aiLL 9/14/23 - ChatGPT Struggles with Health Insurance While Claude Shines

Description

Chapters

Transcript

0:01 foreign
0:29 welcome welcome
0:32 I am
0:35 going to grab a piece of data oh yeah
0:38 this should be fine oh oh this is
0:41 interesting
0:43 today today today
0:52 hmm
0:56 curious curious let's see save
1:02 I'll be right there
1:07 Emilio's wife am I late
1:09 no but you see what's on screen Emilio's
1:12 wife
1:13 this is your PDF
1:18 um let me save it we gonna do some
1:21 experimenting tonight I don't know if we
1:23 can do this but we'll see
1:25 let's see
1:27 um health care plans
1:32 I mean this is actually a really good
1:34 example of the [ __ ] Insanity that our
1:38 that our world is these days
1:41 all right so there's our learning lab
1:44 welcome everyone
1:46 welcome welcome welcome your disc is
1:49 almost full save space by optimizing
1:51 storage yeah if I weren't a data hoarder
1:54 that would be easy
1:55 [Laughter]
1:58 but you know what are you going to do
2:02 what are you gonna do huh what are you
2:05 gonna do what are you gonna do I'm
2:08 trying to quit apps here hold please
2:11 welcome everybody to the AI learning lab
2:15 we will get started here shortly as
2:18 people gather
2:20 say hi to one another
2:23 print hello world by the way 10 out of
2:25 10 on the name print hello world love
2:27 that name Stevo tapping them
2:30 uh Emilio's wife yes real life problem
2:34 so Kyle Stevo it's going down seriously
2:36 I thought about just throwing what was
2:39 that
2:40 just throwing darts at it seriously
2:42 that's kind of how it is
2:44 get you one of those external hhds
2:49 uh but here's my problem with external
2:51 hhds is I've got too many of them and I
2:54 like I'm good with them for about a
2:57 month or two and then I kind of forgot I
2:59 forget to hook them up or I forget to
3:00 take them with me and then they just end
3:02 up in a corner and then I'm like wait
3:04 what was on that one is that the one
3:05 that had the photo library or was it the
3:07 other one
3:08 I'm just waiting for at some point
3:10 they'll come up with infinite storage
3:12 you know for a nickel and I'll just take
3:15 all my drives and just dump it somewhere
3:16 and then GPT it so I can just find stuff
3:20 um
3:21 welcome everybody let's see if we can
3:23 get some folks in here um share the live
3:26 if you'd be so kind and
3:30 that would be Swale
3:35 we'll get some people in here
3:39 all right the hair is looking special
3:42 tonight it ain't good well it's it's
3:44 sort of sort of behaving
3:47 um Winston how are you sir
3:50 welcome
3:53 shared alive with five more friends
3:54 thank you Stevo
3:57 what did you think of the hey gen
3:59 webinar
4:00 um
4:03 my big takeaway from the hey gen webinar
4:06 was they don't have any [ __ ] clue
4:09 what they have on their hands they have
4:12 created a [ __ ] beast of a product and
4:15 he basically said this was a side gig of
4:18 one of our engineers
4:19 laughs
4:21 and we just kind of put it out there and
4:23 because they didn't get any traction on
4:24 it for three three weeks or whatever
4:26 because no one could find it
4:29 um
4:31 they assumed it was kind of a dud and
4:34 then when people actually discovered it
4:35 they're like holy [ __ ] that guy has no
4:39 idea this is the product of his company
4:42 the the the synthesizing voice and
4:44 making avatars talk all that [ __ ]
4:46 there's those companies are a dime a
4:48 dozen right now
4:50 the the video translation I mean like
4:56 I I was I was a little gobsmacked
4:59 how much of an afterthought the product
5:01 is how good it is and how the co-founder
5:05 of the company has no [ __ ] idea what
5:07 he has on his hands they had a backlog
5:10 of 570
5:12 000 videos or something like that or 51
5:14 000 I was some crazy amount
5:17 um my son is attending college for BS
5:20 and cyber security excellent choice tell
5:22 him to move straight to Vegas and go
5:23 work for MGM
5:27 Marla does he I assume he knows about
5:30 the MGM hack they've they've been in a
5:32 three-day a three-day Cyber attack where
5:35 they've got him they must have gotten
5:37 root level to everything because all the
5:39 games are down the hotel rooms that have
5:42 like electric blinds and locking doors
5:45 nothing's working
5:47 they can't check anyone in they can't
5:48 process any money
5:50 it's it's a bad one
5:53 so yeah I think cyber security is an
5:55 excellent choice because you know and
5:58 and you know your son should be aware of
6:00 things like worm GPT that's relatively
6:03 new on the scene which is an autonomous
6:05 agent built on a open source large
6:07 language model designed to sell to
6:09 hackers for 100 bucks a month that they
6:13 can just make autonomous agent go out
6:15 there and do stuff
6:17 um let me get started here welcome
6:18 everyone my name is Kyle Shannon this is
6:19 the AI learning lab that is the AI
6:21 learning lab we also have
6:24 yeah Ally Ally the cute little AI
6:27 learning lab huh how cute is that
6:31 um that image by the way was made with a
6:34 thing called ideogram and it can do text
6:37 it can do images it can do text the
6:40 other thing
6:41 that we'll talk about here is if you're
6:44 new to generative AI if you've not
6:46 played with it I just gave a talk
6:47 tonight at the uh Colorado University
6:51 school for uh for Public Policy or I
6:55 forget the exact name of it basically
6:57 democracy and public policy School
7:00 champ
7:11 they're not going to get us we're good
7:13 thank you though thank you thank you for
7:16 protecting us
7:18 uh
7:21 but it was fascinating I mean it you
7:23 know it's a school they haven't really
