1/13/2025 - AI Agents & the Paranet Protocol: Orchestrating the Future of AI

Description

Chapters

Transcript

0:03 [Music]
0:09 d
0:26 [Music]
0:39 [Music]
0:56 standing
0:58 between you and H is
1:01 [Music]
1:04 insane standing too
1:07 near you and a file makes it
1:12 [Applause]
1:14 clear your trouble to
1:19 [Music]
1:21 meble can't you
1:26 see leaning in
1:28 clothes smelling your perfume that
1:31 scares me
1:35 most leaning
1:38 awayy it feel stronger every
1:44 day your trouble to
1:50 me Rich can't you
1:56 say good evening good people
1:59 [Music]
2:07 she came on him like slow moving
2:15 C his be was warmer than a look in her
2:22 eyes she said on a stool and he said
2:26 what do you want
2:30 [Music]
2:33 she said give me a love that don't
2:36 freeze up
2:39 [Music]
2:41 inside he said I have miled it some
2:44 hearts in my time
2:49 dear but to sit next to you will I
2:52 shiver and
2:57 shade and if I knew love well I I don't
3:00 think I'd be
3:02 [Music]
3:05 here asking myself if I've got what it
3:12 t to melt your icy
3:17 blue I don't know that to
3:20 me no whatever doesn't [ __ ] matter
3:23 good evening happy
3:24 [Laughter]
3:28 Monday have you seen that AI learning
3:31 lab you know it's interesting he's got a
3:33 dog that sings the dog sings beautifully
3:35 he's always on key it's just a fantastic
3:38 situation well with the exception of the
3:40 singer the the man part of it and but
3:43 anyway so he s he starts the song and
3:45 it's like he's he's singing along and
3:47 he's singing along everything's fine and
3:48 then just like doesn't know the song
3:51 it's like well why would you play the
3:52 song if you don't know the song it's
3:54 just I don't understand could someone
3:57 explain to me why people watch this
4:01 I can't either I'm right there with you
4:03 listen we got it's a whole [ __ ]
4:05 situation it's a whole situation I you
4:09 know I've got a dog that's under
4:11 contract he has to sing so I can't not
4:14 sing um I know seven songs and so you
4:18 know we we sing what we sing and then I
4:21 know 150 almost songs
4:26 [Laughter]
4:32 oh good lord good
4:35 people there's so much we could talk
4:37 about tonight I don't even know where to
4:40 begin I don't even know where to begin
4:43 oh I wait so so first of
4:49 all let me let me flip my little view
4:52 over here on the Tik Tock hi ticktockers
4:54 now I'm talking to everyone at the same
4:56 time
4:58 um t to what what's the date 13th six
5:02 more days I heard today R rumor today is
5:06 Elon Musk is going to buy Tik Tok Mr
5:08 Wonderful's trying to buy Tik Tok
5:10 Supreme Court's supposed to weigh in
5:16 Wednesday Tik Tock is
5:19 it I have I have avoided the Temptation
5:22 so far of listen this platform's going
5:26 to go away and I'm going to miss all of
5:28 you
5:30 so I'm gonna I'm gonna give another day
5:33 or two see see if some things start
5:35 playing out but uh a lot of lot of lot
5:38 of heartbreaking sadness right now
5:41 everyone's running over to red
5:49 note so red note if you don't know it is
5:53 the actual Tik Tock owned by the actual
5:57 Chinese government in Chinese it's a
6:00 available on the app store for some
6:01 reason uh apparently there's been a mass
6:05 uh
6:06 downloading red note is now like the
6:08 number one or two app on the Apple iOS
6:11 app store because 170 million people
6:15 were like [ __ ] it let's go to Red note
6:18 you don't want us on a Chinese app why
6:19 don't we go on the one that's actually
6:21 spying on us and people are logging in
6:24 there going where's my Chinese spy I
6:26 want to meet my Chinese spy
6:34 oh good lord so that's happening um let
6:39 me get comments up and running here I I
6:42 am without a producer tonight so red
6:45 notes not red noes red as in the Red
6:50 Army red
6:55 note paranoid people from the 1950s are
6:58 flipping in their grave right now what
7:00 do you mean you're downloading the
7:01 Chinese
7:03 [Music]
7:18 app
7:20 um if if you don't know me if you're
7:22 brand new here God love you God love you
7:25 you're just you're just tutored in a
7:26 little button my name's Kyle Shannon
7:28 this is the AI learning lab we we do all
7:31 sorts of things here if you're here on
7:32 LinkedIn you might think like well this
7:34 doesn't seem very professional you're
7:36 correct uh this is a this is a
7:40 professional
7:42 uh cell phone on my part uh I should not
7:46 be streaming this on LinkedIn but I have
7:49 ADD and I'm lazy and once I got it set
7:51 up it's easier just to keep doing it
7:53 than not so I figure people on LinkedIn
7:55 who know of me as a upstanding
7:58 professional
8:03 oh they'll get over that they'll they'll
8:05 find out soon
8:06 [Music]
8:13 enough and for those of you not on
8:15 LinkedIn if you're on the YouTube that's
8:18 going to be the primary home of all this
8:21 [ __ ] um starting you know if if Tik Tok
8:25 goes away um but right now I'd say it's
8:28 kind of 5050 anyway we're we're
8:30 generally uh streaming live on Tik Tok
8:33 over here and the other three platforms
8:36 X and Linkedin and Alan hilbeck is over
8:41 there on LinkedIn holding down the fort
8:43 for the professionals very nice I like
8:46 it
8:49 [Music]
9:13 all right Vicki's putting up some good
9:16 stuff so don't if you're on YouTube
9:18 couple of things we need there we need
9:20 plays and we need follows subscribers
9:24 doesn't cost anything by the way I
9:25 realized I've been asking people to
9:27 subscribe to YouTube and there's likey
9:29 how much is that going to cost me you
9:31 son of a
9:32 [Music]
9:37 [ __ ] it doesn't cost
9:40 anything
9:42 [Music]
9:45 wait in a
9:49 Westerly
9:51 [Music]
9:54 Direction car is my train
10:00 driving I've been
10:01 [Music]
10:03 wondering what it is I'm running from
10:07 [Music]
10:09 again Cody x i play a different video
10:12 every night I may or may not be awake
10:14 for it thank you that's that's getting
10:17 the that's getting the play count up I
10:19 think we're pretty close actually but
10:20 but thank you very much um I think what
10:24 I'm also going to do I I mean here this
10:27 is part of the [ __ ] up thing about Tik
10:29 Tok going away
10:30 so I paid some decent amount of money to
10:33 to
10:34 download all of my Tik Tok videos and
10:37 push them over to Google Drive so I'll
10:40 have them
10:40 all and
10:43 then I was gonna upload those to YouTube
10:45 shorts but I think YouTube shorts don't
10:47 they have like a three minute cap or Max
10:51 so I don't know so I guess I'll just do
10:53 vertical videos as normal YouTube videos
10:56 which I think is [ __ ] stupid but
10:58 whatever
11:00 love your hair tonight oh boy what have
11:01 I done I don't oh it looks all right
11:05 it's a little it's a little it's a
11:07 little oh you know what I did I I took a
11:09 little I did a little I did a little
11:11 eyebrow trim with the the eyebrows I I
11:15 was looking mighty you
11:18 know uh what's it called Nutty Professor
11:20 like so I did the eyebrows and I did a
11:23 little bit of triming on the sides there
11:24 I need a haircut so I just did a little
11:27 self self-own on the haircut Thank you
11:29 very much the other thing I started
11:31 today is uh I'm just getting too [ __ ]
11:36 [Music]
11:41 fat so I let my I started my it's the
11:45 holidays I'll just start eating more I
11:48 started that like the week before
11:51 Thanksgiving and then I have
11:53 continued that
11:55 gluttonous
11:57 eating agenda uh until this morning and
12:02 I was like [ __ ] it so uh so I'm starting
12:06 intermittent fasting today and I think
12:08 I'm just gonna have like a green drink
12:09 for lunch and then just dinner and I'm
12:12 just going to do that until this body
12:15 feels less
12:17 ridiculous I know from out there because
12:20 of all the Prosthetics and the makeup
12:22 that it looks like I'm [ __ ] jacked
12:25 but situation is if you saw me stand up
12:29 you'd go oh my you know it it would be
12:32 it could be it could be a shocking
12:34 moment Corey Sandler hey Kyle some of my
12:36 Tik Tock lives are seven
12:38 eight nine
12:40 hours oh you're insane at The Potter's
12:43 Wheel I know Corey what are we supposed
12:44 to [ __ ] do with all this content and
12:47 I mean I guess I mean one of the answers
12:49 is Let It Go Right sa La V we we got you
12:54 know there's part part of it's that
13:00 you know the thing that that kind of
13:02 baffles
13:05 me this this really does kind of baffle
13:09 me is that what is it that Tik Tok did
13:13 that is so different than reals and and
13:17 Instagram whatever the [ __ ] they do or
13:19 YouTube shorts and all those they're all
13:22 effectively the same thing
13:26 [Music]
13:30 I mean I guess it was the algorithm
13:32 right just showing people [ __ ] that they
13:33 might like instead of their social
13:36 [Music]
13:40 graph is Blandon a green drink yes I
13:43 will include it in
13:51 that it wasn't image based to start but
13:54 video based yeah but YouTube shorts was
13:56 that although I guess YouTube shorts
13:58 came later YouTube was kind of like you
14:01 know old old people old people video I
14:05 don't
14:07 know I don't
14:10 know and twitch C certainly has this for
14:14 gamers
14:16 [Music]
14:27 right so if you're wondering what we're
14:29 going to talk about
14:32 tonight
14:34 um we're going to talk about agents I
14:36 think I think I might um do a little
14:40 exploring with
14:44 [Music]
14:49 uh with Runway ml with character
14:53 Generation stuff um I've been working on
14:56 a project for work and and digging a bit
14:59 deep ER on Runway ML and act
15:02 one
15:04 um and man it it takes some work but
15:08 what you can do if you put in the work
15:10 is pretty
15:12 stunning so we can go look at
15:15 that
15:17 um Brandon would be excited to know
15:20 maybe we'll play with some jelly pod
15:23 which is a podcast maker that's kind of
15:26 like notebook LM but you get to choose
15:28 the the voices and then Vicky just sent
15:32 me another one that is similar to that
15:34 that's open
15:36 source um which looks
15:40 interesting and what would be cool is if
15:43 I showed the thing Vicky sent me and not
15:46 Brandon's Brandon can watch this later
15:49 and get bitter about it and I think
15:50 that's
15:51 [Laughter]
15:54 funny thank you Lord digital Gods thank
15:57 you for the croses
15:59 the open source one is
16:04 called um well it's on foul. a f.
16:11 a and then it's called play
16:16 AI so foul.
