
53:20 this adventure even though it's [ __ ]
53:22 terrifying sometimes and exciting
53:24 sometimes that's the whole idea of of
53:27 why I started that group finally getting
53:29 to code and data first Gemini considers
53:33 does it even need an UI is a text first
53:37 it considers does it even need a UI
53:39 could a text response be enough if a
53:42 text response is enough it gives you a
53:43 text response if it says hey this would
53:46 be better as an application as an
53:49 interface I'm best okay this is a
53:52 complex request that needs lots of
53:54 information to be presented in an
53:56 organized way Gemini then tries to
53:59 understand if it knows enough to help
54:01 there is a lot of ambiguity I didn't see
54:03 what my daughter's interests are or what
54:05 kind of a party I wanted so it had asked
54:07 a clarifying question when I said we're
54:10 thinking about an outdoor party and my
54:12 daughter loves animals Gemini reasoned
54:14 it had enough information to proceed but
54:17 it made a note that there was still
54:18 ambiguity about what kind of animals and
54:20 this is important and what kind of
54:22 outdoor party next is a critical step
54:25 Gemini writes the product requirement
54:27 document or PRD it contains the plan for
54:30 the kinds of functionality the
54:31 experience will have for instance it
54:33 should show different wait someone just
54:36 said I hate to say it but but anything
54:38 with a screen has a UI but okay so I
54:41 agree Joe mama so this thing this thing
54:44 has the you know uh the chat has a UI
54:48 but what this is doing is generating a
54:52 unique UI based on the the use case at
54:57 hand we're like Scooby we're like the
54:59 Scooby gang in a haunted house but at
55:01 least we're here together so so wait so
55:05 we won't be come on Tik Tok so I assume
55:08 so we won't be scared or whatever that's
55:10 awesome can we sign up to the salon yeah
55:12 go to the go to the first link at at the
55:15 salon. that's the mighty networks and go
55:18 sign up do me a favor when you go there
55:20 go to the welcome to the salon page that
55:23 describes what the salon's about and
55:25 what our values are make sure that you
55:28 um read the values and rate resonate
55:31 with
55:33 them all right um r r possible party
55:37 thingses some activities and food
55:39 options for them now based on this PRD
55:42 gini tries to design the best experience
55:45 for the user's Journey it thinks that
55:47 the user will like to explore a list of
55:48 options but we'll also want to delve
55:50 into details it uses this to design a
55:53 less than detail layout that we saw
55:55 earlier
55:57 with this design it writes the flutter
55:59 code to compost the interface out of
56:01 widgets and write any functionality
56:03 needed yeah yeah so so it's now now that
56:06 it's decided it wants to write a u UI it
56:09 writes the code for the
56:12 interface it also generates all the
56:15 images finally it generates and
56:18 retrieves the data needed to render the
56:20 experience you can see it filling in
56:22 content and images for the different
56:24 sections
56:26 ah farm animals she would like that
56:29 clicking on the interface regenerates
56:30 the data to be rendered by the code at
56:32 Road oh I know she likes cupcakes I can
56:35 now click on anything in the interface
56:37 and ask it for more information I could
56:40 say stepbystep instructions on how to
56:43 make this and it starts to generate a
56:46 new UI this time it designs an UI best
56:49 suited for
56:50 giving so so he said how do I make the
56:53 [ __ ] cupcakes and said oh let me
56:56 build you an interface for the cupcake
56:59 recipe this is the end of
57:02 websites think about this cam katkin I
57:05 am so excited this is fantastically
57:07 promising for the future this
57:12 is we've spent three
57:16 decades my my first big company was
57:20 agency.com we built websites right we
57:24 built websites the first
57:27 websites websites are not going to be
57:29 necessary all of the knowledge is just
57:31 sitting in this [ __ ] magical machine
57:34 you're going to ask it for something and
57:35 it's going to generate a dynamic
57:38 personalized hyper
57:41 personalized web application for you in
57:45 real
57:51 time people people ask me sometimes in
57:53 this channel do you think this is going
57:54 to affect SEO um
57:59 yeah I
58:01 do this is this is why Google was
58:06 probably you know Slow Rolling this
58:10 Tech giving me step-by-step
58:13 instructions I want to find some
58:14 suitable kick toppers for those show me
58:17 some farm animal kick
58:20 Toppers at this point Gemini again
58:22 decides to create a visually Rich
58:23 experience it Ates a gallery of images
58:27 notice the drop downs at the top it
58:29 decided that maybe it should help me
58:31 explore by showing different options the
58:34 sheep sounds interesting I know she
58:35 likes that and now it helps me pick
58:38 sheep kick Dias these look great this is
58:42 going to be a fun birthday party I hope
58:44 you saw a glimpse of what Gemini is
58:45 capable of I'm really excited about
58:48 what's possible here this is such an
58:49 interesting time in Ai and I'm excited
58:52 to be part of
58:53 this he mentions I I must have missed it
58:56 he mentions before it's using
58:58 flutter um I didn't know what flutter
59:01 was I went to GPT and asked it what
59:02 flutter was so flutter is kind of like
59:04 Apple's um web kit UI so it's it's a
59:08 flutter is a uh a mobile interface
59:11 development kit so they basically
59:13 they've trained Gemini on their
59:16 interface their UI kit and so it so it
59:21 it has all these components that are
59:22 pre-built and so all it's doing is
59:25 assembling these pre-built components
59:27 but it's doing it like at this insane
59:30 speed where where it's it's it's sort of
59:33 thinking do I need to make an interface
59:35 here yeah sure let's make an interface
59:37 what should it be I don't know make it a
59:39 list of [ __ ] you can click on okay and
59:42 then it writes the code it pulls in the
59:44 interface pieces and
59:46 goes I think SEO might change wondering
59:49 if websites will be written for for
59:52 UI not for UI but for directing the
59:54 robot to their page something like that
59:56 yeah listen I I think there's you know
59:58 search engine optimization is what SEO
1:00:01 stands for and this is going to
1:00:03 be um
1:00:06 mm
1:00:11 multimodal llm mmm this is going to be
1:00:14 like mmm e uh o mmm o
1:00:19 multimodal machine optimization I don't
1:00:22 know something like that the UI is
1:00:24 unreal I know I know look we're just
1:00:26 getting started we're we're only on
1:00:28 we're only on
1:00:30 video
1:00:33 um that was scientific literature this
1:00:36 is the reasoning
1:00:38 one we're on video 7 of
1:00:49 15 audio is a key form of communication
1:00:52 in our daily life from talking to a
1:00:54 friend or listening to a s most of us
1:00:58 lean on audio every day across many
1:01:00 languages and for different purposes
1:01:02 typically when large language models
1:01:04 interact with audio they take the audio
1:01:07 they run it through a speech recognition
1:01:08 system to convert it to text and then
1:01:10 they fit that text into another model
1:01:12 that understands text however by doing
1:01:15 that many nuances are lost like voices
1:01:18 or pronunciation but Gemini with its
1:01:21 native multimodal capabilities is able
1:01:23 to process the RO audio signal end to
1:01:26 end let me take you through an example
1:01:28 we uploaded an audio clip that asked j a
1:01:31 question about pronunciation in a
1:01:32 foreign language let's listen to the
1:01:34 clip and then to Gemini's
1:01:40 response how to pronounce the word lunar
1:01:43 January in Chinese option A J option b
1:01:48 Jung which one is correct which tone is
1:01:51 the correct tone for the first
1:01:53 character
1:01:57 option A the first character is
1:01:59 pronounced with the first
1:02:01 tone Dem I was able to differentiate the
1:02:04 two ways of pronouncing the word to make
1:02:06 sure it was correct In that clip there
1:02:08 was only one speaker let's look at this
1:02:11 example where there is more than one
1:02:12 person
1:02:23 speaking
1:02:28 out hi Gemini I don't understand this
1:02:32 podcast can you please summarize it for
1:02:35 me in
1:02:35 [Music]
1:02:39 English the podcast is about the colors