7:24 taken a position on AI yet which I find
7:26 astounding
7:29 um
7:30 and you know people in the room were
7:32 some combination of like been in it for
7:35 a decade and are cynical about it to
7:38 haven't really played with it don't know
7:40 anything about it and everyone in
7:42 between and it was interesting anyway if
7:44 you have not played with chat GPT go
7:46 there right now chat.openai.com
7:49 and that's the official chat gbt website
7:51 just go there and start playing just do
7:53 it now public service announcement I
7:55 don't make anything of it I don't give a
7:56 [ __ ] if you pay them or not but I would
7:58 I do uh I think it's worth paying for
8:02 and then uh and then uh if you don't
8:05 know what to say there if you're like
8:06 completely new to this world of
8:08 generative AI prompts.chat that little
8:10 document will will it's like a very
8:13 basic document with a bunch of prompts
8:15 in it but some explanation of how these
8:17 things work
8:18 so that should help
8:20 um the way the channel Works pop any
8:22 comments or questions down below I will
8:24 do my best to answer them and then uh
8:27 yeah we'll get going we'll get going
8:30 we'll have some fun we'll we'll do some
8:31 we'll do some wacky hijinks I like wacky
8:34 High Jinks I don't know about you but I
8:37 am a fan of the wacky hijinks so we'll
8:39 do some wacky hijinks and talk about AI
8:41 might get a little philosophical get you
8:43 know all Jordan Peterson and then we'll
8:46 get all I don't know Jed clamp it
8:49 this is the Jed Clampett voice remember
8:51 Jed Clampett Beverly Hillbillies no
8:54 anyone
8:57 I know all you Irregulars thank you seen
9:01 that that's not champ that's cyber
9:03 security exactly you rang uh
9:08 uh let's see the wife wasn't too happy
9:10 oh wait what is going on
9:13 about the very early get up for the uh
9:16 for the hey gen thing yeah that was a
9:18 little early I suppose
9:20 Kyle sir I have a question yes Ron what
9:23 is it
9:24 um my toilet is Flushing on its own how
9:27 do I fix that all right hang on
9:30 we can solve this
9:41 all right let's see how are we gonna do
9:45 it so if you haven't seen chat gbt
9:47 before this is it it's an unassuming
9:49 little beast isn't it
9:51 um
9:56 you are a master plumber
9:58 and understand the intricacies
10:01 of the modern
10:05 Plumbing lexicon period
10:09 I have a toilet that's flushing on its
10:11 own period
10:13 can you help me diagnose
10:17 what's going on and how to fix it
10:21 [Laughter]
10:23 I can certainly help diagnose and
10:26 potentially fix the issue
10:29 flapper valve seal well I could have
10:31 told you this Ron the most common cause
10:33 of a toilet flushing on its own is a
10:35 faulty flapper valve seal the flapper is
10:37 the rubber or the piece on the bottom of
10:39 the tank that covers the flush valve if
10:41 it's not sealing property the water can
10:43 slowly lean from the tank to the Bowl
10:45 causing the toilet to constantly refill
10:47 and flush to fix this issue you're going
10:49 to try adjusting or replacing the
10:51 flapper valve
10:53 C solution turn off the water supply to
10:56 the toilet flush it to drain most of the
10:58 water out of the tank inspect it see if
11:00 it's worn or damaged replace it ensure
11:03 that the new flapper creates a nice
11:04 tight seal when closed
11:07 your overflow tube it doesn't sound like
11:09 an overflow tube to me though
11:12 and the fill valve yeah that's generally
11:15 if it just keeps running the Overflow
11:17 tube and the fill valve or those things
11:18 water pressure issues doesn't sound like
11:20 that
11:21 leaking flush handle
11:23 nah it doesn't sound like that foreign
11:26 objects or debris do you have poop in
11:28 your toilet maybe maybe you've got some
11:30 poop in your toilet maybe that's the
11:31 problem
11:33 step number one take all the poop out of
11:36 your toilet step number two don't put
11:37 any more poop back in there
11:39 I'm going with flapper valve I'm going
11:43 flapper valve
11:45 welcome to the AI learning lab and
11:47 plumbing plumbing advice simple or
11:50 Emporium
11:56 as jokey and silly as that was I mean
11:59 I've been using Chachi PT to diagnose
12:02 car issues and
12:04 um all sorts of [ __ ] home issues
12:06 electric issues
12:10 um so yeah yeah you can do that you can
12:13 do it you too can make money with chat
12:17 JP day
12:20 all right welcome
12:24 you don't have to summon Jed clamp up
12:26 with jazz PPT here's this yeah now we
12:30 got a story about a man named ahead
12:39 and now the ground came some bubbling
12:42 crude oil that is
12:45 Texas Tea
12:50 oh yeah
12:52 [Laughter]
12:54 all right flap on flap off the flapper
12:58 that's good that's good that's good 90s
13:01 that's good 90s uh commercial PTSD
13:04 everything covered here exactly
13:09 oh by the way the shirt tonight League
13:11 of Rock
13:20 I'm about as qualified for plumbing
13:22 advice as I am for AI advice but I give
13:25 it anyway mansplaining is a Surface
13:27 you're welcome
13:31 me and chat GPT we got you covered
13:34 um hello everyone hey Nancy with shaking
13:36 uh um you like the ants you like the Ann
13:40 Sullivan of Plumbing via chat GPT a
13:43 perfect solution AI is going to help
13:46 customer service reps everywhere I'll
13:48 tell you what with the when you start
13:51 having conversations with pie if you
13:53 haven't done that yet shame on you
13:55 especially if you've been here before
13:58 Go download the PI app Pi data AI is the
14:01 website but they also have an app I
14:04 think it's on both Android and iOS Pi
14:07 personal assistant go find it it's from
14:10 it's from inflection