16:20 a slm
16:22 models here here's I'll do this copy
16:26 that
16:37 so if I put this here and do
16:41 that and do that there it is right there
16:45 on the screen on the big
16:48 screen all
16:50 right don't worry about Brandon he's a
16:53 big
16:54 boy Vicki's already throwing him under
16:57 the bus Brandon it's like your seat
16:59 isn't even it's still warm it's still
17:02 this is like a Shakespearean drama over
17:04 here all
17:08 right
17:13 um I W to I want to I made a uh I did a
17:17 Google deep research thing on AI agents
17:20 as well so over on the AI Salon I shared
17:25 that research document um which I think
17:28 it's pretty [ __ ] amazing Google deep
17:31 research if you have not played with it
17:33 or if you not have not experienced it
17:34 it's pretty [ __ ]
17:36 insane
17:37 [Music]
17:39 um
17:41 so so if you stick around I'll I'll give
17:44 you the locations of the location of
17:46 that thing hang on my glasses are like
17:49 champy took a dump on them what what is
17:52 it with
17:53 glasses like you don't touch them right
17:56 you they're just on your desk and you
17:59 put them them on your desk and they're
18:00 clean and then you pick them up and put
18:02 them on and it's like someone took a
18:04 dump on the lenses what when does that
18:07 happen are there like fingerprint
18:10 fairies that just run around
18:12 [Laughter]
18:17 like get this one really
18:20 deep what the
18:23 [ __ ] good Lord camcat kid all right what
18:26 can you say about the AI diffusion
18:30 AI diffusion rule from today I don't
18:33 know anything about that what did I miss
18:36 what did I miss hang on hold please AI
18:41 defusion
18:54 [Music]
18:57 rule this thing
19:00 Biden Administration proposes new rules
19:02 on exporting AI chips is that from today
19:09 no oh it
19:14 is
19:17 well the horse is already out of the
19:19 [ __ ]
19:21 Barn did you see there's a new model
19:23 today is it a coding model or over the
19:27 weekend it came out it was a coding
19:28 model model that they trained on
19:31 $450 of
19:33 compute that is like as good as 01
19:37 preview it's it's like benchmarking as
19:39 good as 01 preview trained on $450 of
19:45 compute the only thing that
19:49 that
19:52 okay provoking an industry push
19:55 back so I don't I don't know anything
19:57 about this but it's like it's it just
19:59 seems to me it's like well okay so
20:03 China's kicking our ass because our last
20:07 our last exporting rule where we we
20:10 hobbled the chips that went to China
20:12 forced them to innovate so they got
20:14 better at working with shittier
20:17 [Music]
20:19 chips like like at some point someone's
20:23 going to
20:25 recognize that we gota we got we've
20:30 oh [ __ ] like I don't even know where to
20:33 start
20:38 it's
20:41 um the fact that we
20:44 have zero Federal activity about AI
20:48 regulation they're pushing it all to the
20:50 States they're just ignoring it
20:51 federally and they're pushing it all to
20:54 the states and they're allowing the
20:55 states like Colorado that went first on
20:57 this to pass really shitty over Brad
21:00 rules that are regulating the tech and
21:02 not the use cases which every time you
21:04 do that it's a bad
21:06 situation and the states because they
21:12 are
21:14 like radically uneducated about what's
21:17 actually going on with AI they're just
21:19 passing laws so they can go we're out
21:21 there protecting you people you we it's
21:23 good this this is what the politicians
21:25 are like here ye here you leave fellow
21:28 citizens we understand that the AI
21:30 robots are going to kill you and
21:32 therefore and China's evil and except
21:34 for when they're not and but we're going
21:37 to never give them the good stuff un but
21:40 on a state-by-state level we will
21:42 prevent you from using AI so that it
21:45 protects your interest unless of course
21:47 you're an entrepreneur and then good
21:49 luck with that and by the way how do you
21:52 determine whether a piece of AI
21:55 technology is being used in this state
21:56 or that state well you as the
21:58 entrepreneur can figure that out you're
22:00 [Laughter]
22:05 [Music]
22:07 clever this is our [ __ ] legislation
22:10 we got the most profoundly powerful
22:12 technology in the history of humanity
22:14 and we're like ah will figure it out
22:16 [Music]
22:27 sometime good Lord
22:30 Lord good Lord people all right hang
22:35 on he seems he seems a little aggressive
22:39 tonight I don't understand what's going
22:40 on but he certainly does not seem
22:43 stable
22:49 uh Senator Ed Markley this is from Jim
22:53 Walters various posts a few hours ago
22:55 Senator Ed Mary filed a bill to extend
22:58 the deadline Tik Tok for another 270
23:01 days so you're saying we got a
23:06 chance I believe in the finger
23:09 fingerprint Sprites very nice glass is
23:13 one of life's great Mysteries summarize
23:15 it in chat gbt oh that's a great idea
23:17 all right let's go do that we'll do copy
23:19 here hang on we'll go share this with
23:22 all of the world to see we'll put the
23:24 black bar back up because I'm a
23:26 professional uh and then we'll go over
23:29 uh wait
23:30 um dang it there we go black bar back
23:35 over here yeah that's
23:40 working 01 and o02 reasoning models are
23:43 amazing except O2 is now called 03 but
23:46 yes they retired the chat Bots I made
23:48 several months ago yeah I know it's
23:50 pretty amazing uh and people that are
23:52 are uh playing with um 03 or 01 Pro are
23:58 saying that it's quite a amazing that's
24:00 the 200 Buck a month thing uh let's see
24:04 summarize this
24:09 article thinking I'm unable to access oh
24:13 links externally that's
24:16 because uh let's see try now
24:21 [Music]
24:31 [Music]
24:35 let's see what it says
24:37 here the Biden Administration has
24:40 introduced new regulations to restrict
24:42 the export of advanced artificial
24:43 intelligence chips aiming to prevent
24:46 adversarial Nations China from acquiring
24:49 Cutting Edge us technology
24:51 [Laughter]
25:01 have you noticed that every Chinese
25:03 model comes out about three or four
25:06 weeks after the US model comes out and
25:08 it's the exact same [ __ ] thing you
25:11 don't think they have access to our
25:15 technology these I you know the more I
25:18 learn about politics the more I realize
25:20 oh it's just all [ __ ]
25:23 PR um these measures Target high
25:25 performance semiconductors okay
25:27 Department of commerce's Bureau of
25:29 industry and statistics bis will oversee
25:31 these controls fine industry leaders
25:33 have expressed concerns that these
25:34 restrictions could hinder Innovation and
25:37 competitiveness they argue that limiting
25:39 access to International markets May
25:41 reduce Revenue that's it fascinating so
25:44 they're saying don't take re don't take
25:45 Revenue away from Nvidia or if we ever
25:47 get our [ __ ]
25:50 together affecting research developments
25:52 and Investments well I understand the
25:54 reason to try to protect the US
25:56 interests like we've got a a [ __ ] ton of
25:58 elite and we've got a lead in um in
26:03 energy generation as well I
26:05 think the administration contends that
26:08 these measures are necessary to prevent
26:10 the misuse of AI technology and Military
26:12 applications and human rights
26:14 abuses by controlling the distribution
26:16 of advanced AI chips it how about how
26:19 about passing some [ __ ] reasonable AI
26:21 legislation at the federal level here
26:24 [ __ ]
26:26 idiots and I mean [ __ ] idiots in the
26:29 most positive possible
26:31 [Music]
26:45 light is this a place I can rest my PO
26:50 [Music]
26:53 head together my thoughts in sweet
26:56 silence
27:02 there this place where the feelings on
27:09 De from an over exposure to
27:13 violence where's champy champy already
27:15 sang champ sleeping he did his he did
27:18 his Union minimum and he has checked out
27:20 of the
27:21 building oh
27:25 [Music]
27:30 um all right so here's what I'd like to
27:33 do um what are we going to do so here
27:38 okay hang on just just bear with me bear
27:41 with me self- production here
27:45 self-producing Clueless Okay if you've
27:49 not gone to the AI Salon if you've not
27:52 joined the AI Salon you should do that
27:55 go to the
27:56 salon. scroll down until you see button
27:58 that says join our community click that
28:01 button it doesn't cost you anything we
28:04 did over the holidays you may have heard
28:05 of this thing called AI FES of us AI for
28:08 the rest of us we did that we partnered
28:10 up with Anne Murphy and the she leads AI
28:14 Community we put on
28:17 this two-day workshop Aon had 2300
28:21 people show up over two days on a
28:23 holiday
28:24 weekend um that's that's AIS Salon um
28:30 [Music]
28:34 and we just did some we just spent the
28:38 weekend Vicky Baptist and Leah and I and
28:45 Eric Sanner showed up I'm pretty sure
28:48 hang on a
28:50 sec we're gonna go here we're g to go
28:55 here there we go all right
28:58 so here's the AI salon now over here on
29:01 the left hand
29:03 side
29:05 um there's a couple of things so there
29:07 there's a welcome area so when you first
29:09 get to the AI Salon you land on the
29:11 welcome there's a welcome video uh from
29:14 from Leah faston and I we co-founded the
29:17 salon and then we talk about the journey
29:19 to AI Readiness which is which looks
29:21 like play first mindfully create
29:23 generously
29:25 lead which I love that little this was
29:28 sort of born of how this community acts
29:31 and and what we
29:34 do and then there's a a little
29:36 seven-step onboarding process you you
29:39 sort of log in learn about us introduce
29:43 yourself um check out our
29:47 values curiosity playfulness empathy
29:50 bravery generosity what we're all about
29:53 you can watch past meetings if you want
29:55 to see we've been doing this since
29:57 December 7th 2022 so the AI Salon was
30:00 founded the week after chat GPT came out
30:04 um you can RSVP for upcoming events you
30:06 can go into the water cooler and if you
30:08 go into the water cooler one of the
30:10 areas you can go is this thing called
30:13 discussion threads and if you click on
30:16 that the very first one is that research
30:19 document that I created using uh Gemini
30:23 deep research so this is a list of AI
30:26 agent platforms from no code to low code
30:30 to go build your own [ __ ] and with with
30:34 it's pretty it's a pretty cool little
30:35 research thing that that Google deep
30:37 research
30:38 did um and then there's all sorts of
30:41 other stuff and there's there's a news
30:42 area there's a showand tell area where
30:44 you can show off your work it's called
30:45 showof and toolbox where you can show
30:48 off uh images you've made or gpts that
30:52 you've made whatever it might be that's
30:55 really nice from CJ very cool
31:01 so you can go share off show off your
31:04 work and then within each of these
31:06 sections is a tool list of all the
31:09 different tools that are worth checking
31:13 out and then we used to have what we
31:16 called guilds but here's what we did
31:18 some of the guild the guilds there were
31:19 too many of them they were confusing it
31:22 we weren't sure where to put them and
31:24 what we realized was that what the
31:26 guilds were affectingly acting as has is
31:28 like clubs or or sub communities or hubs
31:31 within within the space so so there's
31:34 now just sort of this showoff area where
31:36 you show off your work and then there's
31:38 clubs and hubs so there's AI Festivus
31:40 was you know that that event we did so
31:43 there's a club for that there's a club
31:45 for the Irregulars which is this group
31:47 these clubs are ordered by level of
31:50 activity so for those that run clubs if
31:53 you get more people to use your Club
31:55 you'll go up higher in the list which is
31:56 kind of fun um and it just makes the the
31:58 space a lot more simple but there's so
32:00 much going on here it's so valuable so
32:02 go check all this out all right and then
32:06 in the water cooler under discussion
32:09 threads that's where
32:12 this this Google doc
32:14 is navigating the AI agent landscape
32:18 categories of AI agent platforms no code
32:21 low code developer Focus
32:24 platforms then it describes them's
32:27 benefits links to the site tutorials for
32:31 that particular thing
32:33 videos
32:37 insights now are these the right ones I
32:40 don't know I just had Google deep
32:42 research go do all this but it it went
32:44 out and it searched like 250 websites
32:48 and then made this
32:50 report so there you have it it's Google
32:53 deep research is pretty [ __ ]
32:55 cool but anyway so that's all in there
32:57 so if you want to go start learning
32:59 about um autonomous agent platforms I I
33:02 think now's a good time to do that um
33:05 okay so I want to talk about why I
33:08 believe that is the case let me go find
33:10 something here let me let me while while
33:14 I'm searching let me just do
33:17 this um any questions any thoughts
33:20 thoughts questions
33:25 comments love Serena's graphic I do too
33:34 One-Stop shop for AI thanks for this