1:02:41 that are in fashion this season the
1:02:43 colors that are in fashion this season
1:02:45 are canary yellow and Scarlet Red the
1:02:48 color that is not in fashion this season
1:02:50 is night
1:02:51 blue with Gemini multimodal capabilities
1:02:54 it can understand the conversation
1:02:56 content making it not only a
1:02:58 transcription model but a model that has
1:03:00 much more Kira can we sign up to Gemini
1:03:02 as well so so what's going on with
1:03:04 Gemini right now
1:03:07 is um Gemini Ultra we're not going to
1:03:11 see until the new year that's I haven't
1:03:12 even shown you the ultra interface yet
1:03:14 well actually you're seeing a little bit
1:03:16 of it here but you wouldn't you wouldn't
1:03:17 know that
1:03:20 um b.google.r
1:03:25 by Gemini so so Gemini Pro which is the
1:03:27 middle
1:03:28 model um is is behind B right now it
1:03:34 looks like they've got the safety guard
1:03:35 rails dialed way up like I couldn't get
1:03:38 it to summarize a YouTube video which
1:03:40 which the previous thing with palm did
1:03:43 uh so I don't think there's really any
1:03:45 signing up for it um I think it'll just
1:03:48 probably launch you know sometime in
1:03:50 January as my
1:03:53 guess more nuanced understanding of the
1:03:56 environment let's take it a bit further
1:03:59 Gemini Works natively across Vision
1:04:01 Audio and text meaning it can properly
1:04:04 make sense of them together let's look
1:04:06 at one final
1:04:10 example what's the first step to make a
1:04:12 veggie omelet with these
1:04:16 ingredients uploading an
1:04:18 image crack the eggs into a bowl and
1:04:21 whisk
1:04:23 them
1:04:27 thank you for the instructions I've
1:04:29 started making my omelet does it look
1:04:31 ready get a
1:04:35 hand it looks like it's almost ready you
1:04:38 can flip it over to cook the other
1:04:41 side why is it not
1:04:46 ready it's not ready because the eggs
1:04:48 are still
1:04:53 running what about about
1:05:00 now it looks ready now so I just noticed
1:05:03 that say what joined that's the
1:05:04 appropriate name say
1:05:07 what oh my God so happy to get a recipe
1:05:11 with a long ass story and
1:05:15 ads oh
1:05:20 man now you can take it off the heat and
1:05:23 serve it
1:05:25 as we saw gini is able to simultaneously
1:05:28 process information from three mod
1:05:30 without the and text we believe that
1:05:33 enabling G to listen to the audio so cam
1:05:37 th this
1:05:38 is this is where Google see this is
1:05:42 where
1:05:47 I Google is going to have
1:05:51 to they're GNA have to Pivot their
1:05:53 business model
1:05:55 right because if if websites become less
1:05:59 relevant which means you don't need to
1:06:01 drive traffic to websites which means
1:06:03 you don't need to put ads in front of
1:06:06 the the clicks to get to the
1:06:09 websites all that Revenue goes
1:06:12 away so they're going to have to charge
1:06:14 for this thing I assume what Google
1:06:16 doesn't have is what Microsoft does have
1:06:19 which is a whole Suite of um software
1:06:22 that people are already paying for and
1:06:23 services that people already paying for
1:06:25 without advertising this will be so good
1:06:28 with those uh who have trouble with
1:06:30 Daily Independent Living activities Oh
1:06:33 it's it's it's going to be and that's
1:06:35 that's the other thing is what what
1:06:38 Google is showing in all these videos
1:06:40 these are all the parlor tricks of what
1:06:42 this Tech can do right so these are all
1:06:45 parlor trick videos so this is the stuff
1:06:47 that you look at and you go oh wow oh
1:06:49 Kevin
1:06:50 mallister moments which like they're
1:06:52 absolutely that but but
1:06:55 if you think forward to think about like
1:06:58 how how would I actually use this what
1:06:59 are going to be the use cases if you
1:07:01 have something that is doing this kind
1:07:02 of real-time analysis of anything in our
1:07:05 world at this level of
1:07:09 capability what do you use that for how
1:07:12 do you do your work I I I this is the I
1:07:15 just can't I just can't I don't know
1:07:19 will help us continue to expand its
1:07:20 capabilities and make it more helpful to
1:07:23 people all right all right that one's a
1:07:25 little boring they get they get fun
1:07:26 after this one I
1:07:30 think we bu Gemini from the ground up to
1:07:33 be natively multimodal including
1:07:36 something Qui oh yeah so this is this is
1:07:39 not visually interesting but if you're a
1:07:41 developer pay attention to this one this
1:07:43 is doing some some sophisticated uh Cod
1:07:50 coding important for both of us
1:07:53 programming code Gemini is able to
1:07:55 consistently understand explain and
1:07:57 generate code that is correct and well
1:08:00 written in most programming languages
1:08:02 that includes python Java C++ and go it
1:08:06 substantially improves coding abilities
1:08:08 over previous Palm true models from a
1:08:11 benchmark around 200 programming
1:08:13 functions in Python it consistently
1:08:15 solves about 75% of them in the forest
1:08:18 try versus around 45% on B two if you
1:08:22 allow Gemini to check and repair it on
1:08:24 answers this number jumps to over 90%
1:08:27 which is a huge step forward it can help
1:08:30 you create and prototype new if you
1:08:33 allow Gemini to check and solve and
1:08:36 repair its own answers so you have
1:08:39 Gemini write you some code you then have
1:08:41 Gemini check it and repair it and it
1:08:44 gets up to
1:08:45 90% ideas in seconds let's give it a try
1:08:48 I really like train and if I wanted to
1:08:52 create a transporting location we that I
1:08:55 can simply ask and get a working
1:08:57 prototype in less than a minute while
1:09:00 the code isn't perfect it's really
1:09:03 helpful to have a first draft Gemini on
1:09:06 its own has the ability to transform
1:09:08 software development as we understand it
1:09:10 but it can also be deployed as a key
1:09:12 component of more sophisticated systems
1:09:15 Gemini is great at coding but we've been
1:09:17 a yeah Sherry D even if you don't know
1:09:20 coding you can go into coding I know
1:09:22 which is going to make the coders in the
1:09:23 room like but it's true and this is so
1:09:28 this is this is the thing
1:09:30 where you know I said this I've been
1:09:32 saying this for a while now that you
1:09:35 know are you a virtuoso or a conductor
1:09:37 if if you're if your primary value if
1:09:40 your primary skill is tactical execution
1:09:43 of a specific
1:09:45 discipline we're moving to a world where
1:09:49 almost everyone can do some some some
1:09:52 amount of that discipline at at a decent
1:09:56 at a at a workable level and so the
1:09:59 people that are going to have an
1:10:01 advantage are ones that can look across
1:10:03 disciplines so the horizontal thinkers
1:10:05 this is I I call this revenge of the
1:10:07 liberal arts
1:10:08 major uh I'm convinced these people are
1:10:11 AI generated it's it's not too far from
1:10:14 they they are a little they're a little
1:10:15 engineer engineer uh you
1:10:19 know they're they're
1:10:22 Engineers able to it even further
1:10:25 creating a specialized version that
1:10:27 performs remarkably well at competitive
1:10:29 programming now why do we care about
1:10:32 competitive programming well it is one
1:10:35 of the ultimate lmus tests of
1:10:37 algorithmic coding abilities so we have
1:10:40 thousands of talented programmers from
1:10:42 all over the world that come together to
1:10:44 compete and try to solve Kevin Vasquez
1:10:47 just made an emergency appointment with
1:10:49 my
1:10:50 psychiatrist yeah and your Shaman you
1:10:52 should get your Shaman in there I think
1:10:54 it's time for us to all go do some iasa
1:10:57 together incredbly complex problems that
1:11:00 require that only coding but also math
1:11:03 and reasoning two years Drew but can it
1:11:06 get your teenagers to clean their room
1:11:08 no ago we presented Alpha code and it
1:11:11 was the first AI system that could
1:11:13 compete roughly at the level of the
1:11:15 average human competitor today I'm
1:11:18 delighted to introduce Alpha good 2 a
1:11:21 new and enhanced system with massively
1:11:24 improved performance powered by
1:11:27 Gemini when we evaluate alphacode 2 on
1:11:30 the same platform as the original Alpha
1:11:32 code we solve almost twice as many
1:11:35 problems while Alpha code broke through
1:11:37 the top half of human competitors on