14:12 and
14:13 start talking with that thing and what
14:16 you realize pretty quickly when you talk
14:18 to Pi
14:20 is that
14:22 that's how all customer service should
14:25 be moving forward I would
14:27 by ten thousand by a ten thousand time
14:32 uh
14:34 ten thousand votes to one
14:37 I would rather talk to a synthesized bot
14:41 that's tied to a database that that
14:43 knows all the [ __ ] that I'm gonna ask
14:45 about then talk to a human being that
14:47 hasn't been trained up well enough and
14:48 can't do anything it doesn't have any
14:50 authority to do it so I think customer
14:53 service is going to get
14:55 dramatically better
14:57 and what's here's here's the here's the
15:00 weird thing
15:01 companies are gonna are gonna start
15:03 changing out their call centers for AI
15:06 Bots because it's cheap it's going to be
15:09 cheaper
15:10 and
15:12 we as customers are going to get
15:14 dramatically better customer service I
15:16 think customer satisfaction is going to
15:17 go up for big companies when they make
15:19 this switch
15:21 because what the Smart Companies will do
15:23 is they'll say
15:25 um
15:26 you know people like talking to the Bots
15:28 Better Than People anyway but when the
15:30 Bots hit a snag and can't actually solve
15:33 a problem they'll escalate it to a human
15:35 that's actually trained
15:37 and actually has the authority to do
15:39 something you'll get you'll get
15:40 appropriate escalation
15:43 just to build man I don't know how to
15:45 build that I the scaling of that is
15:47 something crazy and the data ingestion
15:49 and management of that is something
15:51 crazy but I'm telling you the the
15:53 company or the person that makes the it
15:56 just picks up the phone and starts
15:58 talking to you and it knows what the
16:00 [ __ ] it's talking about they are gonna
16:02 print money they are gonna make a
16:05 fortune
16:09 uh flip and Fortune all right Firefly
16:13 great to be here great to have you here
16:14 Firefly welcome
16:16 welcome welcome somebody sent me a
16:19 little crown that was a lovely little
16:20 Crown I got the little pink Rosy Cheeks
16:22 but I was over in that box
16:26 um
16:27 wow I'm wired on coffee and obviously
16:31 easily impressed with myself did 789
16:35 likes my new best
16:40 oh my God we're up to almost 5 000 likes
16:43 already Winston you you were a
16:45 significant percentage of that I don't
16:47 know about you either you're regulars
16:49 Winston's almost up to a thousand likes
16:51 and we're only at 5 000 total
16:53 what are you napping
16:56 I mean sure it's what is it 10 o'clock
16:58 nine o'clock 8 45
17:01 but that's 10 45 on the East Coast I
17:03 know some of you weirdos are on the east
17:05 coast and then we got freaks in South
17:07 Africa Australia they they're coming in
17:09 from all over the world
17:11 they're setting their alarms why are
17:14 they coming here why why I don't know I
17:16 think it's for the competition I don't
17:18 think it's for me I haven't bought a
17:20 conduit for the Irregulars entertainment
17:22 of themselves you all know that's true
17:25 all right
17:30 Winston's totally a show-off but it is a
17:33 competition it is life is a competition
17:36 life is a cavity old friend
17:43 did you know that did you know you can
17:45 make money
17:47 um
17:47 that she didn't know that that's y'all
17:50 pie was wrong about some stuff and when
17:52 I corrected her she apologized and said
17:55 what did she say hey hang on
17:58 my bad yeah I know my favorite thing
18:01 about pie is when you say something and
18:04 you kind of correct her and if you do it
18:05 in a smarmy kind of way she goes
18:09 she can't she can't laugh right so she
18:12 tries to to like speak out h-e-h-e and
18:15 she goes
18:19 it's very disturbing
18:23 hahaha
18:24 oh man oh by the way Emilio's wife I'm
18:27 gonna we're gonna do some experiments
18:29 here let me just answer a few questions
18:30 we're gonna do an experiment with your
18:31 data set
18:33 so because it's something I've got open
18:36 enrollment I've got like seven days to
18:38 do my open enrollment luckily I didn't
18:39 get a PDF like you just showed me
18:41 because oh my God it's awful
18:44 um
18:45 all right let me let me see if there's
18:47 any other questions here it's been a lot
18:49 of hours today on pie does greatly
18:51 improve your prompts yep hey everyone
18:53 farmer here also oh hey oh she's she's
18:56 back welcome back Hey sir I wanted to
18:58 build an OCR trainer for normal and good
19:02 purposes
19:04 can we Workshop prompts
19:07 an OCR trainer
19:11 I don't know what an OCR trainer is
19:12 explain more what that is like you want
19:14 to use Bing or Bard to read
19:18 um like images that have text in them
19:19 and then OCR the text out of them
19:23 we can we can try that maybe we'll do a
19:25 little experimenting tonight because I'm
19:27 trying I'm trying a thing with um
19:30 with Emilio wife's Insurance choices
19:35 you'll see this this spreadsheet's a
19:37 goddamn nightmare
19:39 um and we'll we're gonna see if we can
19:41 get it to work so we might as well do
19:43 some OCR stuff in there too
19:46 Gemini nears release small group of
19:49 companies given Early Access Ted looks
19:51 at life holy crap okay that's a big deal
19:53 that's a big deal that's a big deal I'm
19:56 a level 17 irregular total I like totals
20:00 level 17 Clinton level 17. I don't know
20:03 all you level 12ers are you like are you
20:05 content with your lives
20:08 I mean I hope you're feeling like I hope
20:10 a left the level of Shame you must feel
20:12 at a level 12 knowing that Clemson is
20:15 17.