33:37 you're
33:37 welcome you're welcome cuno yeah this
33:40 was one where um so so agents have
33:44 become the talk
33:48 dour
33:52 and I
33:59 if we think the government is
34:05 ill-prepared I don't listen I don't know
34:07 why I'm laughing because it's [ __ ]
34:10 tragic like like you know there there
34:13 should be a law that lawmakers can't be
34:15 this [ __ ] clueless for something this
34:17 important but here we are um if we think
34:22 that
34:23 legislators and people are ill prepared
34:26 for the current state of AI I I have a
34:29 feeling 2025 is going to be a uh a
34:34 Fantastical popcorn eating movie
34:37 watching uh
34:40 Extravaganza cuno oh you had your
34:42 Minnesota meeting today I that's right
34:44 um who was there it was it was Lori Ryan
34:48 Chris vuno and someone else who was the
34:50 third person that was there
34:53 Chris uh uh uh uh uh uh uh uh uh
35:02 [Music]
35:04 oh kakei that's right that's right
35:06 that's right khaki and khaki so I talked
35:09 to the people that organized that event
35:13 and they said it was a hit were amazing
35:16 um they said there's actually a follow
35:17 on opportunity so great so so about
35:23 three weeks ago there was an AI
35:27 education event for Oregon legislators
35:31 that Anne Murphy went and represented
35:33 the AI salon and today there was one in
35:36 Minnesota and what's what's amazing
35:38 about these events is
35:41 um the thing about the AI
35:44 Salon is not only is it really nice
35:47 people and really generous people and
35:49 really smart people but it's people that
35:53 are are kind of
35:55 impressive and and I I you know I I
36:00 sometimes forget that like like like
36:03 we're just kind of hanging out talking
36:05 about Ai and we're asking good questions
36:07 and things like that but we're doing it
36:09 on a daily basis and and if you take
36:13 people who are really established
36:17 professionals and over the course of two
36:19 years they just sort of immerse
36:21 themselves in a topic and they take
36:24 their professionalism and they sort of
36:26 Route it through that topic you know
36:28 we're some of the most educated
36:30 impressive badass people around this
36:33 idea of generative AI not not AI
36:35 traditional AI so much but more like now
36:37 that chat GPT is here and now that real
36:40 people and real small
36:42 businesses are confronting what these
36:44 tools
36:46 are and and we're learning what's
36:49 possible we're like you legislators need
36:51 to [ __ ] deal with that and what's
36:54 been the result of of two out of two of
36:56 these meetings this year is that AI
36:58 Salon people have showed up educated the
37:01 legislators and the legislators have
37:02 left those meetings going we got to get
37:05 our [ __ ] together so so thank you to to
37:09 all three of you today for doing that
37:11 the the this the spirit of the salon is
37:13 also I reached out to these guys because
37:15 these things come up pretty quickly like
37:18 they're sort of happening in real time I
37:19 reached out to these guys seven days ago
37:23 and and all three of them said yeah I'm
37:25 in let you know let me go do it so
37:27 awesome I'm so yeah I heard it went
37:29 great and I just I didn't get a chance
37:30 to get a note out so so thank you for
37:33 for representing the salon that's that's
37:35 so awesome so
37:37 awesome karuno we're decent at how to
37:40 operationalize AI in various sectors and
37:42 you know Chris it's not just how to
37:44 operationalize it it's not just that
37:48 it's you can sit across a table and when
37:52 someone is
37:53 uneducated about what it even
37:55 is you can hear them not get
37:59 defensive educate the
38:03 room Elevate the conversation above fear
38:08 into
38:10 possibility and then you can talk about
38:12 and here's how you operationalize it
38:13 right it's like you can do that too but
38:16 it's like there's there's a there's a
38:17 much more fundamental skill right now
38:20 which is there's a lot of Fear and
38:22 Loathing around
38:24 Ai and being in the presence of people
38:28 people that are like no no no you don't
38:29 get it yeah yeah there might be risks
38:31 but there's there's all this other stuff
38:33 no one's talking about it's opportunity
38:35 it's optimism it's people Reinventing
38:38 themselves it's people coming out of
38:41 retirement right that was the the other
38:44 thing that the the the folks that
38:46 organized it said today which I thought
38:48 was I thought was really cool is they
38:50 were like you know like like all three
38:53 of them were like basically what they
38:55 said was like gen xers like there were
38:57 were no young people there I'm like I
38:59 I'm like I'm telling you Gen X is
39:02 crushing it with AI like there's
39:04 something about our you know ability to
39:08 be a pain in the ass our ability to to
39:11 have have recognized that we saw this
39:14 with the worldwide web and now we're
39:15 getting to see it again with with AI
39:17 we're like oh this one's
39:19 bigger and I think there's there's we
39:22 we've just got a spirit of let's screw
39:23 some [ __ ] up here I'm really excited
39:26 about that so anyway that's awesome all
39:30 right Silver Fox I'm so part so excited
39:33 to be a part of the AI salon and learned
39:35 so much it's a great group thank you so
39:37 much that's awesome Jeff Flanigan
39:40 digging into future pedia today lots to
39:42 unpack there too much to unpack there so
39:44 much to unpack
39:46 there khaki and how do we leave a seat
39:49 at the
39:50 table and how do we leave a seat at the
39:53 table for oppressed communities we got
39:54 to get them we got to get them in the
39:56 conversation khaki I mean that's the
39:58 that's the that's the deal it's one of
40:00 the things I'm most excited about with
40:03 AI is that it gives everyone a chance to
40:07 dramatically level up their skills no no
40:10 matter what their access in including
40:13 you know a lot of accessibility stuff is
40:15 is going to be bridged um so I think the
40:18 tools are there but it's you know how we
40:20 get people at the table is we get people
40:22 at the table um you know we we've got CJ
40:25 Fletcher within the salon has got luck
40:28 L
40:30 Cultura um where he's he's uh he's got a
40:32 group within the AI Salon that are
40:35 exploring um you know people of color
40:39 um and how they can how they can use
40:41 these tools how they can express
40:43 themselves how they can drive things
40:44 like that so I think it's like I think
40:47 that is an active conversation how do
40:50 you get people in the conversation you
40:51 get them in the [ __ ] conversation I
40:53 don't think it's any different than you
40:55 know what we do here I don't
40:56 specifically Target oppressed
40:57 communities here um but I think it's I
41:01 think it's what this activity is just
41:04 aimed at different audiences right talk
41:07 about it demystify it figure it out have
41:10 fun with it play explore experiment all
41:12 that sort of
41:13 stuff cuno the cool guy from the
41:16 governor's office was very engaged with
41:18 cool use cases use cases are the bridge
41:20 to get people to understand what's
41:22 possible getting involved right now is
41:24 how to get at get to the table I mean
41:27 that so so r rispy Wait ripsy ripsy
41:33 you've got it where where it's like we
41:36 are so early right now it may not feel
41:39 like we're early but trust me we having
41:43 been through one of these before in the
41:45 90s um we're early there are so many
41:49 people that don't even know what's
41:51 happening like don't even have a
41:54 clue what is possible today they don't
41:57 have a clue what was possible two years
41:59 ago they have less of a clue what's
42:01 possible today they have no idea what's
42:03 about to hit them with with autonomous
42:05 agents
42:09 um anything we can do to get anyone into
42:13 this conversation right now will give
42:15 them disproportionate opportunity and
42:19 have a disproportionately positive
42:21 impact on whatever this industry
42:24 becomes Rachel thank you Rachel
42:30 wait
42:31 ripy same experience and I completely
42:34 agree um what is possible today is so
42:37 impactful and it's right here yeah it is
42:40 it is here today like
42:42 the it's I every time I do something now
42:47 like that that research thing that I
42:49 just shared oh the thing that's on
42:50 screen right now hang on let me pop that
42:53 down and go back to
42:55 this this research
42:58 thing it took it took Google deep
43:01 research about I don't know 10 minutes
43:04 to process this but you know here's a
43:08 16-page research document on autonomous
43:12 AI platforms you know with
43:15 citations like like you can just do that
43:19 now well in this case you have to you
43:21 have to have a subscription to it and
43:22 you have to use an individual Google
43:25 account because if you've got a Google
43:27 work group account you're considered a
43:30 pariah if you pay Google they hate
43:34 you but like like this thing's amazing
43:37 here let me show you
43:40 um where can I show you something here
43:43 if I go to
43:47 um
43:55 Runway we're working on a project for
44:02 client generative
44:07 session is it this
44:18 one let me change my sharing so you can
44:21 hear this
44:42 [Music]
44:49 um so this is like a
44:53 virtual sales rep named
44:56 Michelle three 3 2
44:59 1 hello my name is Michelle your virtual
45:02 sales rip it's great to connect with you
45:05 our digital platform allows you to get
45:07 the latest updates on our products and
45:09 services order samples and patient
45:12 savings cards and access educational
45:15 resources to keep you informed and
45:17 prepared for evolving Healthcare
45:20 needs three two I mean like she doesn't
45:26 exist the voice doesn't exist that like
45:29 I actually did the voice acting for this
45:31 and the face acting for this and then
45:33 swapped in her
45:36 voice you know we can do this today like
45:39 this we can do today no special effects
45:42 you know me me sitting at a computer
45:44 like hey welcome I'm your virtual sales
45:47 rep it's just amazing so um all right I
45:53 want to go find something so we can talk
45:55 I want to
45:56 talk I want to to look love Runway ml to
45:59 create 30 second promo videos for my
46:01 clients that's Jim Ross from Self
46:03 Storage from Three Mile
46:06 Storage
46:09 uh Jeff Flanigan I'll be introducing AI
46:12 to a h
46:14 School nearby in a few days parent
46:17 teachers fully on board that's awesome
46:19 that's
46:20 fantastic
46:22 um all right can you explain the
46:25 $450 thing mentioned earlier
46:29 oh yeah uh hang
46:33 on
46:36 um
46:37 search
46:40 01 $450 should do
46:46 it yeah
46:53 [Music]
47:04 sky- T1 open L large language model
47:08 beats 01
47:09 preview Sky T1 an open reasoning large
47:13 language model trained with just
47:16 $450 of large language model I mean I
47:19 assume that means trains with $450 of
47:23 compute um to pate's point it is highly
47:28 likely
47:30 that this is a fine-tuned model based on
47:33 something else that costs significantly
47:35 more than
47:37 $450 right so this isn't just but um
47:42 whatever they did here they've got a 32
47:44 billion parameter model which in in
47:47 large language model SI oh wait I'm not
47:49 sharing this hang on damn
47:52 it maybe share your screen Vicki needs a
47:55 shot collar I tried jelly pod this
47:57 weekend Brandon would be proud very nice
48:00 um yeah so here's the sky T1 beats 01
48:07 preview in in some
48:12 benchmarks
48:14 right cost them 450 bucks to fine-tune
48:17 this thing again I don't know I don't
48:20 know a ton about this but a 32 billion
48:23 parameter model is small right big
48:27 language models are 405 billion
48:29 parameters 32 billion is is the size you
48:33 can you can start running these things
48:35 locally on your hard drive right so
48:37 imagine having something as good as 01
48:40 preview which 01 preview had a an IQ of
48:44 about 120 I think 03 has an IQ of about
48:48 157 um so imagine having a you know 120
48:52 IQ problem solving super knowledgeable
48:57 entity running off your local hard drive
48:59 you don't have to go out on the internet
49:01 anywhere so that's that all
49:05 right okay I guess that shot collar
49:08 didn't arrive in time it
49:12 didn't yeah that's impressive isn't it
49:14 trench I know it's it's Bonkers okay let
49:16 me go find something else here I want to
49:18 go
49:19 find I want I'm trying to think which
49:21 one I want to read first okay I know
49:24 which one I want to read
49:25 first there were two two posts I read
49:28 today backto
49:32 back
49:34 that I've kind of known what's coming
49:37 with autonomous agents but they they
49:40 really solidified for
49:43 me that something different is we're
49:47 we're about to embark on something quite
49:48 different
49:51 so let me go find
49:57 the
50:00 one so I'm on David Shapiro's at Dave
50:05 shap sh a pii Shapiro just Dave
50:10 shappy there's Ethan mullik there's that
50:13 one is this no that's not it it's this
50:15 one is this
50:19 it
50:21 yeah all right so I'm going to push this
50:24 in close so you guys can read
50:36 okay