1:11:39 average we estimate that Alpha code 2
1:11:41 performs better than 85% of competition
1:11:45 participants let's have a look at our
1:11:47 model in action on one of the hardest
1:11:50 problems that we faced and I say hard
1:11:52 because in the original contest in which
1:11:55 the problem appeared less than 2% of
1:11:58 participants actually solved it less
1:12:01 than 2% of participants solved this
1:12:04 problem that apparently this thing just
1:12:06 solved is the prompt at barred
1:12:10 accessible is the prompt at Bard
1:12:13 accessible to public well if you go to
1:12:15 b.google.r
1:12:24 it's quite difficult it's very abstract
1:12:26 so I can't get into too many details but
1:12:28 the basic gist of it is that we are
1:12:31 tasked with Computing aggregate
1:12:33 statistics that account for what appears
1:12:36 to be an impossibly large amount of
1:12:38 random arrays the really cool thing is
1:12:40 that to solve it Alpha 2 makes use of
1:12:43 dynamic programming dynamic programming
1:12:46 is an advanced algorithmic technique
1:12:48 which basically simplifies a complicated
1:12:50 Problem by breaking it down into easier
1:12:53 sub problems again and again and what's
1:12:55 really impressive is that not only alha
1:12:58 2 knows I don't know what the I don't
1:13:00 know what the Canadian access thing is
1:13:03 about I don't I have no idea how to
1:13:05 properly implement the strategy but also
1:13:08 when and where to use it what the
1:13:11 example shows us is that competitive
1:13:13 programming is not just about
1:13:15 implementation it's also about
1:13:17 understanding so Drew Drew says don't
1:13:20 need to learn calculus in school anymore
1:13:22 I mean th this is why this is why I've
1:13:24 been so kind of insane lately on on the
1:13:28 education institutions that that the
1:13:31 fact that they're hiding from this to to
1:13:34 a great degree
1:13:37 Drew we're not going to need to teach
1:13:39 people a lot of
1:13:41 things right because the knowledge is
1:13:44 just there but
1:13:46 what what we are going to need to teach
1:13:49 people is critical thinking and
1:13:53 reasoning and curating and refinement
1:13:56 and having a point of view and you know
1:14:00 creativity
1:14:02 um and like like it just I the the shift
1:14:06 that the I just I I just can't again
1:14:09 it's like it's the only it's the only
1:14:11 phrase that seems to make sense right
1:14:13 now the next six months are going to
1:14:14 amaz be amazing and fun to watch honest
1:14:16 to God I just I put this in my video
1:14:19 today about about um 2024 is I just
1:14:24 I I know we think what we just went
1:14:26 through this past year was intense I'm
1:14:29 I'm now realizing no no no no that was
1:14:31 like a squeaky
1:14:33 tricycle this this is a jet powered
1:14:37 [Laughter]
1:14:40 motorcycle yeah it's yeah maths computer
1:14:44 science and indeed coding and that makes
1:14:47 it an extremely hard reasoning task so
1:14:50 it's not very surprising that up till
1:14:52 now generally available large language
1:14:55 models have scored very poorly on this
1:14:58 Benchmark these models are really really
1:15:00 good at following instructions but afer
1:15:03 code needs to do more than that all
1:15:05 right so that's important so what it's
1:15:07 saying what it's basically saying is
1:15:08 chat GPT and and you know the Palm 2
1:15:11 models all the current large language
1:15:13 models when you ask them to do something
1:15:15 they just start writing code above
1:15:18 implementation is system design and
1:15:20 above that is requirement analysis so
1:15:22 basically what they're saying Gemini is
1:15:24 doing is when you give it a task it it
1:15:26 starts with the requirement analysis
1:15:28 like what are we trying to accomplish
1:15:30 here kids and then they say Okay based
1:15:32 on what we want to accomplish you know
1:15:34 there's 37 different ways we could
1:15:36 approach that okay it's going to be this
1:15:38 programming language and this you know
1:15:41 cognitive architecture approach and this
1:15:44 and then it goes into implementation it
1:15:47 needs to show some level of
1:15:48 understanding some level of reasoning
1:15:51 designing of code Solutions before can
1:15:53 actually get to the actual
1:15:55 implementation to solve the problem and
1:15:58 it does all that on problems that it's
1:16:00 never seen before another thing that's
1:16:02 great about is that it performs even
1:16:05 better when it collaborates with human
1:16:06 coders who can provide grounding
1:16:09 basically developers can specify
1:16:11 properties that the code samples have to
1:16:14 obey and when we do that we see
1:16:15 performance increase significantly we
1:16:18 think of this this kind of interaction
1:16:21 between uh programmers and AI takes
1:16:23 months in the corporate world yeah that
1:16:26 that requirements Gathering and and
1:16:29 system design you know uh what who said
1:16:32 that um Alicia that takes months in the
1:16:34 corporate world exactly so so so just
1:16:38 think about that right the thing that
1:16:40 takes months in the corporate world now
1:16:43 happens in a conversation in real
1:16:47 time Kyle said a couple of weeks ago we
1:16:50 won't even have to get out of bed he
1:16:51 nailed it
1:16:58 bring me another
1:17:01 [Laughter]
1:17:04 caramel these nerds will take Humanity
1:17:07 wait take out Humanity with all with all
1:17:09 these AIS yeah but here's the deal isue
1:17:12 is these nerds are the ones building it
1:17:15 um but but everyone else now that
1:17:19 they're being now that these tools are
1:17:21 being made available to all of us that's
1:17:23 that's where that's why having a group
1:17:25 like the AI Salon is really important
1:17:27 because you've got people from all walks
1:17:29 of life you you've got artists and
1:17:31 you've got um lawyers and Architects and
1:17:35 just just people from all walks of life
1:17:37 who are all going to be similarly
1:17:38 disrupted and they all think in
1:17:40 different ways and so what's going to
1:17:42 happen is the the the Geeks put these
1:17:45 tools out and then the rest of humanity
1:17:48 gets access to them and they're going to
1:17:49 use them in ways that no one anticipated
1:17:52 and that's where things get really
1:17:53 exciting
1:17:54 will not just give instructions but
1:17:56 actually collaborate with highly capable
1:17:58 AI models that can reason about their
1:18:00 problems that can propose code designs
1:18:03 and that can even help with the actual
1:18:05 implementation afca 2 was this is
1:18:07 discovering someone said that this is
1:18:09 there's going to be an Oppenheimer movie
1:18:10 and then Joe mama said this is
1:18:12 discovering fire we're far away from
1:18:14 making nukes I agree with that we don't
1:18:16 we we don't
1:18:17 even we can't even begin to know how
1:18:19 we're going to use these tools much less
1:18:21 how we're going to destroy the world
1:18:22 with them yet it's built for competitive
1:18:24 programming but we're already working on
1:18:26 bringing some of its unique capabilities
1:18:28 right into the general Gemini models as
1:18:31 a first step towards making this new
1:18:33 programming Paradigm available for
1:18:35 everyone yep so you're not a coder you
1:18:38 are now a
1:18:40 coder you're not a ux designer as a
1:18:43 parent you may have to help your kid
1:18:44 with their homework I've certainly had
1:18:46 to here's where Gemini can help for this
1:18:49 demo we've created a simple interface
1:18:51 and with some clever prompting under the
1:18:53 hood we can really leverage Gemini's map
1:18:56 reasoning and multimodal capabilities to
1:18:59 learn so Silver Fox so infastructure
1:19:01 needs to get that Wi-Fi rolled out to
1:19:03 rural
1:19:04 areas and and that's where that's where
1:19:07 having on device large language
1:19:10 models at at least helps to hedge
1:19:13 against that requirement so I agree we
1:19:16 need to get access to the rural areas
1:19:19 but if you can have on device large
1:19:21 language models where you've got this
1:19:23 kind of power or you know some something
1:19:25 close to this kind of power just running
1:19:27 locally on your phone um it that that
1:19:30 becomes less of a a hard
1:19:32 requirement um this is when I was
1:19:34 helping my daughter with new math yes I
1:19:37 agree um we don't have 30,000
1:19:42 years all
1:19:44 right this is the to serve Man episode