20:16 and proclaiming himself a total
20:21 so sad
20:22 so sad
20:25 um
20:27 Winston is a show off now the
20:29 irregular's starting to they're starting
20:31 to Feud we're gonna have a little like
20:32 irregular Civil War in here thank you
20:34 Stevo
20:37 oh man how does one level up I have no
20:40 [ __ ] idea I just every now and then
20:42 it says so and so reach level 17. I
20:45 think
20:45 diamonds are I don't I listen
20:49 as as the young people in here say oh
20:52 isn't that cute look at the elderly
20:53 teaching each other how to use AI so oh
20:57 that's lovely Winston thank you very
20:59 much for that that was a good one oh
21:02 Winston level 17 boom take that clip
21:08 you can't just sit up on the top of the
21:10 mountain like proclaiming Victory not in
21:13 this house
21:14 all right let's let's go do some let's
21:15 go play
21:17 uh if you if you have questions feel
21:19 free to pop them down below uh I'm gonna
21:21 show you
21:24 this thing
21:27 so this is I assume there's no there's
21:29 no like this is just a generic PDF right
21:32 so okay
21:34 so
21:36 look at this horrific [ __ ] document
21:38 remember when I said I think one of the
21:39 killer apps of AI is that it's going to
21:42 take the ridiculously complicated and
21:46 um simplify it so there's your Kaiser
21:48 Permanente plan there's medical plan one
21:52 medical plan 2A medical plan to B
21:56 Medical Plan Three HSA optional and then
21:59 you have all these [ __ ] numbers and
22:01 none of us know what them they [ __ ]
22:02 mean and then oh and then there's more
22:05 and it goes on and then there's more and
22:08 it goes on
22:09 and now now there's Moda so here's a
22:13 here's another one and there's just like
22:15 this sea of [ __ ] numbers and none of
22:18 us knows what it means and you know what
22:20 I you know what this is like this is
22:22 like in the government you know how the
22:23 government has compartmentalization
22:25 where the the way they keep top secret
22:28 is no one knows the whole thing I bet
22:30 that's how insurance companies work
22:32 where there's one person that assembles
22:34 this spreadsheet but they don't know
22:36 what any of the numbers mean and then
22:38 there's whole divisions that are just
22:40 responsible for a column or a row or a
22:42 cell
22:44 all right so she's got this so
22:47 so she's got seven days to pick from
22:50 from all this [ __ ]
22:51 [Laughter]
22:54 [ __ ] Healthcare in America those of
22:56 you not in America this is what we
22:59 [ __ ] deal with so what ends up
23:00 happening is this so our company spends
23:04 I don't know we like we we pay a lot of
23:07 our of our employees health health
23:09 insurance it's it's some ridiculous
23:12 amount of money we're a very small
23:13 company and we're paying I don't know
23:15 150 or 200 Grand a year and then the
23:18 employees pay a bucket on top of that
23:20 and then and then when you go to the
23:23 doctor they're like oh no I'm sorry
23:24 that's not covered that's you're gonna
23:26 have to pay that out of one of these
23:28 whatever you pay comes out of one of
23:30 these numbers they don't tell you which
23:31 one and like oh I'm sorry that's out of
23:33 network or it just anyway it it's
23:36 criminal so that document
23:39 we're gonna go now try to do something
23:41 with it with chat JPay Tay and we're
23:44 gonna do we're gonna do chat chipity
23:48 I think I think we'll do both things in
23:50 chat gbt so wait uh oh can only select
23:53 from Moda no Kaiser okay great good
23:56 piece of information to know
23:58 so we're going to start with
24:01 chat with your PDF and then I heard a
24:05 hack on this okay so
24:09 so for for those of you that don't know
24:13 chat GPT when you pay the 20 bucks a
24:16 month you get access to gpt4
24:20 you get access to this thing called
24:21 Advanced Data analysis which is really
24:24 insane and we might play with that and
24:26 then you get an you get access to the
24:28 thing called plugins
24:30 within plugins
24:36 um you can install lots of them but in
24:38 any given chat session you can only
24:40 activate up to three and I'm gonna just
24:44 pick one right now
24:47 there is a plug-in called
24:51 chat chat with PDF ask your PDF I think
24:56 it's ask your PDF
24:57 I think it's ask your PDF
25:00 so we're going to turn that on we're
25:01 going to turn web pilot off now I heard
25:03 a hack okay so here's the hack
25:06 someone told me
25:08 oh I might have to create an account but
25:10 that's okay
25:11 for Emilio's wife we'll do it
25:15 we're gonna go to ask
25:20 your
25:26 pdf.com
25:32 upload document okay
25:36 so we're going to upload this document
25:38 healthcareplans.pdf
25:42 analyzing document all right
25:45 so far so good so so here's what I heard
25:49 so I can
25:54 I don't know if I could well if I've got
25:56 a PDF that's on the web I can just put
25:59 that URL in
26:01 um
26:02 in
26:06 chat GPT with the ask your PDF plugin
26:09 and it'll analyze it but what I heard is
26:11 better is that you upload the thing
26:14 to ask your PDF
26:18 and then there's supposed to be a number
26:20 here
26:24 if anyone knows how to do this let me
26:26 know
26:27 resume conversation here we go ID Okay
26:31 so
26:33 I might ask your PDF let me make this
26:35 bigger
26:39 the summary states there are no lifetime
26:41 maximum maximums on any of the plans
26:43 okay so so this is already started so so
26:46 I've uploaded this healthcare plan
26:48 document it's got this ID right here so