50:38 so
50:42 um Kim anismus chubby who's a who's a AI
50:49 commentator and she said you know the
50:51 news that meta wants to replace
50:53 mid-level Engineers is spreading we're
50:55 taking six we're talking six figure
50:57 salaries here people and meta wants to
51:00 replace them completely and this was
51:02 from
51:03 Zuckerberg probably in 2025 we at meta
51:07 as well as other companies that are
51:08 basically working this are going to have
51:11 an AI that can effectively be a sort of
51:14 midlevel
51:16 engineer that you can have at your
51:18 company that can write code so that's
51:20 Zuckerberg just talking about you know
51:22 what they're doing at
51:24 meta all right can everyone see do I
51:26 have everything set up correctly I think
51:27 I do I'm on a
51:30 sec and then David Shapiro responded to
51:34 this and I want to read through his post
51:37 because
51:38 it there's a level of clarity in it
51:43 that I really appreciate um but it's but
51:47 it's it's a big deal it's a big deal if
51:50 you don't know David Shapiro he has been
51:52 an AI researcher for years he he has
51:55 been you know ay that does you know
51:58 highlevel systems troubleshooting for
52:00 his day job and wrote five books on AI
52:04 he's done a number of uh cognitive
52:07 architectures open- Source projects to
52:09 do AI cognitive
52:11 architectures and about three or four
52:15 months ago he quit
52:18 Ai and he didn't quit AI like um I'm
52:23 just done with it I'm out of here screw
52:25 you people he quit I like they've
52:29 figured it
52:30 out
52:33 that based on his
52:35 knowledge and knowing what's
52:39 coming that he realized that him doing
52:43 anymore research on this was pointless
52:45 because they basically figured it out
52:47 and they've gotten it right and if if
52:49 you listen to the engineers coming out
52:51 of open AI right now they're no longer
52:53 talking about AGI they're not they're
52:56 now talking about ASI right Sam Alman
52:59 basically said we we now
53:02 know we're not speculating we know how
53:05 to get to
53:06 AGI and what we're up to now is getting
53:08 to ASI and if you don't know those two
53:10 acronyms AGI artificial general
53:12 intelligence where the AI is as good or
53:15 better than most people at most
53:17 economically viable jobs that's
53:20 AGI ASI is super intelligence where
53:23 these things become smarter than all the
53:25 humans you you know the best possible
53:27 human in every possible
53:31 category super intelligence right so so
53:34 we are on a path right now for that so
53:38 here's what here's what Shapiro wrote in
53:41 response to meta is going to
53:44 replace midlevel and Below Engineers
53:48 with autonomous agents so David Shapiro
53:51 let me lay this out fresh because after
53:53 years of building cognitive
53:55 architectures
53:57 I've learned something fascinating about
53:58 intelligence versus agency everyone's
54:01 hyping up AI agents like they're the
54:03 next breakthrough but they're missing a
54:05 crucial piece of the puzzle that took me
54:07 way too long to figure out here's what
54:10 threw me for a loop raw intelligence
54:13 isn't the same as knowing what to do
54:14 about it think about it chat Bots were
54:17 relatively straightforward because they
54:19 were just waiting for humans to start
54:21 the conversation give them a prompt
54:23 they'll Dazzle you with their smarts I
54:25 call this the genius in the jar phase
54:29 but true agency that's where things get
54:32 spicy okay so we're in the genius in the
54:35 jar phase with current llms right
54:37 they're super smart but they require us
54:39 to deal with them and I and one of the
54:42 things that I hear unable to verify hang
54:45 on sorry I'm having H I'm having uh Tik
54:49 Tok is making me do physical dexterity
54:51 Mind Games okay I did it great
54:57 he quit because he's suffering from from
54:59 an auto autoim immune condition as well
55:02 but AI helped fix it yeah exactly
55:04 exactly well it's funny Danielle he quit
55:07 AI but like all he talks about now is AI
55:10 so he's moved from researcher to
55:13 commentator uh which I think is that
55:15 makes sense to me
55:17 okay but true agency that's where things
55:20 get spicy the real Cha challenge hit hit
55:22 me one day on building cognitive
55:24 architectures you're sitting there with
55:26 this this incredibly powerful system
55:28 that can think about literally anything
55:30 and you suddenly realize wait how does
55:33 it how does it decide what to think
55:35 about it's like having having infinite
55:38 processing power but no executive
55:40 function as a card carrying ADHD member
55:44 with no executive function I I resonate
55:47 with this so current ai ai is like add
55:50 right all this capability no way to
55:53 focus it
55:58 that's hilarious
56:00 um uh okay the human brain handles this
56:03 naturally but programming it that's a
56:05 whole different ball game you'd think
56:07 that solving the what Tod do problem
56:09 would be straightforward with enough
56:11 computing power right wrong I've had to
56:15 dive deep into ethics and philosophy
56:17 just to create a basic decision
56:19 framework utilitarianism uh theology
56:24 deontology um thrown
56:26 uh throw all that in and you've barely
56:29 scratched the surface because then you
56:31 need to build out the entire model of
56:33 what an agent can do what it can't do
56:36 what it should and shouldn't do and most
56:38 importantly why that's the agent model
56:42 layer okay that's important for the next
56:45 thing I'm about to show
56:47 you it's got to know what I am and how
56:51 do I work before it can start even using
56:54 its digital hands I call this the
56:56 constraints capabilities in context or 3
56:59 C's of the agent model um and just when
57:02 you think you're getting somewhere you
57:03 hit the broader context problem your
57:05 agent needs to understand its
57:07 environment available tools specific
57:09 tasks access levels and we haven't even
57:13 touched on goal tracking and T task
57:15 prioritization yet so again this is a
57:18 guy that's built these systems right
57:20 it's like trying to build Jarvis from
57:22 scratch and realizing why Turing never
57:26 cck crack this part of the puzzle Allan
57:28 Turing the Turing test you know I I'll
57:31 know uh computers have reached
57:33 intelligence when I can have a
57:35 conversation with one and not be able to
57:37 tell the difference if it's a human or
57:39 if it's a machine right that Turing um
57:42 this is why Turing never cracked this
57:44 part of the puzzle which is the how do
57:47 you use you know this intelligence
57:50 resource utilization keeping track of
57:52 what you need and have and how close you
57:55 are to solving the actual problem llms
57:58 can do each of these steps no problem
58:00 but coordinating it all a bit
58:04 harder that's what makes today's llm so
58:07 interesting let let me see see there's
58:09 comments here Shapiro pioneered the
58:11 hierarchical framework well for me in
58:13 which the agents have an upstairs office
58:16 that deals with the human in the loop
58:18 yeah exactly yes AI can write code but
58:21 that AI that writes code cannot do my
58:24 job it doesn't have the ability to
58:26 insert itself agentically into what
58:28 already exists exactly okay um what else
58:32 do we have here any other
58:36 comments
58:38 okay that's what makes today's llm so
58:41 interesting there's still narrow AI in
58:44 their operation but they're packing some
58:47 serious general intelligence so when he
58:49 talks about there's still narrow AI what
58:52 he's talking about is classical AI you
58:55 would take a narrow task like beating a
58:59 chest Grand Master or diagnosing one
59:02 specific kind of cancer and you would
59:04 train an AI on that it would just have
59:06 this narrow expertise what David's
59:08 saying here is even though today's large
59:12 language models have this general
59:14 knowledge capability they're narrow AI
59:17 because they still require us to figure
59:19 out what to do with them okay um there's
59:24 still Nar narrow AI in their operation
59:27 but they're packing some serious general
59:28 intelligence features under the hood
59:31 World models reasoning planning problem
59:33 solving it's all baked in but true
59:36 agency that's still the Holy Grail we're
59:38 chasing true agency these things being
59:41 able to do it on their own that's the
59:43 Holy Grail we're chasing and it's not
59:45 necessarily just the model it's the rest
59:47 of the body so to speak okay think about
59:51 it if your brain was just floating in a
59:52 jar somewhere no eyes ears or hands you
59:56 couldn't really do anything either to
59:58 make matters a bit more confusing AI
1:00:00 agents exist in cyberspace where what
1:00:04 they see and hear is just too different
1:00:07 from our meat space context right and
1:00:09 their hands are API calls so he's
1:00:11 basically saying like we don't know how
1:00:13 to use we we don't know how to give
1:00:16 these brains in a jar bodies right
1:00:20 okay and then here's how he closes it up
1:00:22 look here's the punchline that most
1:00:25 people haven't wrapped their heads
1:00:26 around
1:00:27 yet we cracked this
1:00:37 problem this is one of the reasons he
1:00:39 quit
1:00:41 AI here's the punchline most people
1:00:43 haven't wrapped their heads around yet
1:00:45 we cracked this problem those years of
1:00:48 research into cognitive architectures
1:00:50 and agent Frameworks weren't just
1:00:52 academic exercises we were building the
1:00:55 foundation for some
1:00:57 exponentially bigger than anything we've
1:00:59 seen before the real kicker once you've
1:01:02 got one working agent system you don't
1:01:06 just have one you've got the potential
1:01:08 for
1:01:13 billions so take take it in so what he's
1:01:16 saying is we've got this brain in a jar
1:01:18 what we haven't figured out well the
1:01:21 hardest problem to figure out is how do
1:01:24 you give that brain in the jar the body
1:01:27 the ability to go do [ __ ] that's the
1:01:29 part he saying we've cracked and he says
1:01:32 if you crack it for one agent you've
1:01:34 effectively cracked it for all of them
1:01:37 okay um the same agent perfectly
1:01:41 replicated operating at scale across
1:01:44 every digital system on Earth copy paste
1:01:47 repeat we're not just talking about
1:01:50 replacing a few mid-level Engineers here
1:01:52 and there we're talking about a tsunami
1:01:55 of artificial EX executive
1:01:57 function that's about to reshape every
1:02:00 industry every workflow every digital
1:02:04 interaction um the people tweeting about
1:02:08 2025 aren't being optimistic if anything
1:02:11 they might be understating it just how
1:02:14 fast this is going to move once it
1:02:16 starts the dominoes are already
1:02:18 following and trust me you're going to
1:02:21 want to be ready when they
1:02:24 hit every we should all print out this
1:02:27 paragraph and like hang it by our
1:02:32 desk they've already cracked the
1:02:37 problem and once they release it and if
1:02:41 they do it for one agent they've
1:02:42 effectively done it for for all the
1:02:44 agents okay questions
1:02:50 thoughts it feels like the universe just
1:02:53 got a lot bigger yeah I we'll wait till
1:02:55 you see what I show you next
1:02:57 next I have convos with GPT all the time
1:03:01 it knows a lot but it doesn't provide
1:03:03 insights that we can well don't so
1:03:06 listen sacred is
1:03:10 profane don't assume that that's going
1:03:12 to be the case for much
1:03:14 longer like that's exactly what he's
1:03:17 talking about
1:03:18 here his whole point is that these
1:03:21 things are smart but they're limited
1:03:23 because they require us to make that
1:03:26 kind of decision making and insight and
1:03:29 understanding that's about to change and
1:03:32 and it's about to change kind of
1:03:34 instantly and
1:03:36 overwhelmingly it's not like you're
1:03:37 going to get one agent and then three
1:03:39 months later you'll have 10 agents and
1:03:41 then no this is going to be like once it
1:03:43 happens it's just they're going to start
1:03:45 to proliferate and then as those agents
1:03:48 start figuring out the other kinds of
1:03:50 agents that need to be built and the
1:03:51 other kind of Frameworks that need to be
1:03:52 built they're going to prol proliferate
1:03:54 even faster
1:03:56 okay got it cool right it's a big deal
1:04:00 it is a big deal Jeff
1:04:02 Flanigan it's a big
1:04:05 deal
1:04:09 okay all right let me go find this other
1:04:11 thing I want to
1:04:17 find this is Robert scobble
1:04:34 he forwards a lot of [ __ ] it might take
1:04:35 me a second to find
1:04:53 this is it did he do it as a
1:04:59 [Music]
1:05:10 okay this
1:05:11 one
1:05:13 okay if you don't know who Robert scoble
1:05:16 is he he's been an apple Fanboy for
1:05:20 years he's a San Francisco based you
1:05:22 know uh technology reporter and and and
1:05:26 you know he's he is deeply entrenched in
1:05:29 the scene in Silicon Valley he was there
1:05:34 for the first I think he said it Siri
1:05:36 was launched in my home um I live