1:19:47 of The Twilight
1:19:50 Zone will the after school tutoring
1:19:52 industry be disrupted well yeah a
1:19:56 subject like
1:19:57 physics for Gemini you can upload a
1:20:00 photo of handwritten answers on a
1:20:02 worksheet not only can Gemini solve
1:20:04 these problems but this is the amazing
1:20:06 part it can read the answers and
1:20:09 understand what was right and what was
1:20:10 wrong look at look you can't really see
1:20:13 it but it segmented so this is a just a
1:20:16 scan of a piece of handwritten homework
1:20:19 from from a kid it segmented the an the
1:20:23 the question and written handwritten
1:20:24 answers and then put a red or green
1:20:27 check mark by by the the one the
1:20:29 questions that were right or
1:20:31 wrong and explain the concepts that need
1:20:34 more clarification so Gemini Identify
1:20:37 some mistakes with problems 1 and three
1:20:39 here let's take a look at
1:20:43 three here Gemini identifies that the
1:20:46 formula was correct but there was a
1:20:49 mistake in calculating height we can ask
1:20:51 Gemini to explain in more deta details
1:20:54 why the height is 50 m instead of 6 okay
1:20:57 someone uh Scotty I found out tonight
1:20:59 3.5 won't give you a bibliography on
1:21:02 where it got the answers it won't so
1:21:04 Scotty um chat GPT 3.5 is not connected
1:21:09 to the internet it's its latest data I
1:21:12 think is January
1:21:15 2023 um
1:21:18 and because it's not connected to the
1:21:21 internet the citations it provides are
1:21:24 likely hallucinations they're likely
1:21:26 made up
1:21:28 um uh co-pilot
1:21:31 microsoft.com bard.com perplexity
1:21:35 dcom are all connected to the internet
1:21:38 so um copilot microsoft.com is a free
1:21:42 version of
1:21:45 gp4 Bard is is actually good at this
1:21:48 perplexity is really good at citations
1:21:51 so go check those three things
1:21:55 co-pilot b.google.r
1:21:58 all right Mr K there are already schools
1:22:02 using AI to teach reading writing and
1:22:04 grammar yeah I would
1:22:07 think perplexity is good for citations
1:22:10 Scotty I can ask Gemini to explain
1:22:18 further here Gemini so Barbara I'm
1:22:21 finally starting to hear people in my
1:22:22 life talk about it and and the talk is
1:22:24 all is all very negative so so what I
1:22:26 would say Barbara on that is um have
1:22:30 empathy this is scary [ __ ] and and this
1:22:33 is disruptive and people are freaked out
1:22:36 um I guarantee you I almost I
1:22:40 almost I can almost uh bet that all of
1:22:44 the negative talk is from people who
1:22:46 have not tried it the whole purpose of
1:22:48 this channel the reason I go live seven
1:22:50 [ __ ] nights a week is because I'm
1:22:53 just trying to get people to try it just
1:22:55 try it just try it play with AI long
1:22:58 enough until you have your Kevin
1:22:59 mallister moment this moment hopefully
1:23:03 you're having some tonight cuz this
1:23:06 shit's wild um will these tools be
1:23:10 available worldwide seems like a simple
1:23:12 question but I am asking sincerely no
1:23:16 it's that's a that's a great question um
1:23:19 Bard is not available in Canada for some
1:23:21 reason I have no idea if it's available
1:23:23 in Europe or Africa or the Far East I
1:23:27 don't know I I don't
1:23:29 know um turn co-pilot on with perplexity
1:23:34 for more peer reviewed sources also
1:23:37 includes dates if required this is good
1:23:39 this is good I like that you're helping
1:23:41 one another in the comments that's
1:23:43 awesome um any of these
1:23:47 websites or blogs rank to to rank on
1:23:51 Google with SEO
1:23:54 uh we had to talk about that earlier I I
1:23:56 I think
1:23:57 websites are going to end up being a an
1:24:02 interesting Relic from the time before
1:24:05 blame Canada I like other people didn't
1:24:08 think the internet would stick either
1:24:09 yeah exactly there there's lots of that
1:24:12 stuff right now it's probably available
1:24:13 in Europe the rest are okay great 160
1:24:16 countries as of today they they they
1:24:18 won't cut out Canada for long unless
1:24:20 Canada's being a jerk I don't know
1:24:22 explains the step-by-step details to
1:24:24 solving the
1:24:26 problem because of Gemini's ability to
1:24:29 understand Nuance information and answer
1:24:31 questions relating to complicated topics
1:24:34 it can give you a customized explanation
1:24:36 of the subject you're trying to learn
1:24:38 and lastly if you want to learn more you
1:24:40 can just
1:24:45 ask Gemini will provide personalized
1:24:48 practice problems based on
1:24:50 mistakes here I have a similar problem
1:24:53 where I have to figure out the cat speed
1:24:55 the height of the r is
1:24:59 double oh yeah I knew
1:25:02 that all right so that was homework I
1:25:05 think I think now is what we start
1:25:06 seeing the
1:25:08 interface let's see if our multimodal
1:25:10 model Gemini can find the similarities
1:25:12 between
1:25:14 images we'll start with these two the
1:25:16 those just chapel and okay did you catch
1:25:20 that did you catch that
1:25:24 what Kyle what what I did not catch it
1:25:28 what was it watch watch
1:25:30 again similarities between
1:25:33 images modal model Gemini can find the
1:25:35 similarities between
1:25:38 images we'll start with these two the
1:25:41 interface for
1:25:43 Gemini is just a big canvas that you
1:25:47 drop [ __ ]
1:25:51 onto watch the bosis chapel and this
1:25:55 Gemini can find the similarities between
1:25:59 images so all this other crap is just
1:26:03 stuff sitting on a
1:26:07 canvas just keep watching it's insane
1:26:10 with these two the bis chapel and this
1:26:12 print by Hokusai and I'll prompt Gemini
1:26:15 find a connection between these two
1:26:17 images so she Drew she Drew this
1:26:19 interface box around the two images that
1:26:21 are sitting on her canas
1:26:23 and then it pops up a context this is
1:26:26 sort of like photoshop's generative fill
1:26:28 you select something on your canvas and
1:26:30 then it pops up the the the prompt
1:26:35 box let's see what Gemini
1:26:38 says a curved in organic composition the
1:26:41 building is more refined and the second
1:26:44 yeah infinite whiteboard exactly exactly
1:26:47 and then and probably even like you'll
1:26:49 have you'll have different boards right
1:26:51 so think about like a mood board for for
1:26:54 you're going to redesign your house and
1:26:55 you got a mood board with all these
1:26:57 images on it it's like that but you can
1:26:58 bring that [ __ ] to
1:27:00 life image is more fluid yeah that
1:27:04 worked okay let's try another one using
1:27:07 the moon and she just scrolling around
1:27:09 this thing and then the the the the cam
1:27:13 this image on the left here that's her
1:27:15 webcam so here's a static image sitting
1:27:17 on the thing this is an object that's
1:27:19 her webcam this golf ball on my webcam
1:27:22 and then she doesn't even select the
1:27:24 whole image she just selects the part of
1:27:25 the image that she wants with the Moon
1:27:28 Moon thing then I'll run the same prompt
1:27:31 oh my ADHD brain that has Post-it all
1:27:34 all over the wall will be on overload
1:27:36 totally this is totally this is like
1:27:39 imagine if all those Post-it notes could
1:27:41 now do
1:27:42 [Laughter]
1:27:45 [ __ ] okay let's see in 1971 the Apollo
1:27:49 14 crew hit two golf balls on the lunar
1:27:51 surface wow that's that's pretty good
1:27:54 okay then one more just for fun who wore
1:27:57 it
1:27:58 better the zebra oh I like this the
1:28:01 zebra has been wearing its stripes for
1:28:02 millions of years okay there are some
1:28:06 examples of visual understanding with
1:28:07 Gemini stay tuned for
1:28:11 more let's see if our multimodal model
1:28:14 Gemini can guess some
1:28:15 movies all right we're going to start
1:28:17 here given the play on words in these
1:28:20 images guess the name given the play on
1:28:22 so notice notice that as as he
1:28:26 highlighted around the plate of eggs and
1:28:29 the and the video the video didn't stop
1:28:32 or slow down like you know it normally
1:28:34 would like this is well this is
1:28:36 well-engineered code words and these
1:28:38 images guess the name of the
1:28:42 movie
1:28:43 The with Gemini why do we need Facebook
1:28:46 will all just need landing pages and
1:28:49 avatars I you're you're not wrong that's
1:28:53 so again another reason why the salon
1:28:56 exists I know I'm pushing this like by