26:51 I'm going to copy that
26:53 and then we're going to go over to
26:54 chatgpt and apparently this so so
26:59 ask your PDF has already done some
27:01 pre-processing on this PDF it clearly
27:03 kind of understood what it was but I'm
27:05 going to say
27:07 um
27:07 do you do
27:11 you understand
27:14 what's in
27:17 the
27:20 PDF with this number
27:24 and then I'm going to paste that big
27:26 long number
27:28 and let's see what it says it should
27:30 fire off the CH ask your PDF yep it
27:32 fired off the ask your PDF plugin and it
27:35 should basically just give us what it
27:36 had over there like yeah this is a
27:39 a thing
27:41 the PDF
27:44 appears to be a document summary
27:46 benefits planned for the 2023-2024 year
27:53 okay so it understands what's in the PDF
27:57 the document provides details on various
27:59 Medical Services
28:02 see so so what you're gonna have to do
28:05 Emilio's wife is I go through here I
28:07 don't quite know what to ask it so so
28:09 you you might need well well actually so
28:12 what I'll what I'll do is we'll have
28:15 chat GPT analyze the plans so what were
28:18 the plans you said was it called Moda
28:24 um
28:26 I can only choose from the Moda plans
28:33 Moda right Moda yeah I can only choose
28:37 from the Moda
28:39 plans
28:41 and there are so many numbers and
28:43 options that
28:48 it makes my head spin period
28:52 can you
28:56 describe the Moda options
29:00 and give me a balance between monthly
29:03 premium
29:04 and
29:09 deductibles and out-of-pocket expenses
29:13 Etc
29:14 without giving me too much detail
29:21 period
29:24 ideally what I want is a recommendation
29:27 from you that strikes the right balance
29:29 between these two period
29:38 some
29:39 marketers
29:42 show you which of their options is the
29:46 best deal
29:47 period can you do that for me with the
29:50 Moda plans
29:53 now that may be way too much of a prompt
29:55 but like what I guess what I'm trying to
29:59 do here is I'm I'm actually trying to
30:01 imagine so imagine a a version of this
30:04 in the future where we can talk to these
30:06 things
30:07 and because this is this is kind of what
30:10 I would say to a uh to an insurance
30:14 agent so let's see let's see what it
30:16 does
30:18 so it's going to ask your PDF
30:26 certainly let's break down the Moda
30:28 Health Plans based on
30:37 generic 12 bucks for 31 day supply 36
30:41 bucks for 90 day supply so there's the
30:44 pharmacy benefits
30:47 wait Moto plans one through four oh I
30:51 see plans one through four have all the
30:53 same Pharmacy benefits and then five
30:56 through seven have all the same Pharmacy
30:57 benefits
30:59 General notes the individual deductible
31:06 this but this is just for drugs
31:11 plans one through four seem to offer a
31:13 balanced approach
31:18 all right so
31:20 see plans one through four seem to offer
31:22 a balanced approach with reasonable
31:24 costs for generic drugs and preferred
31:26 Brands however the exact monthly premium
31:28 deductible are not mentioned is that
31:31 true I don't know that that's true
31:33 hang on
31:36 where
31:40 is this it
31:43 let me go down
31:45 Moda plans
31:47 out of pocket now it's got the out of
31:50 pocket here
31:52 okay so it hallucinated all right this
31:54 is good to know so hallucinated
31:55 basically told us I couldn't see all the
31:57 [ __ ]
31:58 so let's go yell at it
32:01 um
32:05 I'm looking at the document and the out
32:07 of
32:09 pocket and monthly premiums are
32:11 mentioned
32:13 for the Moda plans
32:17 ammo
32:19 D.A
32:21 plans
32:24 you only showed me the pharmacy
32:28 section
32:30 please look at the entire PDF
32:33 and make sure that you
32:36 understand the entire plan before
32:40 speaking to me
32:45 give that guy give it a little dressing
32:46 down even hey I can't figure it out
32:49 exactly this is all by Design you know
32:58 oh look it has it it fired off the thing
33:00 twice what did it what did it find
33:03 so the first time
33:06 it got back the pharmacy stuff I
33:08 apologize for the oversight
33:13 and this looks like it's got deductibles
33:15 and stuff okay so it basically did two
33:17 different things here we go
33:20 office and virtual care 20 to 30 bucks
33:22 wait for let's break down the Moda plans
33:25 based on the document
33:33 uh
33:56 foreign
33:59 let's break so now it looks like it it
34:02 got us
34:07 some amount of the information but
34:13 okay it it doesn't tell us which plan
34:16 this is and it doesn't tell us the cost
34:19 um
34:25 I'm very disappointed period
34:31 you don't seem to understand
34:35 the document
34:37 period
34:40 there are seven different Moda plans
34:44 you only appear to have given me
34:46 information about one of them and you
34:49 still haven't given me information
34:51 about
34:54 deductibles and out of pocket Etc please
35:02 try to
35:04 look at the entire document
35:07 and give me my options
35:10 we may have to bail on this and go try
35:12 code interpreter
35:19 we can also try I don't know if Claude
35:22 can read PDFs but we can go try Claude I
35:25 haven't really played with it
35:28 I apologize for the oversight let's
35:31 delve deeper Moto plans one through four
35:34 so wait so this says the cost of
35:38 Primary Care office for plans one
35:41 through four is 30 bucks let's go look
35:44 at the actual document
35:47 oh this is [ __ ] maddening
35:51 medical plan one
35:54 primary visits 20 bucks 20 bucks medical
35:58 plan 3 is 25 4 is 25.