1:05:39 streamed the first ride in the first
1:05:41 Tesla ever right like this is a guy
1:05:44 that's connected so what he
1:05:48 writes this will be the most important
1:05:51 video I've ever
1:05:53 done this is a guy that's been making
1:05:56 videos and making posts and and and
1:05:59 being on the Forefront of what's coming
1:06:01 from a technological standpoint for like
1:06:04 35
1:06:05 years that first sentence that doesn't
1:06:08 land lightly on me because he's not a
1:06:11 bullshitty kind of guy like if he thinks
1:06:13 something sucks he thinks it sucks now
1:06:15 when he goes all in on something like
1:06:16 he's all in on on uh on the holck and
1:06:20 you know and Google or Apple Vision Pro
1:06:23 and all that sort of stuff so there's
1:06:24 things he goes all in on that that you
1:06:27 know you may or may not agreee are going
1:06:28 to be a big thing but he's not a g guy
1:06:31 that says stuff like this too often so
1:06:33 this will be the most important video
1:06:35 I've ever done and again in the context
1:06:37 of I had just read the David Shapiro
1:06:40 post and then I read that sentence and
1:06:42 I'm like what the [ __ ] is going on
1:06:45 History will be the judge but I'm pretty
1:06:47 good at figuring out the future each of
1:06:49 my seven books properly predicted decade
1:06:52 Long's Trends my first on social media
1:06:55 was written before Twitter X even
1:06:57 started keep in mind Siri was launched
1:06:59 in my home I have streamed the first
1:07:01 ride ever of of the Tesla of the first
1:07:04 Tesla um I've had many other firsts but
1:07:07 this is the beginning of a new operating
1:07:10 system the guy that created Amazon's
1:07:13 Marketplace which is the main reason
1:07:15 Amazon is profitable company today this
1:07:18 will be a hugely important technology
1:07:21 and you could only learn about it here
1:07:23 the Accenture booth at CES was not open
1:07:26 to the public you had to be invited in I
1:07:29 am the first to take you behind the
1:07:30 scenes and Jensen announced Nvidia was
1:07:33 partnering with Accenture on robot
1:07:36 orchestration this is it so CES report
1:07:40 video 28 so what I'm going to do here is
1:07:43 I'm gonna change my sharing so that you
1:07:45 can hear this and we're gonna we're
1:07:46 going to listen to this video and if
1:07:48 anyone's still here you don't have to
1:07:50 still be here if if you're not here I
1:07:52 understand
1:07:58 um I think this shit's really
1:08:01 important you know to the to the David
1:08:03 Shapiro thing be prepared like how do
1:08:05 you be prepared I don't know you watch
1:08:08 [ __ ] like this and you try to get your
1:08:10 head around it like I can't quite get my
1:08:12 head around this the guy that we're
1:08:14 about to meet not this guy but the
1:08:16 skinnier guy he talks to in a
1:08:18 second is the guy that wrote dos like
1:08:22 the dis operating system that Microsoft
1:08:24 bought and turned into Microsoft that's
1:08:27 the guy who started this the company
1:08:29 that they're talking about right now
1:08:30 that guy right there that's him all
1:08:33 right so here we
1:08:39 go we're backstage at the Centra this is
1:08:43 the Cs that you never usually see how
1:08:47 the Nerds are running the
1:08:49 world and uhler I'm uh one of the lead
1:08:52 nerds x Microsoft Amazon and I'm the COO
1:08:55 of
1:08:56 uh autona and we do the paranet which
1:08:59 does all the orchestration for the demo
1:09:01 that you're going to see a little bit
1:09:02 later so so hold on a second because
1:09:05 perinet I I got a little I drove to
1:09:07 Seattle to see pinet and understand it
1:09:10 this is a new network protocol that
1:09:12 arranges or organizes orchestrates
1:09:15 drones robots autonomous vehicles and
1:09:18 human beings and human beings we pull
1:09:20 human beings and you're going to see
1:09:21 that in the demo later where Allen will
1:09:23 actually fault in for a robot the robot
1:09:26 will fault in for Allan seamlessly so
1:09:29 this is a distributed asynchronous
1:09:31 operating system Jim Hardy our founder
1:09:34 over here who you seen on a present
1:09:36 video I met with him at his house Philip
1:09:38 over here is the brains behind the
1:09:40 operation yeah Brian's running all the
1:09:42 robotics today on our side of the house
1:09:44 yeah we've got our great Partners
1:09:46 agility robotics over here yeah the
1:09:49 coolest robots you're ever G to see this
1:09:51 is I badass on air but I just did now
1:09:54 the the the thing we're g to see out
1:09:57 front is a warehouse of the future right
1:10:00 it's a work cell in a warehouse of the
1:10:02 future and then we're going to also see
1:10:04 the real life version of that being
1:10:05 orchestrated then we'll see an Nvidia
1:10:08 Omniverse version where that starts to
1:10:10 zoom out and integrate multiple
1:10:12 warehouses yeah this has impacts on not
1:10:14 just Warehouse we were talking about the
1:10:16 fires Malibu that just burn down a
1:10:19 community right sorryy and you guys
1:10:23 still Brian has a GEOS background we did
1:10:26 a bunch of aerial Imaging we used to fly
1:10:29 uh for FEMA and other First Responders
1:10:32 and so what we're going to be partnering
1:10:34 with coming up in the next year or two
1:10:36 is drones and drone manufacturers to be
1:10:38 able to do that distributed get there
1:10:40 where you need them yeah at the time you
1:10:42 need them same with hurricanes we used
1:10:44 to fly hurricanes right after the
1:10:46 aftermath and you've got no
1:10:47 infrastructure so how do you get
1:10:49 infrastructure quickly right starlink is
1:10:51 one answer but then starlink to drones
1:10:54 trying to help the first respond
1:10:56 so the conversation going on right now
1:10:59 on Tik Tok about you know where the
1:11:01 women nerds are asking the same thing
1:11:03 where the where are the women at yeah
1:11:04 the Nerds just want to date right um I
1:11:07 was just thinking how would women
1:11:09 influence this scene so he here's
1:11:14 what one of the things I am most excited
1:11:17 about with the AI Salon community and
1:11:20 this community here um is the generative
1:11:25 I seen so so you got to realize these
1:11:28 guys are sort of well you'll see here
1:11:32 this thing that they're talking about is
1:11:34 at the protocol level like at the IP
1:11:36 protocol level it's just it's more of a
1:11:39 bro scene do I wish there were more
1:11:41 women in it yeah I wish there were more
1:11:43 women in it but the generative AI thing
1:11:47 that requires empathy to be good at
1:11:50 it is where I think women are kicking
1:11:53 ass the fact that we had three women
1:11:55 representing the AI salon and educating
1:11:58 our politicians about AI today makes
1:12:01 like makes my heart really warm but
1:12:04 these are very different layers the
1:12:05 layer that we're talking about here is
1:12:07 this fundamental protocol level we we'll
1:12:09 get to what this is about in a second
1:12:10 it's only 15 minutes so we'll we'll they
1:12:13 they get into they're just kind of
1:12:14 shooting the [ __ ] right now but they
1:12:16 start getting into what the paranet is
1:12:18 here in a second but uh yeah we need
1:12:21 more [ __ ] women in it but anyway I
1:12:23 like I I think that because generative
1:12:29 AI is a is a language and empathy like
1:12:32 if you are more empathetic you're
1:12:35 actually better at generative
1:12:37 AI than people that are not and I think
1:12:41 this is an opportunity for people that
1:12:43 have good empathy skills to really step
1:12:45 into their [ __ ] power and that that
1:12:47 I'm super excited about so so let's
1:12:50 let's distinguish that a bit but yeah is
1:12:52 is the tech industry still bro [ __ ]
1:12:55 riff
1:12:56 yep it sure
1:12:59 is this
1:13:01 par yeah turn up the volume on this when
1:13:04 I start again
1:13:05 sure though has impacts on everything
1:13:09 from a music festival that's going to
1:13:11 need autonomous cards to get to it a
1:13:13 warehouse where you're going to need to
1:13:15 control a bunch of robots and humans
1:13:18 working together what what else does
1:13:21 what does paranet do that is really
1:13:23 magic here and what why am I here
1:13:25 because I'm here wait Shane Shane saw
1:13:29 said I love working with creative women
1:13:30 then he drew an arrow pointing to
1:13:32 himself Rel reluctant empath I'm telling
1:13:35 you man the the the again it's like
1:13:38 revenge of the liberal arts Majors re
1:13:40 like if you're good at if you're good at
1:13:42 people you're gonna be good at this
1:13:44 stuff
1:13:45 okay it's more like CIS rific Yeah well
1:13:49 yeah that too um us wom women are
1:13:53 infiltrating it okay here we go so now
1:13:54 he's going to talk about what the
1:13:55 paranet is yeah paranet p a r a NE T and
1:13:59 think internet paranet you you'll you'll
1:14:02 hear in a second this is way down at the
1:14:04 protocol level is what these guys are
1:14:06 building but once you understand what
1:14:08 the [ __ ] they're talking about combined
1:14:10 with what David Shapiro just said all of
1:14:13 a sudden you go holy [ __ ] like this is
1:14:16 coming and it's coming fast okay here we
1:14:19 go wait who's the guy on the
1:14:23 salon ton skier who's the guy on the
1:14:25 salon talking
1:14:27 about discovering ai11 uh I don't know
1:14:30 he he he left there was he was he he he
1:14:33 was not uh he was not living the values
1:14:36 of the salon so he's no longer there I
1:14:38 don't know what he was talking about and
1:14:40 he was not willing to share it so that
1:14:43 that's that's no longer
1:14:46 uh an active conversation see
1:14:50 orchestrating all of these B thing it
1:14:52 does is it can orchestrate in a
1:14:54 heterogeneous as
1:14:56 almost anything so this is the critical
1:14:58 piece right it can orchestrate in a
1:15:01 heterogeneous
1:15:04 fashion almost anything okay go on
1:15:08 humans robots computers Enterprise
1:15:11 workflow when you're at that you know
1:15:13 Coachella and youve got bottlenecks of
1:15:15 people trying to come through a facility
1:15:17 right well why wouldn't you just
1:15:18 automatically open up other Gates that
1:15:21 actually can have like CV security and
1:15:23 just monitor the flow and move it back
1:15:25 back and forth instead of bottlenecking
1:15:27 the way it happens to right so
1:15:31 so he they'll get more into this but
1:15:33 like what he's talking about here is
1:15:37 imagine if all of the systems of the
1:15:39 world that includes humans and systems
1:15:42 that don't normally talk to one another
1:15:45 what if those systems could all talk to
1:15:47 one another
1:15:49 dynamically that's what they're talking
1:15:50 about
1:15:53 here okay I'm not a big Coachella I
1:15:55 don't look like no I went and you know
1:15:57 they have 3,000 Uber cars driving people
1:15:59 around with humans someday soon there's
1:16:01 going to be autonomous networks right U
1:16:04 whether it's or or Tesla your roboxing
1:16:08 right I'm imagine sech pricing except
1:16:10 for you're not your goal isn't
1:16:11 necessarily to cost you know charge more
1:16:14 but to in fact get more people through
1:16:16 yeah make more people happy what what is
1:16:18 so what what is it actually doing and
1:16:21 why is this new kind of network protocol
1:16:24 really it's a smart net turn it to Jim
1:16:26 for hey Jim come on
1:16:28 over what why what how does this work
1:16:32 and why do we need a new network
1:16:33 protocol for this new robotic autonomous
1:16:36 ah that we're heading into so this
1:16:39 there's like 9,000 languages right yeah
1:16:41 it's been documented there's 4,000
1:16:44 protocols um so 9,000 computer languages
1:16:47 4,000 protocols so think tcpip HTTP
1:16:51 those are all protocols 9,000 languages
1:16:54 4,000 protocols
1:16:56 what's missing has been this this
1:16:58 protocol stack called The OSI stack with
1:17:00 seven layers on it
1:17:01 forever and we know that internet really
1:17:05 is insecure we have to kind of plug and
1:17:07 patch things and buy a bunch of software
1:17:08 it's $9.5 trillion dollar worth of cyber
1:17:11 crime in the world which is just going
1:17:13 to grow with AI so we saw this as a
1:17:16 problem we saw that it's not just going
1:17:18 to be humans using the internet anymore
1:17:20 it's be machines using the
1:17:22 internet P Mo how many of those are in
1:17:24 use 9,000 is silly well 9,000 is silly
1:17:28 Pate until you've got like some obscure
1:17:32 language that's running some control
1:17:34 system at a nuclear power plant um that
1:17:37 needs to be interfaced with that that's
1:17:40 that that's what he's talking about is
1:17:42 the ability to interface th the ability
1:17:45 to coordinate things across domains um
1:17:49 that that currently doesn't exist and do
1:17:51 it in a secure way machines are using
1:17:52 the internet how are they talking to
1:17:53 each other so what if you create a new
1:17:55 new protocol but a new layer so what we
1:17:58 did most is what most people don't
1:18:01 realize is when the web came out the
1:18:02 internet had been out since 1983 tcpip
1:18:05 right and we had email SMTP all these