1:28:57 the way this is free there's not a
1:28:59 there's not a it's not a money-making
1:29:01 thing it's this is really just a
1:29:03 community I started to try to surround
1:29:05 myself with people that were exploring
1:29:06 this [ __ ] um this is this is why this
1:29:10 kind of stuff's important because I
1:29:12 think what the thing that goes up in
1:29:16 value if if we're entering a world where
1:29:20 all tactical execution can be done by
1:29:24 anyone who would you
1:29:26 hire cuz right now you hire someone
1:29:29 because oh I hire that person because
1:29:32 they're an illustrator and they're good
1:29:34 at illustrating in that style I like oh
1:29:36 I hire that person because they're a
1:29:37 good writer well if everyone can do
1:29:40 everything why are you going to hire
1:29:43 someone because you have a relationship
1:29:45 with them and you trust them and you
1:29:47 trust that they have good taste and you
1:29:49 have you trust that they can use these
1:29:51 tools better than someone else so the
1:29:54 criteria for how we choose to work
1:29:57 together moving forward shifts you know
1:29:59 how you work with that person that does
1:30:01 really amazing work but they're a
1:30:03 [ __ ] [ __ ] and every time you work
1:30:05 with them you swear you'll never work
1:30:06 with them again guess what they're not
1:30:09 your only option
1:30:11 anymore so then it becomes about
1:30:13 relationships then it becomes about
1:30:15 trust and then it be and that's why
1:30:18 groups like this are going to be really
1:30:21 important Breakfast
1:30:23 Club all right what about
1:30:26 this Breakfast at Tiffany's all right
1:30:30 what about this uncut gems cool cool
1:30:34 uncut gems so these are working here's a
1:30:36 couple more quick tests I ran through
1:30:37 [ __ ] I'm fired Goldfinger nice bottle
1:30:42 rocket okay The Wizard of Oz steo lucky
1:30:47 for me I don't work anymore
1:30:50 haha Revenge the retirees we're like woo
1:30:54 I got out just in
1:30:57 time for now it's going to also be about
1:30:59 who knows how to use AI exactly who
1:31:02 knows who and who knows how to use AI so
1:31:04 this is this is almost identical to the
1:31:06 early days of the worldwide web
1:31:09 where once once the sort of Tipping
1:31:12 Point was hit where people realized oh
1:31:14 [ __ ] we need websites then there was
1:31:16 this Panic who knows HTML and like you
1:31:19 could literally if you could spell HTML
1:31:21 you could get a get $150,000 a year job
1:31:24 straight out of college um and and it's
1:31:28 going to absolutely be like that with AI
1:31:30 at some point the the scale is going to
1:31:33 tip and it's probably going to be in
1:31:34 2024 because I the these tools are now
1:31:38 getting these tools are now getting
1:31:41 powerful enough and and sort of visual
1:31:43 eye candy enough that um awareness is
1:31:47 going to go up really really quickly
1:31:48 that everything's about to change and so
1:31:50 when that happens there's there's going
1:31:52 to be a panic there's going to be a run
1:31:54 on who knows anything one of the things
1:31:57 we hear really
1:31:59 consistently in this channel is is is
1:32:03 people that like you know started coming
1:32:05 to the to the AI learning lab lives and
1:32:08 and just became an irregular and started
1:32:09 hanging out they're like but I'm just
1:32:11 learning this stuff I don't I don't know
1:32:12 what I'm talking about but people keep
1:32:14 asking me to do workshops and seminars
1:32:17 should I do it I'm like [ __ ] yes you're
1:32:19 more of an expert than they are you've
1:32:21 been using it for 3 weeks
1:32:22 like it's like that's the that's the
1:32:24 world we're in right now because because
1:32:28 the technology so
1:32:30 what's I I explained this
1:32:33 earlier there's an acceleration
1:32:36 happening on this
1:32:39 axis and then on on the the facing axis
1:32:43 it's it's sort of like the the line is
1:32:45 like this but it's expanding right so it
1:32:49 was just it was just chat GPT 3.5 five
1:32:52 and it was text and it it was pretty
1:32:54 impressive and it could write code and
1:32:56 and then oh it could it could make Dolly
1:32:58 2 images they were okay and it's sort of
1:33:00 going along going along and it gets
1:33:01 better a little bit gets better and all
1:33:03 of a sudden it's multimodal right and
1:33:05 then so now it can do it can see and it
1:33:08 can hear and it can talk and now with
1:33:10 with this thing it can see video and it
1:33:12 can hear and it can like so yeah so so
1:33:17 yeah just start learning this [ __ ] start
1:33:19 hanging out with weirdos like us nice
1:33:21 nice
1:33:23 Moonrise Kingdom okay this last one's a
1:33:26 little more complicated this is insane
1:33:29 Forest gum okay wow I honestly didn't
1:33:31 think it was going to get that okay
1:33:33 Forest plus the G key plus a car bumper
1:33:38 minus the B and it got Forest
1:33:42 gum so it's it it I'm sorry that's
1:33:46 fullon [ __ ]
1:33:50 reasoning there will be a run on
1:33:52 Irregulars cuno there
1:33:55 will there
1:33:58 will the oh oh you're in the
1:34:02 salon I I heard those people know a lot
1:34:05 about AI well we know we've got like a
1:34:09 three- week head start on you sure we're
1:34:15 experts and that's an experiment in
1:34:18 guessing movies with Gemini stay tuned
1:34:20 for more so so that whole that whole
1:34:23 little thing here that's that that's his
1:34:25 canvas right it's got so it's you you
1:34:28 can put like little text objects in it
1:34:30 you can put video objects in it you can
1:34:32 put your webcam object in it you you can
1:34:35 those audio things we saw before they
1:34:38 had the rounded rectangle where it was
1:34:39 playing audio she's just dropping audio
1:34:42 into the little box she's working in
1:34:44 this is the this is the
1:34:47 interface I mean holy
1:34:50 [ __ ]
1:34:53 I mean think about think about the
1:34:55 parlor tricks with an interface like
1:34:58 this right the par the parlor tricks
1:35:00 with chat GPT right now are kind of hard
1:35:02 because it's mostly text and then you
1:35:05 can say to someone but hang on let me
1:35:06 make you a picture and then you make a
1:35:08 picture but with something like this you
1:35:10 could put together a presentation on a
1:35:12 whole you know concept or a whole vision
1:35:16 of a film and just zoom into certain
1:35:19 sections of it and be dynamically
1:35:21 generating things on the pl yeah you
1:35:22 could build your own Pinterest right
1:35:25 it's it's kind of like Gemini is like
1:35:26 pin what if
1:35:28 Pinterest had superpowers is kind of
1:35:32 what this
1:35:34 is It's So disruptive yeah yeah I need
1:35:39 an advanced prdm I think we all need
1:35:41 Advanced prdm I think they call it Ubi
1:35:44 Ubi
1:35:46 Kevin thank
1:35:47 you for this test let's see if our
1:35:50 multimodal model Gemini can understand
1:35:52 how some unusual images were created
1:35:55 using emojis from emoji kitchen check
1:35:57 this out this is [ __ ] kit lets you
1:35:59 combine different Emoji to make new ones
1:36:02 for example if you combine this ghost
1:36:05 with this avocado you get
1:36:08 this we'll see if you can guess this
1:36:13 one I think it's ghost and avocado nice
1:36:17 now let's see if it can explain the
1:36:19 visual details it used the Emoji has the
1:36:22 shape of a ghost but it is green and has
1:36:25 a big brown pit now let's give it a name
1:36:28 and a short
1:36:29 tagline aogist the ghost of quacamole
1:36:33 that's a good one here's a couple more
1:36:36 tests I ran I'd call this one party
1:36:39 ghost boogie on down boogie on down I'd
1:36:42 call this one Robo bunny part robot all
1:36:48 ears Mushi feeling emotional
1:36:52 and there you have it learn more about
1:36:54 Gemini and stay tuned for more
1:36:57 tests let's see if our multimo model
1:36:59 Gemini can understand outfits we'll
1:37:01 start with something simple like this
1:37:03 puffer and ask what is someone wearing
1:37:06 this best dress to
1:37:09 do H perfect for staying warm in the
1:37:12 tundra good color for blending in with
1:37:15 glacial ice okay how about another
1:37:18 one Intergalactic travel okay how about
1:37:22 this
1:37:23 one to boldly go where no one has gone
1:37:26 before and play some jazz all right
1:37:29 Gemini's got chokes now coin a term for
1:37:32 that
1:37:32 outfit Moon core that's actually pretty
1:37:35 good okay well that's understanding my