36:03 so so it's just wrong so those are all
36:05 wrong okay
36:07 so so this is good we're doing this
36:11 primary care costs
36:15 and incentive care visits 25 wait let me
36:20 go
36:24 primary care now it got it got these
36:27 wrong these these are just wrong
36:30 outpatient no it just got them wrong
36:33 okay
36:34 all right fine uh we're gonna tell chat
36:37 GPT to take a hike although no you know
36:40 what we're gonna do we're gonna go
36:43 what evil mind created that chart I know
36:45 it's it's really bad okay we're gonna go
36:48 try uh Advanced Data analysis let's see
36:51 if this can do PDFs
36:53 um
36:54 DD
36:57 there's more than one way to skin a kid
37:00 we're gonna skin this kit Healthcare
37:03 plans.pdf
37:05 can you Analyze This PDF and tell me
37:08 what it is
37:10 let's just see if it can do that
37:15 Share Bear join welcome share bear
37:17 what's happening
37:19 I'll start by extracting the text from
37:21 the PDF file once I have the text
37:23 content I'll analyze it to determine its
37:25 content and provide a summary good thank
37:28 you share bear loves me some gummies I
37:30 love them gum this gum is tasty
37:33 the extracted content from the PDF
37:35 appears to be related to health care
37:37 plans here's a summary of what I've
37:38 extracted
37:40 the name Kaiser Permanente network is
37:42 repeated multiple times there are
37:45 different plans mentioned
37:47 fine
37:48 would you like a more detailed analysis
37:50 or any other specifications sections
37:53 you'd like to one of the companies is
37:57 moda
37:58 and I'd
38:01 like you to explain
38:05 their plans to there wait hang on there
38:11 plans
38:12 dang it
38:14 plans to me
38:16 as if I were a third grader period
38:21 these plans are purposefully convoluted
38:26 and I'm just trying to make a simple
38:27 choice
38:29 I'm hoping you can help
38:33 I am in deep
38:35 agonizing
38:37 frustratingly
38:40 saucy
38:42 pain
38:44 as a result of trying to understand
38:47 this purposefully
38:49 obfuscated information
38:53 about
38:54 my health as a human being period
38:59 please make the pain stop
39:02 [Music]
39:02 [Laughter]
39:09 let's break it down imagine you have a
39:13 toy store membership card depending on
39:16 which one you want to buy okay
39:18 deductible this is like the entrance fee
39:21 you pay this amount before your
39:22 membership cards start helping you get
39:24 the discount so if you have a hundred
39:26 dollar deductible you pay the first
39:27 hundred dollars on your own out of
39:29 pocket maximum think of this as the
39:31 safety net
39:33 oh my God in network versus out of
39:35 network blah blah
39:38 okay from the document the Moda plans
39:40 seem to offer
39:42 you pay a certain amount 25 30 bucks and
39:45 as your membership card so it's
39:47 continuing to explain it like we're in
39:49 third grade which is cool
39:51 virtual care if you talk to a doctor
39:53 online it's like a virtual toy some
39:56 plans let you do this for free While
39:58 others might ask you for a bit
40:00 specialty Visions okay
40:04 Okay so
40:05 [Music]
40:08 that helps explain things
40:12 generally but I'm hoping that you can
40:15 actually
40:17 look at the specific data for each plan
40:21 give me an overview
40:25 and explain to me
40:30 the pros and cons of each plan period
40:34 at the end of explaining the pros and
40:37 cons
40:38 I'd like you to recommend two plans that
40:42 you feel strike the greatest balance
40:44 between
40:45 small monthly
40:48 premiums and decent coverage
40:55 all right
40:57 let's see
41:03 Absolutely I'll go through the details
41:06 of the mode of plans
41:08 and I'll highlight the pros and cons I I
41:10 have my gut is telling me that code
41:13 interpreter or Advanced Data analysis is
41:15 going to be able to do this better
41:17 because it because it understands what
41:18 what csvs are
41:20 and assuming that it it maintained the
41:23 column structure of the PDF it should be
41:25 able to dump that so it says it seems I
41:27 made an oversight let me correct that
41:30 and extract the details for the Moto
41:32 plan so what did it do here
41:36 Moda compile
41:39 service Clean Service strip so it looks
41:42 like it's stripping out stuff
41:45 it seems I'm having difficulty
41:57 however I can still work with the
41:59 details we have and provide some okay so
42:01 it just choked
42:03 networker
42:05 um let's see
42:07 let me stop generating
42:11 rather than trying to analyze right now
42:14 can you take
42:17 the Moda data
42:24 foreign
42:25 and put it in a data table so that you
42:29 can
42:31 manipulate it
42:33 like any other CSV
42:43 message and conversation not found okay
42:45 so we're going to start over all right
42:46 hang on starting over
42:48 uh that's another thing sometimes these
42:50 things are just they kind of crap the
42:52 bed and you can't do anything so I'm
42:53 gonna say
42:56 this is a PDF of health care plans
42:59 I'd like you
43:02 to
43:06 collect the information for the Moda
43:09 plans
43:15 Moda is a hard word for OCR or for voice
43:18 to text
43:22 and put them in a data frame so that you
43:25 can interact with them just like any
43:27 other CSV period is that possible
43:34 ah I we are we having fun yet we're
43:37 having fun right we're having fun we're
43:38 having fun by the way for those of you
43:39 that are new here so so what we're
43:41 trying to do is we're trying to do an
43:42 actual real world application
43:45 um one of our lovely Irregulars Emilio's
43:47 wife has this horrific
43:51 set of options which I'm sure we all do
43:53 right now it's open enrollment season
43:55 and here's all the [ __ ] crap you
43:59 gotta sift through and figure out isn't
44:01 it clear all of this look just look at
44:04 these numbers and pick one of these 17
44:06 plans and we'll still screw you and take
44:09 too much money out of your check and
44:11 when you go to get something done we'll
44:13 refuse to pay it yay in Insurance in
44:17 America the best yay I hope someone in
44:20 here works for one of these God forsaken
44:21 companies
44:24 you should
44:26 hurt at night
44:31 I don't mind most corporations but
44:33 health insurance companies
44:39 bad all right so
44:43 where are we where were we let's see