1:18:07 protocols when the web came out HTTP it
1:18:09 allowed you to transfer HTML from a
1:18:12 server to a client and the rest is
1:18:14 history there's billions of people using
1:18:15 these these things and all kinds of
1:18:17 other Protocols are out there for doing
1:18:20 you know iot and modbus mqtt and so
1:18:23 forth and those are all good protocols
1:18:24 they all nothing changes so the IP
1:18:27 infrastructure and the internet as it
1:18:28 exists today nothing changes but it's
1:18:31 broken from security and Trust
1:18:34 standpoint so we said let's create a new
1:18:36 overlay an overlay like the web is an
1:18:38 overlay when you use zoom it's an
1:18:40 overlay only you and your friends are
1:18:42 talking in Zoom no one else gets in why
1:18:43 can't we create a network overlay such
1:18:47 that machines can be up there so we did
1:18:50 that but then what are these machines
1:18:52 but we actually had the yeah this gym is
1:18:54 the DUS
1:18:56 uh create something like the original
1:18:58 operating system like dos you know yeah
1:19:00 which you were involved in yeah and um a
1:19:04 couple years after that uh Andrew S
1:19:07 tanon who wrote the book computer
1:19:08 networking also wrote a sacred is
1:19:11 profane this is stepping on a lot of
1:19:13 toes it's it is stepping on a lot of
1:19:15 toes but agents are going to step on a
1:19:18 lot of toes as well and and yeah I I
1:19:21 agree what what what you're talking
1:19:23 about is
1:19:26 there there are massive Industries built
1:19:29 on top of the insecurity of the current
1:19:34 Internet and if you all of a sudden
1:19:37 throw in autonomous agents that can go
1:19:39 take action on everyone's behalf that's
1:19:42 where that security vulnerability starts
1:19:45 to become completely just unimaginable
1:19:48 how bad it could be and that's that's
1:19:50 what he's trying to attack here a
1:19:53 textbook called distributed operating
1:19:54 system
1:19:56 there aren't any distributive operating
1:19:57 systems so what we needed well something
1:20:00 like this could be used to start World
1:20:01 War II I think what he's talking about
1:20:03 is this could also be used to prevent
1:20:05 was this huge history of the internet IP
1:20:08 infrastructure and for somebody to come
1:20:11 along autona and the paranet to build
1:20:14 that overlay so now we can have
1:20:15 distribut operating system what that
1:20:16 means is you have to have new machines
1:20:18 we're not taking new CPUs or gpus like
1:20:20 that we're talking about a new machine
1:20:21 it's called an actor and this stuff goes
1:20:24 all the way back to
1:20:25 MIT um and the early pure thinkers of of
1:20:30 AI they thought about actors and
1:20:33 planning languages so we took the idea
1:20:35 of actors and said okay what if every
1:20:36 machine every software every API every
1:20:38 robot every web page every database
1:20:40 every Everything could be an actor like
1:20:43 Paul was talking about with you know
1:20:44 World Management systems and robots and
1:20:46 so forth he was referring to we need to
1:20:48 get all these things on this new secure
1:20:50 trusted Network again it's the Internet
1:20:52 it's IP infrastructure but it's the
1:20:54 paranet paranet now is this overlay and
1:20:56 machines only talk to each other through
1:20:58 this protocol now even though they're
1:20:59 written in C or python or whatever
1:21:01 doesn't matter we can quickly aoriz them
1:21:03 put them on a network and when they're
1:21:04 on that Network they can securely talk
1:21:06 to we can quick quickly aoriz them so
1:21:08 basically what they're talking about is
1:21:09 creating like a metadata wrapper that
1:21:12 points to something doing something down
1:21:15 at the at the a lower level in trust we
1:21:19 had had a b did a bunch of things around
1:21:20 security that we can talk about another
1:21:23 time but yeah you want to keep
1:21:25 Iran so if if you're all if if anyone's
1:21:29 new here I see Jason came in who this
1:21:31 and what what have we been talking about
1:21:32 tonight so what we're talking about
1:21:34 tonight is David Shapiro's basically
1:21:36 said that
1:21:38 that the the main problem with agents is
1:21:42 how do you take this smart brain this
1:21:44 llm this brain in a jar that's not
1:21:46 really connected anything and right now
1:21:48 currently requires our expertise to make
1:21:51 it work and how do you solve the how
1:21:55 does it decide what Tod do problem and
1:21:57 what Shapiro argues is that problem has
1:21:59 effectively been cracked and once that's
1:22:02 released on the world once you solve
1:22:04 that for one agent you've effectively
1:22:05 solved it for all agents so that was the
1:22:09 first thing we talked about this is this
1:22:12 is a guy that has been in the business
1:22:15 for years effectively proposing or has
1:22:17 built a protocol lay layer called the
1:22:20 pro the paranet that sits on top of the
1:22:24 current Internet technology stack and is
1:22:27 a coordination layer that coordinates
1:22:30 between objects and functions of
1:22:33 everything that's happening on the
1:22:34 internet so that's that's what he's
1:22:36 talking about out of your network yeah
1:22:38 but but yeah we want to problem with the
1:22:40 whole internet right now if something
1:22:41 gets on your on your network if it gets
1:22:43 through zero trust security you're done
1:22:46 right you have to start over right the
1:22:48 paranet says if you're using a protocol
1:22:50 if a no gets infected the network
1:22:52 realizes it's bad and gets rid of it so
1:22:54 we call that so that's that's an
1:22:55 important thing right there right in in
1:22:57 in the current Internet if you're hacked
1:22:59 and someone gets into your system it's
1:23:01 done what he's saying is if you're
1:23:03 operating at this paranet level and
1:23:05 someone grabs a node the system will be
1:23:07 smart enough to know that that node is
1:23:09 [ __ ] up and it'll actually just shut
1:23:11 it out of the the the
1:23:15 Swarm negative trust security so to
1:23:18 answer your question why do we need a
1:23:19 network because trillions of machines
1:23:22 are coming yeah right we're going to be
1:23:25 machines right and they have to talk to
1:23:26 each other and they all have to work
1:23:28 together but the only way to do that is
1:23:30 through an overlay and a exclusive
1:23:32 protected secure trusted uh
1:23:35 collaboration protocol and this protocol
1:23:37 is pretty cool because it actually does
1:23:41 um Q&A so machines are doing Q&A with
1:23:44 each other
1:23:45 so everything's getting get more
1:23:46 intelligent py don't out every single
1:23:49 thing he's saying so so here's the thing
1:23:52 this is likely a private company he's
1:23:54 trying to spin up a protocol layer but
1:23:55 this is clearly a commercial entity
1:23:58 right so whether or not this thing ends
1:24:00 up being a protocol whether or not this
1:24:03 thing ends up even launching I don't
1:24:06 know like I don't know where this is the
1:24:08 reason I'm showing this is
1:24:11 that there is fairly significant work
1:24:15 being done on the orchestration layer so
1:24:18 if the current um Frontier Model
1:24:21 companies solve the how do we use this
1:24:24 smart brain we have an ajar problem and
1:24:26 you have something like this that can
1:24:28 start to coordinate it things are going
1:24:30 to accelerate very fast he's spouting
1:24:32 nonsense Kyle I have actual credentials
1:24:34 for this type of thing yeah but it's
1:24:36 it's not binary just because he said dos
1:24:39 was the original operating system I
1:24:41 assume he meant for for PCs rather than
1:24:44 mainframes and and Minis and things like
1:24:46 that I get your point but don't throw
1:24:48 the baby out with the bathwater because
1:24:49 it's it there's this isn't necessarily
1:24:53 just this specific guy and this specific
1:24:55 company there are probably lots of folks
1:24:57 doing this
1:24:59 but one of the questions I've always had
1:25:02 is if you have if you introduce billions
1:25:05 of things onto the internet with lots
1:25:08 and lots of different protocols how do
1:25:09 you coordinate all that and so that's
1:25:12 what that's okay so whatever if you if
1:25:18 you want to listen to pate and what I'm
1:25:20 showing you is completely useless all
1:25:22 the AI and agentic Frameworks all that
1:25:24 stuff you hear by agents they all need
1:25:26 some place to talk robots need a place
1:25:28 to talk but the weird thing about robots
1:25:30 is that they they are in the OT side of
1:25:32 the world not the it set right OT is
1:25:34 where all the hard engineering goes the
1:25:36 electrical engineers they now need to
1:25:37 place well how do I get how do I put
1:25:39 this in your Warehouse how do I put this
1:25:40 in your factory yeah well you got to do
1:25:42 Network you security no you don't just
1:25:43 put on the paranet it discovers
1:25:46 everybody else out they're all secure
1:25:47 and they run so that's that's this is
1:25:49 like air traffic control for the
1:25:50 internet well it's like air traffic
1:25:52 control above the internet like yeah
1:25:55 it's not a bad it's not a bad thing
1:25:57 where you've got you know planes landing
1:26:00 on the internet right so you've got
1:26:01 robots and software and all sorts of
1:26:04 things doing all this stuff down at the
1:26:06 internet level and this is saying how do
1:26:09 you get those things to talk to one
1:26:12 another it's the paret in a nutshell
1:26:15 wow this has deep implications on the
1:26:18 future of everything doesn't it well
1:26:20 because because as humans just like s TP
1:26:25 had exactly our original thoughts SMTP
1:26:29 basic protocols but start with security
1:26:33 everything is secure from the beginning
1:26:35 to the end sorry Jim email was never
1:26:37 secure uh the web wasn't really secure
1:26:40 until they put s on it but the horses
1:26:41 left the barn so bad things happen but
1:26:45 you need more than just encryption you
1:26:46 need a trusted Network that's
1:26:48 intelligent you need you need security
1:26:50 actors on your network that are looking
1:26:51 at all the other actors doing the work
1:26:53 so you need to be able to have have a
1:26:55 programmable Network that's intelligent
1:26:56 that can know what's going on one of the
1:26:59 biggest problems in networking
1:27:01 particularly in data centers is just
1:27:02 setting up routing
1:27:03 tables there's millions of hours every
1:27:06 day being spent setting up your router
1:27:08 tables to your switches your network on
1:27:10 the paret and this new protocol of
1:27:12 course you still have routing tables you
1:27:13 have routers and your switches but
1:27:14 you're on this new semantic layer now
1:27:16 now all the actors these new machines
1:27:18 they don't go through routing tables
1:27:19 they just say here's my skills here's
1:27:21 what I could ask and then all the
1:27:23 different nodes with all the actors
1:27:24 skills they all discover each other and
1:27:26 no one gets to ask a question they're
1:27:28 not they're not able to ask yeah and
1:27:30 that's how you create security and Trust
1:27:31 going forward now I'm I'm here because
1:27:34 I'm trying to get ahead of Elon
1:27:36 Musk yesterday I met a a voice company
1:27:40 and he immediately announced a new roock
1:27:42 with multi-party voice to respond right
1:27:45 and uh it's hard to get ahead of that
1:27:47 man right um let's say he's building a
1:27:51 paranet of his own to run his autonomous
1:27:54 car now and his robot yes so he has a
1:27:56 vertical why does he need to throw all
1:27:58 that away and come and see you no one
1:28:00 needs we're not seeing through
1:28:04 any overlay
1:28:06 augment yeah so why would he need to
1:28:09 come and see you and learn about paranet
1:28:11 learn about this new network protocol to
1:28:13 integrate with the the Tesla system he's
1:28:16 building so Tesla
1:28:18 is um he's all about autonomy right so
1:28:21 he's doing humanoid robots Optimus he's
1:28:23 doing
1:28:25 you you came here in a in a robot car I
1:28:27 did my robot drove me here so that
1:28:29 that's on a network but the thing is
1:28:32 nothing that humans build is truly
1:28:34 autonomous it has to be on some sort of
1:28:37 network with information orchestration
1:28:38 collaboration workflow planning and
1:28:40 observation right yeah so we built a
1:28:43 generic version of whatever he built
1:28:46 that has those five properties again
1:28:48 orchestration workflow
1:28:50 collaboration planning and observation
1:28:53 so we build a network that you can do so
1:28:55 you could build the Tesla Network on the
1:28:56 paranet you don't need to because he's
1:28:58 already done it but if you need to build
1:28:59 an autonomous system for your factory
1:29:01 for your Warehouse yeah someone just
1:29:02 made a comment about beta VHS versus
1:29:06 beta yeah I mean this is so so here's
1:29:10 the thing this is a guy that's that's
1:29:13 talking about you know building the
1:29:15 protocols to be able to make VHS right
1:29:18 and Tesla's doing the same thing he's
1:29:20 done over at or Elon Musk is doing this
1:29:24 over at Tesla those are going to be