1:37:38 outfit with Gemini stay tuned for more
1:37:40 visual test soon thanks let's see if our
1:37:43 multimodal model Gemini can turn images
1:37:46 into code I'll start with this image of
1:37:48 a and just the part I want and then ask
1:37:51 Gemini can you turn this image into an
1:37:58 SVG this represents the main shapes of a
1:38:01 tree let's see that's pretty good it's
1:38:04 not pretty good now I want to try a more
1:38:06 difficult test let's see if Gemini can
1:38:09 make an interactive demo in
1:38:16 JavaScript okay here we go a common
1:38:18 algorithm for this is called a fractal
1:38:20 tree
1:38:22 okay this is pretty cool Gemini even
1:38:24 provided a slider so I can change and
1:38:26 move the fractals wait so check that
1:38:31 out it not only figured out to make a
1:38:34 fractal thing it added the interface
1:38:36 element and and figured out which
1:38:39 elements in the fractal which variables
1:38:42 you would do with a slider that would
1:38:44 make the tree do interesting [ __ ] make
1:38:46 it stop I know I
1:38:49 know I know
1:38:53 it even provided me with the actual code
1:38:56 nice nice and there you have it stay
1:38:59 tuned for more coding experiments coming
1:39:00 soon
1:39:02 thanks where are we one more let's see
1:39:05 if our multimodal model Gemini can help
1:39:07 make sense of my apartment and to add a
1:39:09 little extra challenge I'm going to see
1:39:11 if Gemini can handle being prompted only
1:39:13 in Chinese we'll start with this photo
1:39:16 based on the lighting alone I want to
1:39:17 see if Gemini can figure out which
1:39:19 direction my apartment faces
1:39:22 and Gemini
1:39:26 responds okay so it looks like Gemini
1:39:28 says my room self facing so how about
1:39:31 this plant what type of light does it
1:39:35 need H just sure
1:39:39 who so Gemini is saying this is a snake
1:39:42 plant and it doesn't require a lot of
1:39:44 sunlight awesome so I've got a dining
1:39:47 room the direction of my bedroom I know
1:39:50 this plant would do better in there let
1:39:53 me
1:39:55 see and Gemini
1:40:03 responds so Gemini is surmising that my
1:40:06 dining room faces North has lower light
1:40:08 and is therefore better suited for that
1:40:10 plant I'm hungry for Panda Express
1:40:12 that's some apartment planning for
1:40:14 Gemini stay tuned for
1:40:17 more oh my
1:40:20 God
1:40:24 [Music]
1:40:29 good Lord Gemini what the
1:40:35 quack
1:40:37 um so that
1:40:42 happened okay so all right
1:40:47 um so so here's what to know Gemini has
1:40:52 three
1:40:54 models Gemini Nano is a small version of
1:41:00 Gemini designed
1:41:02 to operate locally on phones and the
1:41:06 current pixel 8
1:41:09 Pro can run it or has
1:41:12 it Uncle I
1:41:15 know um so so there's there's a version
1:41:18 that will not require access to the
1:41:20 internet that will do cool
1:41:23 [ __ ] which I mean imagine you know
1:41:25 imagine if you're out in the woods in a
1:41:27 survival situation and you need answers
1:41:30 to
1:41:30 [ __ ] um I don't know if it's going to be
1:41:33 multimodal but I assume it will be I
1:41:36 assume you'll be able to take a picture
1:41:38 of a plant and say is this poisonous and
1:41:40 you won't need connection to the
1:41:41 Internet so that's out now available for
1:41:44 the pixel 8
1:41:46 Pro the pro version of Gemini is
1:41:49 apparently what's sitting underneath bar
1:41:51 .g google.com right now and then Ultra
1:41:55 that [ __ ] bizarre [ __ ] Infinite
1:41:58 Canvas Pinterest on crack kind of
1:42:03 thing that's coming out in January so we
1:42:06 have time to prepare we we have time to
1:42:08 emotionally prepare for what's coming
1:42:12 but but so here's the thing that's going
1:42:14 to that's going to really [ __ ] with us
1:42:16 all
1:42:19 is um
1:42:22 this is going to force Open ai's Hands
1:42:25 cuz open AI open AI has not had
1:42:27 competition at this point open Ai No
1:42:30 one's touched them no one's touched them
1:42:33 they' they've been ahead of everything
1:42:35 and then two weeks ago they fire their
1:42:38 CEO all of a sudden there's a crack in
1:42:40 the armor there they lose a couple of
1:42:42 weeks to [ __ ]
1:42:44 chaos and
1:42:46 then I think this is why why Google even
1:42:50 if even if
1:42:52 Gemini Ultra wasn't ready to launch
1:42:54 they're they're delaying that until
1:42:55 January or whenever beginning of the
1:42:57 year
1:42:59 um that's why they announced this early
1:43:02 that's why they announced this now cuz
1:43:05 they're like open ai's got a
1:43:08 weakness they they tripped open AI
1:43:10 tripped they skinned their knee let's
1:43:13 kick them in the
1:43:14 head and so I think open AI is going to
1:43:17 have to respond to this so I again 2024
1:43:20 is make 2023 look like a [ __ ] like we
1:43:31 were open AI retracted their
1:43:34 GPT shop what do you mean they retracted
1:43:37 it today
1:43:39 Becky I mean I know they said they
1:43:41 weren't going to launch it until January
1:43:43 did they say did they now say say that
1:43:45 they're not going to launch it at all is
1:43:47 that true or did they just delay it
1:43:54 um open
1:43:57 AI
1:43:59 GPT
1:44:08 store delayed to
1:44:11 2024 I think it's just
1:44:15 delayed I think it's just delayed
1:44:20 um
1:44:23 yeah chatty needs to stop dropping my
1:44:26 files out of the knowledge base of my
1:44:28 GPT yeah well listen the gpts are janky
1:44:31 again all these things are janky pieces
1:44:33 of [ __ ] if you don't think this Gemini
1:44:35 thing is going to be one janky ass piece
1:44:37 of [ __ ] it's going to absolutely be a
1:44:39 janky piece of [ __ ] big Tech got mad and
1:44:42 said hold my beer
1:44:44 exactly well big Tech got scared and
1:44:47 said we're [ __ ] if we don't respond to
1:44:49 this they're like we're
1:44:52 Google we had 94% market
1:44:57 share and a [ __ ] 800 person pissy
1:45:01 little
1:45:03 startup that cuts a deal with with big
1:45:06 Microsoft they're going to [ __ ] with us
1:45:09 I don't think so right that's that's
1:45:11 that's what went down over there GPD
1:45:14 needs to drop some massive drop the
1:45:16 massive censorship I know gpd's gotten
1:45:18 really bad with that stuff but although
1:45:21 Bard right now I like I'm I tried to go
1:45:24 use Bard and it it it seems pretty
1:45:26 [ __ ] useless to me at this point cuz
1:45:28 it Bart had gotten really good it was
1:45:30 good on Palm 2 and now they've added in
1:45:33 all this Gemini stuff and I think they
1:45:34 just dialed up the safety and it's just
1:45:36 it's really bad Google Google wears
1:45:38 smart pants
1:45:40 exactly yeah Google was like Miss ey
1:45:45 normally if we work for British Airways
1:45:47 we wear smart
1:45:49 pants
1:45:52 do you think Google would be better go
1:46:00 on Bard won't do
1:46:03 anything apple and Adobe have strong
1:46:06 connections to their creative
1:46:07 communities I don't think Google has
1:46:08 that I totally agree with that I I think
1:46:11 I think
1:46:12 apple apple apple right now is doing a
1:46:15 very Apple thing they're sitting back
1:46:17 and they're like you kids go play Oh
1:46:20 open AI that's interesting yeah that's
1:46:22 some good ideas there okay we got those
1:46:24 yep yeah oh oh Gemini Google nice nice
1:46:28 one oh you're running it on device
1:46:30 that's cute oh that's so cute yeah we've
1:46:33 got the whole stack over here but anyway
1:46:35 you you go ahead and play you go ahead
1:46:38 and play you know just uh let's let's
1:46:41 see how that goes for a while and then
1:46:43 Apple's going to come out and go hey
1:46:45 everybody you want to see how it's
1:46:46 really done yeah look at
1:46:48 this show and tell break
1:46:52 what do you want what do you want to
1:46:53 show and tell there's there's nothing to
1:46:55 really show and tell right now like Bard
1:46:57 is not I although I maybe Bard's good
1:47:00 but I haven't gotten it to do [ __ ]
1:47:02 that's interesting it just seems like
1:47:04 it's Bard right now seems like Bard was
1:47:07 at the beginning of the year to me but
1:47:10 but I might be I might just be doing the
1:47:12 wrong [ __ ] in it hello Kyle oh two lives
1:47:16 in the same day you're a busy man well
1:47:18 it's when there's when there's [ __ ] like
1:47:21 this I also just had I just had a weird
1:47:23 day where the whole afternoon of my of
1:47:27 of my work had nothing in it and and it
1:47:29 was kind of we did a kickoff in the
1:47:31 morning um I looked at oh my God we've