44:46 um of course I'll start back so it
44:48 appears that it's been successfully
44:50 extracted but the formatting might be
44:52 messy
44:53 I can see references to various plans if
44:56 you proceed it seems I couldn't find any
44:58 sections labeled Moda plan
45:01 oh that's because it's an image
45:05 uh this is painful
45:08 this is [ __ ] painful
45:12 so so they have a logo here but they
45:15 nowhere do they have Moda they've got
45:17 medical plan one Connexus Network
45:21 all right so I assume that's different
45:23 than Kaiser right
45:25 yeah Kaiser Permanente Network okay okay
45:28 so
45:31 Connexus Networks
45:33 c-o-n-n-e-x-u-s so so
45:36 notice what we're having to do here
45:38 we're having to essentially think like
45:40 programmers
45:42 to even be able to ask chat GPT
45:45 how to look at this document now if we
45:48 had this as a CSV file which I'm sure we
45:51 could probably get it as one it it would
45:53 probably be easier
45:55 but
45:57 this is where just getting really
46:00 curious with these tools is really
46:02 important so so what we're going to say
46:04 is this
46:05 um
46:06 the Moda plans are the ones that are
46:09 connected to the Connexus
46:13 Network period can you pull the data
46:17 related to that Network
46:20 and put it in a day but
46:25 data table for us
46:30 no not save the table data table come on
46:34 you dumb dumb
46:38 sure I'll look for the sections related
46:41 to the Connexus Network
46:46 [Music]
46:47 what are we gonna name this collection
46:50 of doctors that refuse to do any work uh
46:53 let's call it Connexus it sounds like
46:56 connection and Lexus because I want a
46:59 new Lexus
47:01 I would like the Lexus 450 XL gold
47:05 platinum diamond reserve the 2024 model
47:09 so Connexus will get me my new car I
47:14 apologize for the inconvenience it
47:15 appears that extracting text from the
47:17 PDF might be more challenging let me
47:19 take a different approach
47:21 so now I'm starting to get as mad at
47:22 Chachi BT as I am at the insurance
47:24 companies
47:27 what did he do did it just stop
47:30 wait let's take a different approach if
47:32 you can provide a deep brief description
47:34 of the layout of the document
47:36 oh God
47:39 now I want you to be magic
47:43 is this text yeah plans one to four
47:48 plans one through four
47:51 plans continued
47:53 okay so we can say
48:04 sure the Moda plans have bold text at
48:08 the top of the pages that say plans 1-4
48:14 and another set of plans marked
48:18 plans 5-7
48:23 please gather the information from those
48:27 seven plans
48:28 so that we can start making a decision
48:33 all right let's see if we can get
48:35 something here
48:36 I'll give this one more shot we'll see
48:38 we'll see if we can get it to at least
48:40 grab the data and if not I'm gonna I'm
48:42 gonna pop it over into Claude because I
48:47 think Claude it'll just be interesting
48:49 to see can Claude do this better than
48:51 than chat gbt I doubt it but you never
48:53 know
48:54 and then I don't think I don't think we
48:57 can upload this to Bard
48:59 I don't think we can upload this to Bing
49:03 because both of both of those can read
49:04 images
49:06 so it says it still can't
49:08 they're still not able to get it
49:11 um there's one other way to maybe skin
49:13 this cat hang on
49:16 let's see
49:19 uh I think this will work
49:32 copy
49:35 foreign
49:42 [Laughter]
49:50 it's like Madden except not fun
49:54 that didn't work did it nope
50:01 uh
50:04 delete let's go see if chat gbt figured
50:07 anything out
50:08 I guess I'm not
50:12 oh wait
50:14 great we managed to extract headers from
50:16 different plan sections
50:22 nope I forgot to import the pandas
50:25 Library
50:28 all right so now it's making its own
50:29 data frame so it looks like maybe it got
50:31 some some data for the data frames
50:36 it's gonna It's Gonna Fill In missing
50:39 values with not a number see they've got
50:41 they've got collapsed cells and stuff
50:43 like that so the thing that makes it
50:45 easier to read as a human makes it much
50:47 harder for the
50:49 for the computer to parse the data but
50:53 it's it's attempting to do its thing so
50:55 what you're looking at here if if you've
50:57 joined new this is code interpreter or
51:00 Advanced Data analysis within chat gbt
51:03 we uploaded a PDF of Health Plan
51:05 information and now we're just trying to
51:08 get it to understand the data in the PDF
51:11 you know as if it were in a CSV
51:14 and it's basically sort of coaching
51:17 itself through trying to make it work
51:19 and I have a feeling it's going to fail
51:21 miserably yeah
51:25 the additional cost to your column seem
51:27 to have fewer entries than the plan name
51:29 and network name columns that this
51:30 mismatch is what's causing me to have a
51:34 headache
51:37 chat EVT is like I need a [ __ ] drink
51:40 [Laughter]
51:46 um would you like me to proceed with any
51:48 specific analysis or manipulations of
51:50 the data
51:52 um
51:54 why yes I'd like you to help me
51:56 understand the specifics of each of the
51:59 seven plans and then recommend two of
52:02 them that you think strike the best
52:04 balance between cost and coverage
52:11 I have zero confidence that's gonna work
52:15 zero confidence zero confidence
52:23 coverage this would include benefits
52:25 medical services covered any extra
52:26 features Etc cost this generally
52:29 includes premiums deductibles
52:30 out-of-pocket maximums co-payments Etc
52:35 plan one we don't have the detailed
52:37 specifics to go behind no additional
52:40 specifics
52:42 no it's got nothing wait plan plan five
52:47 no no this is a fail fail me fail all
52:50 right done now we're gonna go try Claude
52:54 I I I I I I hate to be the bearer of bad
52:58 news Amelia's life I think you might
53:01 need to actually look at these stupid
53:03 documents uh but let's try Claude
53:08 uh
53:09 uh when AI couldn't save the day
53:11 actually
53:13 um
53:14 what what would potentially do it if if
53:20 Clyde can't do it is if you got if you
53:22 reached out to the to the HR people
53:26 or the insurance people it's probably
53:28 the HR people whoever's the managing
53:30 group dartboard strategy exactly