1:29:26 competing protocols does this guy end up
1:29:29 winning I don't [ __ ] know Does Elon
1:29:31 Musk end up winning he'll probably at
1:29:32 least be in the game right um but
1:29:35 someone has to crack this layer and and
1:29:38 what what I was surprised by is how far
1:29:42 how far down the road this particular
1:29:45 version seems to be that's that's a
1:29:47 generalized protocol now one of the
1:29:50 hardest things to do in the world is
1:29:51 introduce a new protocol
1:29:54 right how do you get people to adopt
1:29:56 your protocol well you know that's VHS
1:29:59 versus use us there's GNA be a lot of
1:30:01 failures use tless Network yeah I don't
1:30:04 know what he's using for his humanoids
1:30:05 but he probably should use something
1:30:06 like the paranet yeah because we built
1:30:09 it to allow robots become actors in the
1:30:12 network the thing that people don't
1:30:14 assuming he has a paranet himself
1:30:17 because he's building stuff like that to
1:30:19 have his robots work together and his
1:30:21 autonomous cars work together and figure
1:30:23 out where they're supposed to be to pick
1:30:24 up a ride or stuff like that can he hook
1:30:27 his paranet in to your paranet and now
1:30:30 they could ours level ours is Lay layer
1:30:33 eight okay I'm just going to say it when
1:30:35 you do the layer above seven it's eight
1:30:37 here we are we're at this layer he's not
1:30:38 doing layer eight he can certainly plug
1:30:40 into that yeah right so and that
1:30:43 explains how a Ford could join their
1:30:45 robots into the Tesla Network into the
1:30:48 paranet right working with the robots
1:30:51 I'm about to see up in front of the H
1:30:53 right yeah and so then you know it's
1:30:56 just not going to happen overnight yeah
1:30:57 so we're partnering with Accenture right
1:31:01 so so here's the hey we're just a humble
1:31:03 startup trying trying to write a new
1:31:05 protocol this is not going to happen
1:31:07 overnight means that this is somewhere
1:31:10 between vaporware and something right
1:31:12 and if you listen to all the stuff that
1:31:13 c c talking about CEs it's all about
1:31:15 autonomy and Ai and trust and people
1:31:18 working with AI systems together and
1:31:20 agents well that's the parent I mean
1:31:23 that's what it is so that's why we're
1:31:24 partnering with them working through
1:31:26 this and it's very early like we just
1:31:27 you know yeah we don't there's only a
1:31:30 few thousand human ey BR that's there
1:31:33 will be million many millions yes soon
1:31:37 if you ask if you yeah so his company I
1:31:40 don't I forget what his company's called
1:31:42 his company is uh has partnered with
1:31:45 Accenture so this is this is backstage
1:31:48 at CES two days ago where Accenture did
1:31:51 this Warehouse of the future demo
1:31:55 that all of the robots were Co
1:31:58 coordinated on this paranet um protocol
1:32:01 yes the musk it's even more than that I
1:32:03 I I did ask bus I asked him for his
1:32:06 spreadsheet actually predicting how many
1:32:09 robots he's gonna sell and he said you
1:32:11 don't need a spreadsheet it's gonna be
1:32:13 so [ __ ] many of
1:32:17 them because we were talking about
1:32:19 people with bad needs and you know just
1:32:21 that one little tiny Market 60 million
1:32:23 people without bad needs they need a
1:32:24 robot to help them out right just that
1:32:26 one little sliver so I wanted that's FR
1:32:29 he has to predict all the markets right
1:32:32 because he has something like that he
1:32:34 has he has grock you know that knows
1:32:36 everything at least does know it yet but
1:32:39 it soon will oh as soon as I post this
1:32:42 video grock will know all it needs to
1:32:44 know yeah soon as you put it up that's
1:32:46 an talk to Gro what does Gro need to now
1:32:50 you just needs to ask hey what's what
1:32:51 should what should I be doing with the
1:32:53 baronet needs to know what it needs what
1:32:56 does grock need to know to to hook into
1:32:58 the parad yeah so what what are the API
1:33:01 calls that grock needs to
1:33:04 know more to come we publish a dev kit
1:33:07 grock will be all over yeah and uh
1:33:11 perinet is we're on the web with grock
1:33:13 fan when we publish a dev kit so they
1:33:15 don't have a dev kit yet so this is I
1:33:19 assume this is really well thought out
1:33:21 probably some papers probably some stuff
1:33:24 that not thought out to P's Point um
1:33:27 probably spitting duct tape at this
1:33:29 point so they so I don't know how how
1:33:32 far in they are paranet you'll find it
1:33:34 at
1:33:35 pet. you know and aon.com autonoma
1:33:40 doccom paranet doai and autonoma Doc O
1:33:45 N O ma a okay and does this cost money
1:33:51 to integrate right now what's your
1:33:54 business model right now you're in the
1:33:56 back room on the Cs right you're in the
1:33:57 back like this where you kind of this is
1:34:00 the dirty room make they never like to
1:34:03 let the press back here because it's
1:34:04 dirty not supp talk see he's even
1:34:06 shaking his head we're working through
1:34:09 various models right now right so what
1:34:10 we want to do is some combination we
1:34:11 don't know our pricing open source and
1:34:13 Licensing such that we make it easy for
1:34:16 the world to use this one of the things
1:34:17 we did was we we went and we got arpanet
1:34:20 doai and so we want the researchers of
1:34:22 the world all the universities to be
1:34:23 able to use
1:34:24 freely and arpanet was really the
1:34:27 beginning of sharing information across
1:34:29 the world arpet will be distributed
1:34:32 Intelligence on a paranet network that's
1:34:35 run by researchers so we're going to
1:34:37 make that free everyone and then um for
1:34:42 paranet it's really kind of like a
1:34:45 licensing world with certain amount of
1:34:47 bandwidth on the network haven't worked
1:34:49 out that perfectly yet but basically we
1:34:51 have a distributed Network operating
1:34:52 system every not is out there if you're
1:34:55 using it commercially that'll be a
1:34:56 license and then how much bandwidth how
1:34:58 many actors with how many skills are
1:35:00 being you know interacting at any point
1:35:03 in time kind of like a network fee so
1:35:05 still working through that still early
1:35:07 we haven't got our day kit out we'll
1:35:08 have that in a few months and then you
1:35:10 can put that in a Gro yeah few months
1:35:12 from a dev kit right so so these guys
1:35:13 are early so so because they have a good
1:35:17 team they partnered with Accenture they
1:35:19 they might even have partnered with
1:35:20 Accenture to help build some of this
1:35:22 [ __ ] so um so anyway that's what's going
1:35:25 on all right enough enough comedy jokes
1:35:32 so there you have it I think those two
1:35:35 things back to back you know on the same
1:35:39 day they were significant for me here
1:35:40 here's the here's the implication
1:35:43 independent of whether or not perinet as
1:35:46 that guy described it is his thing or
1:35:48 some other
1:35:50 thing one of the things that I have not
1:35:52 been able to wrap my head around is how
1:35:54 do you
1:35:55 introduce millions or
1:35:58 billions of autonomous agents into an
1:36:01 insecure environment like the internet
1:36:04 without it being absolute [ __ ] chaos
1:36:08 um it's it's got to be solved and I you
1:36:10 know I think individual companies are
1:36:12 going to solve it on their own so open
1:36:14 AI is going to have their version
1:36:16 grock's gonna have its version um you
1:36:19 know or you know Tesla and all that are
1:36:22 going to have their version is going to
1:36:24 have their version the Chinese are going
1:36:25 to have their version how do all those
1:36:27 things talk to one another this is the
1:36:30 first thing I've heard that I start to
1:36:31 say okay I can at least see conceptually
1:36:34 how you could create a layer above what
1:36:36 exists right now which feels like an
1:36:38 abject [ __ ]
1:36:39 nightmare
1:36:42 um all
1:36:47 right Windows is still basically the dis
1:36:50 operating
1:36:51 system uh uh
1:37:00 sacred is profane you may be splitting
1:37:02 hairs paint for better or worse that's
1:37:06 my whole
1:37:10 stick I mean listen paint is it's it's
1:37:15 all about definitions and Clarity I am
1:37:18 about muddiness and what's possible so I
1:37:22 think that's why having p here's a
1:37:24 really good voice a really good balance
1:37:25 to my to my optimism and and fuzziness
1:37:30 but um here's the point my the point and
1:37:34 the reason that I shared let me go back
1:37:37 and show you where it
1:37:46 is the reason that I
1:37:49 shared this document so if you go to the
1:37:53 AI salon I put this in in I put a post
1:37:57 of this in um the code showand tell
1:38:02 section I put this in
1:38:04 Pates um in Pat's uh club and I put this
1:38:09 in the water cooler
1:38:11 conversation so this is a document that
1:38:13 I did today using goo Google deep
1:38:17 research of um AI
1:38:21 agents broken into three categories no
1:38:24 code low code and um and developer
1:38:27 focused
1:38:28 platforms
1:38:32 um you don't necessarily have to go
1:38:34 build agents you don't has have to
1:38:37 necessarily um dig deep into this but I
1:38:41 I feel that right
1:38:43 now just like building custom gpts was a
1:38:46 good augmentation to understanding
1:38:49 prompting and chat GPT in
1:38:52 general understanding what agents are as
1:38:55 they are today and understanding what
1:38:57 their limitations are
1:39:00 today is going to allow you to be as
1:39:03 prepared as possible for the thing that
1:39:06 David Shapiro described which is
1:39:08 basically once one shows up The Dominoes
1:39:12 fall very very
1:39:15 quickly and it seems to me unless
1:39:18 everyone's [ __ ]
1:39:20 lying that that first Domino is going to
1:39:22 fall in 2020
1:39:25 and it it there might be multiple
1:39:27 dominoes that fall in 2025 it could be
1:39:30 2026 but even if it's
1:39:33 2026 That's mighty [ __ ] close for for
1:39:37 a series of Technologies being Unleashed
1:39:40 on the world that can do many people's
1:39:43 jobs do the jobs not just do parts of
1:39:46 the jobs that require the human to be
1:39:48 there but do their jobs so so
1:39:54 I kind of feel like 2024 was the year
1:39:58 2023 was the year we got excited 2024
1:40:02 was the year we got to take a breath and
1:40:04 sort of take in what was Unleashed on
1:40:05 the world 2025 feels to me like the year
1:40:09 where shit's getting serious and so why
1:40:12 I wanted to talk about this tonight is I
1:40:14 want people to
1:40:17 um I I would strongly encourage you if
1:40:20 you haven't been paying attention to
1:40:21 agents just start paying attention to
1:40:23 them
1:40:24 um do your best to sift the wheat from
1:40:27 the chaff and if you've got a little
1:40:28 curiosity and geekiness in you come play
1:40:31 with this list and just go look at what
1:40:32 these platforms are and what they claim
1:40:35 they can do right even if you just see
1:40:37 that just what do they claim they can do
1:40:39 there's also a bunch of Open Source
1:40:41 things out there's there's there's one
1:40:42 that was just talked about I think it
1:40:43 scrapes Reddit and does some [ __ ] for
1:40:46 you that's been open sourced in the past
1:40:48 three or four days um they're they're
1:40:51 starting to get pretty interesting and
1:40:54 if you're one of those people that knows
1:40:55 how to turn loose an agent as soon as
1:40:57 the platforms are are available you
1:41:00 might get a competitive head start on
1:41:03 you know whatever you're trying to do in
1:41:04 the world so all right with that
1:41:08 questions
1:41:11 [Music]
1:41:17 thoughts I on China the impression I get
1:41:21 is gpts went nowhere generally
1:41:25 speaking I've still not made a custom
1:41:30 GPT I'm introducing a hypothetical agent
1:41:33 on
1:41:35 Wednesday oh are you for real CEF or or
1:41:39 is that just a
1:41:40 joke because right now I'm in
1:41:42 hypothetical land but I here's the thing
1:41:44 I don't think we're GNA be in
1:41:45 hypothetical land that that much longer
1:41:52 [Music]
1:42:00 all right
1:42:05 fantastic 2025 the year of
1:42:09 disruption yeah it's
1:42:12 like yeah to say the
1:42:15 least Kyle as I said earlier can we all
1:42:18 hide in a blanket fort in your living
1:42:20 room for a while Daisy May like that is
1:42:24 the appropriate
1:42:26 sentiment the appropriate sentiment from
1:42:29 where I sit
1:42:31 is like
1:42:34 listen a lot of what I'm talking about
1:42:36 tonight still lives in the realm of chat
1:42:39 TMZ right which drives Pate [ __ ]
1:42:42 crazy a lot of what I'm talking about is
1:42:44 speculation but there's enough
1:42:47 indicators
1:42:49 that this might not just be hype there's
1:42:52 enough indicators is that yeah in 2025
1:42:56 some shit's going to drop that
1:42:58 fundamentally changes the game and yeah
1:43:00 I feel like an appropriate response to
1:43:03 that is I want my doie give me a hot
1:43:06 chocolate my doie and a warm blanket I'm
1:43:08 just gonna sit out the next three years
1:43:11 I think that's an appropriate response