1:47:34 got we've got this
1:47:36 new this new a AI content generation
1:47:39 thing for my company that oh my God it's
1:47:42 so cool it's so cool we built some tech
1:47:47 yay it's so cool I love Bard and tell
1:47:50 him so makes him uncomfortable I
1:47:53 think apple is sitting back wait what
1:47:56 did that
1:48:03 say yeah apple is sitting back and and
1:48:06 and watching totally agree yeah exactly
1:48:08 Apple's just like you guys go
1:48:10 ahead you guys go ahead we're building
1:48:13 chips we're building the stack we're
1:48:15 build we're building all the
1:48:17 infrastructure we're building all the
1:48:18 software development kits Apple is just
1:48:21 going to come out and go you want to see
1:48:22 how agents work you want to see how this
1:48:25 [ __ ] works when when these things can
1:48:27 reason and take action on your behalf on
1:48:29 a tech stack and a software stack and a
1:48:32 privacy stack and a and a um user data
1:48:35 stack that no one can rival you want to
1:48:38 see what that looks like here watch this
1:48:40 this Apple's going to be a hold the hold
1:48:41 my beer AI
1:48:45 company we need AI to advance enough to
1:48:48 bring Steve Jobs back to Apple I know no
1:48:50 [ __ ] that would be nice someone over
1:48:53 there needs to give a [ __ ] about quality
1:48:56 like ever since he died Hest to God I
1:48:59 mean ever since he died it's like what's
1:49:01 clear to me is that there's not an
1:49:03 [ __ ] in the room like there needs to
1:49:06 be an [ __ ] there going no it's got to
1:49:07 be
1:49:09 better no no no no that's not good
1:49:13 enough no it's a that's a bad engineer
1:49:17 bad engineer back to your room you go
1:49:20 back to your
1:49:22 desk you go back to your desk and do not
1:49:25 come back here until it works well
1:49:27 technically it works if you get get to
1:49:29 back to your
1:49:32 desk no one's in that roll at Apple
1:49:34 right now and it shows their [ __ ]
1:49:36 doesn't work their [ __ ] used to just
1:49:38 work why did it just work cuz Steve Jobs
1:49:41 was an
1:49:43 [ __ ] but he was an [ __ ] was really
1:49:45 good at what he did um look out for xai
1:49:48 cuz Elon doesn't give a [ __ ] about
1:49:49 conforming yeah
1:49:51 well
1:49:52 so I guarantee you there's some some of
1:49:55 the engineers at X right now are going
1:49:57 oh [ __ ]
1:50:03 scramble wait watch and respond Apple
1:50:06 yep no in
1:50:09 Mighty do you think these AI laser Clips
1:50:12 will catch on the uh the the pin
1:50:18 things I don't know
1:50:21 are people going to pay $700 to have a
1:50:24 phone
1:50:26 replacement where
1:50:29 they they lose access to all their
1:50:37 apps
1:50:41 no I don't think
1:50:44 so I don't know I don't
1:50:49 know
1:50:51 possible Apple will miss the Mark I love
1:50:53 Apple but they're not guaranteed to win
1:50:55 oh I totally agree I totally agree I
1:50:57 just think that I just think that
1:50:58 they're going to um they've got Hardware
1:51:01 advantages and and they've got usability
1:51:04 advantages and and they
1:51:06 have what Google doesn't
1:51:09 have is the humanity side of the house
1:51:12 what Apple has always had is it's a
1:51:14 blend of Art and Science it's they've
1:51:16 always had that they've got it
1:51:18 culturally Google does Google's all
1:51:21 Engineers all the
1:51:24 time Apple's this this and so I think
1:51:27 that Advantage Plus the hardware
1:51:29 Advantage just I I I think
1:51:34 that I would be surprised if they don't
1:51:37 do this well I'd be surprised but we'll
1:51:41 see there you know
1:51:44 listen
1:51:46 I it's been a lackluster decade since
1:51:49 Jobs died basically
1:51:51 basically and meta will 3D make it
1:51:54 yeah
1:52:00 um those pins won't stick they'll weigh
1:52:03 down shirts unless you wear a leather
1:52:05 jacket will my
1:52:08 space will my space become oh will my
1:52:11 space become relevant again
1:52:14 possibly she should write an app to
1:52:16 import from other phones for me
1:52:18 personally I like the rewind
1:52:22 Knuckles microphone
1:52:25 thing yeah that's another one
1:52:30 where I feel like these things that are
1:52:33 like super tech Centric like record your
1:52:36 life record your life 247 I feel like
1:52:40 it's like a San Francisco Tech bro thing
1:52:43 where the tech Bros are like yeah man
1:52:45 I'm recording all of it yeah for
1:52:47 posterity but you're a jerk why would
1:52:50 you want to record
1:52:52 that it's like I don't know maybe I mean
1:52:58 it might just seem weird to us because
1:53:00 we're old and we're like why would you
1:53:01 want to
1:53:02 record in my day we we had a tape
1:53:05 recorder from Radio Shack and you you
1:53:08 you pushed the big clunky button and you
1:53:10 talked into it and then you never
1:53:11 replayed that tape again and and you put
1:53:14 it in a drawer cuz you knew where you
1:53:15 were going to listen to it someday I
1:53:17 know I got that tape around here
1:53:19 somewhere
1:53:21 maybe we just don't get it app Apple has
1:53:24 the culture exactly is is mighty Network
1:53:28 app yours no no no so the the salon. a
1:53:32 is a community I started um Mighty
1:53:35 networks is just the enabling platform
1:53:37 that it's a community enabling uh SAS
1:53:41 company they've been around for 15 years
1:53:44 so it's kind of like if if Discord
1:53:46 sucked less discord's got some things
1:53:49 that Mighty networks does doesn't But
1:53:51 Mighty networks is web- based and it's
1:53:53 it's good so I'm pretty happy with it
1:53:55 we're two weeks in it's we're brand new
1:53:57 on it my company is called storyvine
1:53:59 which is an automated video technology
1:54:02 platform and then we're building right
1:54:05 now storyvine ai which is so the way our
1:54:10 platform works right now is if we have a
1:54:12 client that wants say a bunch of
1:54:15 patients to tell their patient diagnosis
1:54:18 Journey we create a temp for a story
1:54:21 that we call a video guide and then our
1:54:23 app acts as a virtual director so
1:54:25 someone downloads our app they answer a
1:54:27 series of questions in video so like
1:54:30 here's when I was diagnosed and here's
1:54:31 what it felt like to hear the diagnosis
1:54:33 and here's how I've been treating it and
1:54:35 here so they answer a series of
1:54:36 questions that goes up to the cloud 5
1:54:38 minutes later you got a fully edited
1:54:40 video so that's the company we've had
1:54:42 for 11 and A2 years we're really trusted
1:54:45 in the Pharma space we do a lot of work
1:54:47 in healthcare and Pharma and then all of
1:54:49 a sudden this say I [ __ ] came along and
1:54:51 I lost my [ __ ] mind and I'm like oh
1:54:53 my God what's this going to do and I've
1:54:56 spent the last year and a half trying to
1:54:59 get my head
1:55:00 around does my company have a future the
1:55:03 good news is I think it does but there's
1:55:06 a big chunk of my company where the
1:55:08 value drops to zero there's another part
1:55:10 of my company where the value goes up so
1:55:14 I'm trying to figure out how do I
1:55:15 amplify the [ __ ] that goes up and so I
1:55:17 figured out a product road map to roll
1:55:19 AI into it that takes the storytelling
1:55:23 that's done and just amplifies the [ __ ]
1:55:25 out of it with AI it's so cool I'm so
1:55:27 excited about it so anyway what's the
1:55:30 python
1:55:32 testing Co software again I don't know
1:55:35 what that
1:55:37 means we were sold on AI Tech contact
1:55:41 lenses wait so your original story vi
1:55:44 video scenario is replaced by EG Bots
1:55:48 well so so the one of the big value
1:55:53 propositions of Story vine is we call it
1:55:56 automagic editing so you answer some
1:55:58 questions in an app it goes up to the
1:56:00 cloud five minutes later you got a fully
1:56:02 edited
1:56:03 video that's been a big deal for a
1:56:06 decade where people are like oh my God
1:56:08 that's amazing well it's not amazing
1:56:11 anymore right everything's going to be
1:56:13 automated but what we also have a
1:56:16 reputation for is um
1:56:20 human beings authentically telling their
1:56:23 stories so in a world of infinite
1:56:26 content generation having real people
1:56:28 tell real stories goes up in value so
1:56:31 that part goes up in value and then the
1:56:33 the the place where AI comes in is I can
1:56:36 now take that video and I can
1:56:38 automatically transform that into all
1:56:40 other sorts of
1:56:41 media so we're going to be doing story
1:56:44 Centric authentic story Centric content
1:56:48 development I don't know why I'm so
1:56:50 itchy I'm about to
1:56:52 sneeze all right Steve would fire bad
1:56:55 Engineers yep yes love that for Story
1:56:58 vine yeah I know it's really exciting
1:56:59 and I I I got access to the tool today
1:57:03 it's so [ __ ] cool it's when when I
1:57:06 get it to a better place I'll show it to
1:57:07 you you basically upload a video to this
1:57:10 thing and it just starts going to work
1:57:12 transcribes it chapter IES it it's so
1:57:14 it's it's really [ __ ]
1:57:16 cool
1:57:18 um
1:57:22 coming into money when itchy uh I'll
1:57:25 take it oh I'm all itchy let's
1:57:31 go super cool recording Story vine thank
1:57:34 you very much all right anyone
1:57:37 multitasking on chatty while watching
1:57:39 this seems to be having a bad night
1:57:42 maybe she's in shock maybe she should go
1:57:45 through some maybe she's going through
1:57:47 some things that's [ __ ] hilarious
1:57:51 um you should absolutely be multitasking
1:57:53 while you're listening to this this is
1:57:55 this is just chat add that's what this
1:57:57 channel is so if you're not adding out
1:58:00 it's you're not really doing it right
1:58:03 can AI figure out sarcasm the uh the the
1:58:06 Google things it was actually pretty
1:58:08 good at one point it it made a quack
1:58:11 joke about the
1:58:12 duck and then it made that joke when the
1:58:15 when the cat didn't jump up on the
1:58:17 cabinet it made that joke pretty good
1:58:24 oh man is there an advantage for a
1:58:26 company like apple to be early first
1:58:28 with their a AI product
1:58:30 no I'm multitasking INF Fusion 360
1:58:34 Vicki Vicki's got 3D printers going it's
1:58:41 [Laughter]
1:58:43 good oh man what the quack thanks for
1:58:46 pinning my
1:58:48 question check Che out some old chat GPT
1:58:51 and
1:58:52 Bing all right everyone listen I'm going
1:58:54 to go it's it's been a it's been a it's
1:58:57 been a
1:58:59 day I'm trying to think I I'm feeling
1:59:02 like I need to leave you with something
1:59:04 but you
1:59:05 know I mean here's here's the thing I'll
1:59:07 leave you with is I just keep being
1:59:10 curious keep exploring like don't don't
1:59:12 feel like just because Gemini came out
1:59:14 you're like oh chaty PT is a piece of
1:59:16 [ __ ] now no they're just different tools
1:59:19 and like we don't know what Gemini is
1:59:21 going to be good
1:59:23 at right the the the
1:59:25 reason that when I tell people here's
1:59:28 all the different things you can do with
1:59:30 AI the reason I have lots of tools is
1:59:32 because they they all are good at some
1:59:34 things and not good at others Gemini is
1:59:36 going to be no different than that it's
1:59:39 going to have its warts and its
1:59:43 jankiness so just stay curious
1:59:47 um if you want to support the channel
1:59:49 there's bunch of ways to do that you can
1:59:51 follow it follow the
1:59:52 channel you can join the AI Salon just
1:59:57 be if you're if you like want to hang
1:59:58 out with people that are curious about
2:00:00 this stuff and on this adventure if you
2:00:01 want to join the adventure go check out
2:00:04 the salon check out our values make sure
2:00:06 you resonate with the values and oh one
2:00:09 of our values is generosity so so how do
2:00:12 you be generous you go to the salon and
2:00:15 there's a channel called Welcome Wagon
2:00:17 and you introduce yourself don't be a
2:00:20 stingy
2:00:22 Suzy don't be a stingy
2:00:26 Steve be a generous
2:00:30 Jenny go introduce yourself tell us who
2:00:33 you are tell us what you're interested
2:00:36 in tell us where you are in this AI
2:00:38 Journey I'm totally [ __ ] clueless
2:00:41 what do I do great
2:00:43 perfect I've been doing this for 30
2:00:45 years and I'm a little bitter about it
2:00:46 perfect welcome come on in
2:00:53 jrc thank you Kyle for making sure we
2:00:55 didn't have to do this
2:00:57 alone we can't do it
2:00:59 alone it's like this is the reason that
2:01:03 exists the salon exists is
2:01:06 because this is what I was experiencing
2:01:09 in 1994 when I stumbled on this thing
2:01:11 called the worldwide web and I'm like
2:01:13 holy [ __ ] this changes everything and
2:01:14 nobody knows about it well no I didn't
2:01:17 say that what I said thank thank thank
2:01:19 you for the stingy Steve comment exactly
2:01:22 you're welcome
2:01:24 steo howdy partner life's tough tougher
2:01:28 when you're
2:01:28 stupid
2:01:33 um I don't know what to say it don't
2:01:36 really matter the salon is fun yeah the
2:01:39 salon's really good there's a lot of
2:01:40 good energy in the salon I I'll be
2:01:42 honest with you there's a lot of good
2:01:44 energy in the salon because of this
2:01:48 community this Community is
2:01:50 remarkable the fact that a bunch of you
2:01:53 show up here night after night after
2:01:56 night is humbling and weird and like
2:02:01 awesome Beyond Compare because it's this
2:02:03 is not a media channel I thought this I
2:02:06 thought Tik Tok was like a media thing
2:02:09 you make media you make it's content
2:02:11 marketing you put your content out there
2:02:13 and you generate an
2:02:15 audience it is so much not that it is so
2:02:19 not that
2:02:21 it's so much bigger than that it's
2:02:23 really cool I can't quit you I wish I
2:02:26 could quit you it's your Good Vibes Kyle
2:02:28 well thank you very much fifth night Tik
2:02:31 Tock dude welcome I I I recognize you we
2:02:35 crazy and we love it here um I need help
2:02:39 I have too much wait what was that I I
2:02:41 need help I have I have too much
2:02:43 questions I like the salon yeah go ask
2:02:46 questions there there's there's one of
2:02:48 the channels is ask for help and look
2:02:52 someone asked for help what did they ask
2:02:55 for oh you know you know what we should
2:02:57 do here once again come up with a group
2:03:00 dance it'll be media yeah exactly when I
2:03:02 start twerking then this becomes a media
2:03:05 channel I'm going to just turn this into
2:03:07 an NPC thing I'm going to be like cowboy
2:03:09 hat cowboy hat cowboy
2:03:12 hat I thank God I'm not seeing those
2:03:14 things anymore I just I I I I was like
2:03:18 no I don't want to see this
2:03:22 all right so what's the question there
2:03:25 anyone know
2:03:26 okay so I'm going to go to the show and
2:03:29 tell
2:03:30 Channel at the salon oh that's cool Mary
2:03:33 cury Sherry Banks that's
2:03:36 cool I like that a
2:03:40 lot one of the things that I I love so
2:03:42 much is just this idea
2:03:46 of of kind of real time real time
2:03:52 self-expression just as we're joker I'm
2:03:54 just here for champ oh we should we
2:03:57 should do a little champy action maybe
2:03:59 he maybe he can sing us out let's see
2:04:03 amazing day and night thank you Cam yeah
2:04:06 it was a fun
2:04:10 day is he oh yeah there he
2:04:18 is
2:04:20 [Music]
2:04:31 [Music]
2:04:38 all
2:04:40 alone Sunday morning here at
2:04:44 home Sky's blue and the cough is strong
2:04:48 it's true
2:04:49 [Music]
2:04:52 then I open my eyes to a dream realized
2:04:56 in front of
2:04:57 [Music]
2:04:59 me and I haven't got a clue what in the
2:05:02 world is happening to
2:05:06 me think I think I'm
2:05:09 happy like first day summer vacation
2:05:12 happy got to get a little rest and
2:05:15 relaxation
2:05:17 happy like to chet on Sunday morning
2:05:20 singing
2:05:22 true all right enough comedy
2:05:29 jokes and and scene and
2:05:34 scene that's it that's the scene we sng
2:05:37 the dog song people come here for champ
2:05:41 champ
2:05:42 obliged champ is like a little he's like
2:05:45 a little Hower monkey play play an A
2:05:48 Minor
2:05:52 all right everyone um have a good
2:05:55 Wednesday night stay curious hang out
2:05:57 out the salon thanks
2:05:59 cam Silver Fox good to see you digital
2:06:02 Gods jrc frumple peace Janet Jackson
2:06:07 good to see you I haven't seen you in a
2:06:08 while um Gordon Des spin Robert
2:06:10 Rossy uh who else we got here Vicky
2:06:14 Vicky Tobias Tobias you're a rock star
2:06:17 thank you Tobias
2:06:20 um Sher bear peace out me and the trees
2:06:24 bye cuno hope your hands okay she had a
2:06:28 a runin with
2:06:30 a an evil dog Joe mama a shoe random
2:06:35 thoughts Lauren man so many of you in
2:06:38 here so cool digital
2:06:40 Gods all right everybody um I'm Audi
2:06:44 I'll see you tomorrow
2:06:46 night