1:43:14 um and if you want to be prepared if you
1:43:17 want to not be blindsided by this stuff
1:43:20 then I think it's I think it's worth
1:43:21 putting time into this so there you go
1:43:24 Jim Ross this makes my head hurt but
1:43:26 it's amazing I know but Jim Ross so so
1:43:30 you know okay so where you are right now
1:43:34 with lovable dodev so so Jim Ross is if
1:43:38 you don't know Jim Ross awesome dude
1:43:40 he's got a self- storage business called
1:43:42 thre Mile Storage where he helps people
1:43:46 run storage businesses right where they
1:43:48 own storage unit businesses um and in
1:43:51 the past week he sent me two different
1:43:53 Publications that he's written with
1:43:54 lovable dodev where he's like oh my God
1:43:57 like the one thing you said was you
1:43:59 wanted to do this a year or two ago and
1:44:02 the dev that you met with said it would
1:44:04 be a minimum of $5,000 and you did it in
1:44:07 a couple of hours right or an hour or
1:44:09 whatever it was and then and then you
1:44:11 sent me the next application you wrote
1:44:13 like the next day right so so this agent
1:44:17 stuff right now is head spinny we don't
1:44:20 quite get it kind of stuff but at some
1:44:22 point there will be some version
1:44:24 of an agent spinner uper that'll be like
1:44:26 lovable dodev
1:44:28 like we will be able to use these things
1:44:31 um quite easily like it may take a year
1:44:35 or two to get the the systems or you
1:44:38 know to get the right company to to
1:44:40 introduce the right interface for it
1:44:46 um but I think that I think that it's
1:44:49 it's it's coming really fast and you
1:44:51 know like for your business Jim it could
1:44:53 be
1:44:54 you know rather than having to hire a
1:44:56 manager to go manage all of your
1:44:58 different properties you could have you
1:45:00 know a suite of Agents go manage all of
1:45:02 your properties right so you could
1:45:04 literally become a oneman show that's
1:45:06 managing you know dozens of properties
1:45:09 around the country is
1:45:12 possible oh good lord all right
1:45:17 appropriate response push the pink bow
1:45:19 button you can make money with
1:45:24 my gpts are absolute obsolete
1:45:28 already um where are you all posting
1:45:32 your music but beyond the salon um if
1:45:36 it's good enough I put I put Quantum
1:45:38 Cipher on Spotify I just put it up there
1:45:40 as a
1:45:42 song one man show that's the plan
1:45:45 exactly
1:45:50 exactly looks like Cosmic lovers on a on
1:45:53 a roll to get someone hired which is
1:45:56 awesome gonna hire from the salon
1:45:58 fantastic love
1:46:00 that love
1:46:02 that yeah I would say you know posting
1:46:04 your music on uh on uh on the salon's
1:46:09 good I know that there's at least one
1:46:12 there's probably like five now I would
1:46:14 go on futurepedia
1:46:16 doio and see if you can find like is
1:46:20 someone creating a Spotify for AI music
1:46:24 I personally think that's a limited
1:46:26 model because I don't
1:46:29 think I don't think it's going to be
1:46:32 that much longer that we can actually
1:46:33 distinguish between AI music and music
1:46:36 like it's e it's either good music or
1:46:38 it's not um but I think short term there
1:46:41 are probably going to be platforms that
1:46:43 are like post your AI music here and
1:46:45 we'll create the you know the Spotify
1:46:48 for AI music I'm sure there's a bunch of
1:46:49 those out there in fact I kind of
1:46:51 remember there being one or two of them
1:46:54 um all right I'm putting my music on
1:46:57 United Artists there you go it goes to a
1:46:59 ton of platforms that
1:47:03 way I don't think people can tell very
1:47:06 well today
1:47:08 um have I have I watched megalopolis I
1:47:12 have not watched it yet I've been
1:47:15 listening to other users music in
1:47:18 soono oh um oh what's his name um
1:47:26 [Music]
1:47:28 Jay um he was in our prototyping meeting
1:47:32 today he's using XML in sunno here I'll
1:47:36 just show you this
1:47:38 quick um where did he send it I think
1:47:42 I've got
1:47:44 it frostbitten
1:47:55 so here's this song look at the look at
1:47:58 the prompt for this song or the
1:48:01 lyrics can you see that yeah you can
1:48:03 here let me make it
1:48:07 bigger you can't really see it too good
1:48:10 but that's all
1:48:11 XML but it's
1:48:14 like interval four
1:48:18 4101 drop outro so he's sort of setting
1:48:21 up different modules
1:48:23 and then he defines the patterns so
1:48:26 interval one has those musical patterns
1:48:29 interval two has those those musical
1:48:35 patterns number one feel Soul
1:48:38 interactive module conform and then in
1:48:41 the lyrics of the song he's dropping in
1:48:44 when he wants different patterns and he
1:48:46 says this works that that it's
1:48:48 generating songs that basically follow
1:48:50 the XML recipe up above so I'll share
1:48:53 this in uh in the salon he he might have
1:48:57 already shared it there but I think it's
1:48:58 kind of cool so I'll go
1:49:01 to um music and
1:49:05 audio and then I'll go
1:49:10 from
1:49:13 at
1:49:16 frostbitten his zany XML music prompt
1:49:28 check out the song
1:49:33 and in quotes
1:49:37 lyrics and then ask him how the hell
1:49:42 he's doing
1:49:45 this I had I had chat GPT analyze the
1:49:48 XML and it's it's not based on anything
1:49:51 that that I know so might be that he
1:49:53 just came up with this idea and is
1:49:55 having chat GPT create a custom XML
1:49:58 schema but that's in the music and audio
1:50:01 show in there you
1:50:04 go all right
1:50:07 people we don't need to notify
1:50:13 everyone it's getting late I think it is
1:50:18 time time to get
1:50:21 rolling rolling rolling
1:50:23 rolling any final thoughts
1:50:33 questions our students will be loaded up
1:50:35 and securing AI certifications through a
1:50:37 new AI course in high school very
1:50:48 cool Silver
1:50:50 Fox you are you are welcome welcome
1:50:53 thank you AI machine learning dude frost
1:50:56 bitten can't wait to see yeah it's it's
1:51:00 cool I I I want to get my head around it
1:51:02 but I also you know what's amazing about
1:51:04 the frost bit's approach to sunno is
1:51:08 he's taking a full-on programming
1:51:10 approach to it and it's working I take
1:51:12 this sort of just vomit words at it and
1:51:16 just keep vomiting words until I get it
1:51:19 right
1:51:20 like the breadth of how you can get
1:51:24 results out of large language models and
1:51:27 and these tools it's it's so wide like
1:51:29 any way your brain works so so long as
1:51:32 you're treating it like this
1:51:34 collaborative interaction with these
1:51:36 tools you can get you can get it to work
1:51:38 there's there's kind of an infinite
1:51:39 number of ways to do it oh I've vomited
1:51:42 everything at it oh that's funny
1:51:44 frostbitten says that's awesome um
1:51:48 Brandon hey I just landed what did I
1:51:49 miss we spent the whole hour on uh jelly
1:51:52 pod
1:51:54 no we
1:52:00 didn't I call suo a curation process oh
1:52:03 my God that thing I did today in Runway
1:52:06 trying to get those those agent you know
1:52:09 those those avatars looking like
1:52:11 something not from a [ __ ] nightmare I
1:52:14 mean it was
1:52:16 probably I don't know 30 or 40
1:52:19 Generations per little segment I did
1:52:24 um I am excited for these tools to be
1:52:27 less janky but you know the opportunity
1:52:30 is is the the biggest opportunity is
1:52:33 when the tools are the jankiest let me
1:52:35 just leave you with that if if you're
1:52:37 frustrated with the Jank of AI right now
1:52:40 that means that you're in in a good
1:52:42 place because it means you're early and
1:52:44 your opportunity is bigger than you
1:52:46 realize all
1:52:49 [Music]
1:52:51 right yeah the people that make fun of
1:52:53 AI music because it's not real as silly
1:52:55 I the people that make fun of AI music I
1:52:58 can promise you have never made AI music
1:53:01 right if you look at it from the
1:53:03 outside you can watch someone just go
1:53:06 push a button and out comes music and
1:53:09 what the what the
1:53:12 doomers um contend is that that's what
1:53:16 everyone's doing everyone's just being
1:53:18 lazy most people are not being lazy that
1:53:21 are trying to make music that's worth
1:53:23 listening to they're putting a lot of
1:53:24 time into it they're putting a lot of
1:53:26 effort into it they're putting a lot of
1:53:27 production into it um it's just how you
1:53:30 do that production is different than how
1:53:33 we currently do it but didn't we have
1:53:35 the same [ __ ] when we went from you know
1:53:37 eight track tape recorders to digital
1:53:40 audio workstations and midi and sampling
1:53:44 I mean [ __ ] sampling in the 80s that's
1:53:47 not music that's just stealing other
1:53:48 people's music right like we've been
1:53:50 through this before it's the same
1:53:52 [ __ ] [ __ ]
1:53:54 it's like but whatever this is fancier
1:53:58 this is fancier
1:54:02 sampling they're the same type of people
1:54:04 that were saying Electronica wasn't
1:54:05 music 30 years ago exactly
1:54:09 exactly you know the other thing that
1:54:11 that absolutely blows my [ __ ] mind is
1:54:14 that the same people that are afraid of
1:54:17 AI replacing them are like well then
1:54:20 there's you know if it's this good then
1:54:22 there's clearly no place for artists
1:54:25 it's like
1:54:26 [ __ ] the whole point of an artist
1:54:28 like what is an artist an artist is
1:54:31 someone who looks at the same world that
1:54:34 you and I see and they see it
1:54:36 differently right so artists are going
1:54:39 to look at the same AI tools that you
1:54:42 and I see and they're going to look at
1:54:44 them differently and they're going to go
1:54:46 oh what if I did this with it and then
1:54:48 what if I took that and combined it with
1:54:49 this and then they're going to make a
1:54:51 thing that the rest of us are going to
1:54:54 like go wait how H what is that I've
1:54:58 never seen that before right because an
1:55:01 artist put their point of view into
1:55:04 using these tools that doesn't go
1:55:08 away like like I just I like the [ __ ]
1:55:11 Simplicity I just feel like the
1:55:14 political dialogue because it's so
1:55:16 binary has kind of set us up with this
1:55:19 binary AI good AI evil it is much more
1:55:23 [ __ ] gray than that right much more
1:55:26 [ __ ] gray than that so anyway and
1:55:29 artists are more important than ever so
1:55:33 anyone that tells you they're not
1:55:35 clearly isn't using this [ __ ] all
1:55:39 right Corey
1:55:41 Sandler artiste extraordinaire looking
1:55:44 at things a bit differently a different
1:55:46 box
1:55:47 exactly I mean you should have seen when
1:55:50 we were all off building custom GP s
1:55:53 Cory Sandler's gpts were [ __ ] insane
1:55:56 well they still are crazy glasses
1:55:58 designers and just like you know artists
1:56:02 are artists artists will do things
1:56:05 differently so and artists are still
1:56:07 required it's about thinking outside the
1:56:09 box yeah it's about looking at the world
1:56:12 and having the confidence in who you are
1:56:16 to know what you want and then just go
1:56:18 get these tools to do that for you all
1:56:21 right Kyle please yes
1:56:24 what what am I let's see Kyle my boss
1:56:31 ask teenage Engineers I don't quite know
1:56:34 what you're asking Cosmic lover just DM
1:56:37 me offline we'll talk about it tomorrow
1:56:39 um all right cool I am gonna get on out
1:56:43 of here can Mr ASMR make another
1:56:45 appearance in Tik Tock no I think well
1:56:48 so here's the thing I made that video
1:56:50 the the uh hey are you
1:56:53 scrolling because I get this same woman
1:56:57 like I don't know two times a week hey
1:56:59 are you scrolling same woman same video
1:57:02 and I assume that everybody got that
1:57:05 video but they don't like so I made that
1:57:08 video and I think most people didn't
1:57:10 understand it unless you're an
1:57:13 irregular check teenage engineering
1:57:16 sampler okay I'll check it out
1:57:18 archetypal
1:57:19 architect if you like carpentry
1:57:23 got an AI replacement I'd be like cool a
1:57:26 new tool yeah exactly
1:57:31 exactly all right
1:57:34 cool all right beautiful let's see I got
1:57:37 a
1:57:38 [Music]
1:57:39 video I blocked
1:57:42 [Music]
1:57:47 it all right everybody I'm going to get
1:57:49 out of here I will see you tomorrow so
1:57:51 wait tomorrow's Tuesday nothing going on
1:57:53 it'll be same bat Time same bat place so
1:57:55 8:00 pm tomorrow I don't think there's
1:57:57 anything going on but I don't have a
1:57:59 producer telling me all
1:58:02 right all right um Cosmic lover we'll do
1:58:06 that tomorrow
1:58:08 and because I don't want to start
1:58:10 something new especially if I haven't
1:58:11 seen it before I don't want to start
1:58:12 something new and run down a 30 or 45
1:58:15 minute rabbit hole right at the end here
1:58:17 um all right peace out I will see you
1:58:21 guys uh
1:58:23 tomorrow trying to think if there's
1:58:25 anything else I need to talk about no
1:58:27 not right now all right peace out
1:58:29 everybody