53:33 um reach out to the people that gave you
53:35 this PDF and see if they can get access
53:37 to the
53:39 to the CSV to the to the Excel files
53:42 from Moda
53:44 um that was used to put in that stupid
53:46 [ __ ] chart they gave you start a new
53:48 chat okay so we're gonna say
53:50 um
53:53 I let's see I
53:58 am allowed to
54:03 only
54:05 choose
54:09 Moda plans
54:12 in this Healthcare document there are
54:17 seven plans
54:25 from Moda
54:27 and I want to
54:34 understand
54:36 the pros and cons of each of them
54:40 period can you help me analyze
54:45 the details of each of the seven plans
54:50 all right let's see what Claude does
54:52 with this thing
54:54 the last time I tried a PDF with Claude
54:57 it did not like it at all
55:04 based on the healthcare plan document
55:05 here's a comparisons Moto plan one
55:07 Connexus Network that seems right
55:10 400 deductible
55:13 oh this is looking right hang on hold on
55:16 is Claude coming to our rescue
55:23 plan one
55:25 hundred dollar copay wait Moda plans
55:28 only
55:32 what did it say
55:42 400 individual 1500 family deductible is
55:47 that right
55:57 foreign
56:00 I can't read this document
56:03 100 copay where's the oh continue that's
56:07 why
56:08 okay plan one
56:10 four hundred dollar deductible 1500 okay
56:13 that looks right and then plan two was
56:15 800 2700 okay let's go let's go back to
56:18 cloudy cloudy Claude
56:26 in some remote plans
56:28 one to four have co-pays
56:30 and lower deductibles while five and
56:32 seven have coinsurance and higher
56:35 deductibles
56:36 lower numbered plans
56:42 have lower out-of-pocket Maxes Moda 6
56:45 has the lowest deductible and out of
56:47 pocket
56:48 oh my God like even this thing
56:51 explaining it simply is making my head
56:55 [ __ ] explode
56:57 um
57:00 listen comma I'm a dumb dumb and
57:05 even your simple explanation at the
57:07 bottom is confusing me
57:10 beyond my imagination period
57:14 I went to school period I read stuff
57:17 period I know how to use a calculator
57:20 period
57:22 why is this so hard to understand
57:23 question mark
57:26 can you please
57:28 explain to me
57:30 the difference between
57:33 the one through four plans and the five
57:36 through seven plans
57:38 even more simply
57:41 now was was all of that
57:45 that shame showing off necessary no it
57:50 makes me feel better I understand health
57:52 insurance can be confusing so let me try
57:54 explaining
57:56 plans one to four you have a co-pay for
57:59 a doctor the deductible and out of
58:01 pocket are lower
58:03 you start getting coverage right away
58:05 since many services have co-pays in the
58:08 deductible is lower
58:09 five to seven you pay the percentage of
58:12 costs coinsurance after meeting the
58:15 deductible for example you might pay 25
58:17 percent oh I get it
58:19 so this is actually helping okay
58:21 Emilio's wife so I don't know if you can
58:23 understand this [ __ ] so basically
58:26 the okay so okay so this is interesting
58:29 okay so now so so here's now here's what
58:32 I want to understand
58:33 okay I get it conceptually now can you
58:36 please
58:37 do the analysis
58:40 and help me determine
58:45 what is going to cost me
58:50 the least on an annual basis for the
58:54 most coverage
58:56 assuming
58:59 average
59:02 doctor visits for a Gen X couple
59:10 this is actually kind of cool I mean the
59:12 fact that the fact that Claude could
59:14 figure this out and chat EBT couldn't
59:16 this is a really good example of
59:20 um
59:21 the reason to kind of have so so one of
59:24 the things I you know I show all the
59:26 time is I've got these different tools
59:28 right I've got this is the adults table
59:30 the the the these are the biggies right
59:34 and then this is the Kids Table
59:37 and there's seven of them right there's
59:39 CH there's Chachi PT Claude bard
59:42 Bing perplexity I don't really use
59:45 that's just there to play with pi I
59:47 don't really use other than to talk to
59:48 it Poe I don't really use so Bing so
59:51 there's kind of four that are my go-to's
59:52 these these three plus Bing
59:55 and so this is a really good example
59:56 where chat GPT both with plugins and
59:59 code interpreter didn't do this well
1:00:03 oh look what it did here this is cool
1:00:05 this is really cool
1:00:08 okay let's run through an example
1:00:10 assumptions you and your spouse are in
1:00:12 your 40s Gen X ooh
1:00:14 this obviously doesn't know what Gen X
1:00:17 is
1:00:18 um you plan to get physical routines you
1:00:21 plan to take one generic prescription
1:00:23 regularly
1:00:24 um so so so let's see
1:00:27 um
1:00:29 let's up the ages to
1:00:31 late 50s and daily prescriptions to
1:00:35 three
1:00:38 I have no idea what's going on in your
1:00:40 life with me Leo's wife but I know in
1:00:42 mine that should get us in the
1:00:44 neighborhood
1:00:50 mm-hmm
1:00:53 okay here's a comparison how much does
1:00:55 Claude cost claude's free
1:00:58 um but they just announced Claude Pro
1:01:00 and so what I think they did was they
1:01:02 actually
1:01:04 they're actually reducing the amount of
1:01:08 uh interactions you can get
1:01:11 for the free version and then if you pay
1:01:14 the 20 bucks you can do more the the
1:01:16 real power thing about Claude is that it
1:01:19 can take big documents and apparently it
1:01:21 can read health insurance plans better
1:01:24 than chat gbt okay here's the cost
1:01:26 you're in your late 50s three generic
1:01:29 prescriptions a day two sick visits one
1:01:31 urgent care routine physical
1:01:34 so
1:01:35 so it this is interesting it kind of did
1:01:38 this like like a like a like a food
1:01:43 um
1:01:43 whatever you call it a food uh
1:01:48 like a Yelp right with the three dollar
1:01:50 signs so Moda plan one the premiums are
1:01:54 three dollar signs the co-pays are three
1:01:55 dollar signs the total is three dollar
1:01:57 signs Moda plan two the premiums are
1:02:00 lower
1:02:01 the co-pays are high and the overall is
1:02:04 high Plan Three is
1:02:07 small premiums all right let's see
1:02:13 so all of the totals are about the same
1:02:18 um
1:02:20 with increased prescriptions Moda plan
1:02:22 one Remains the lowest cost option oh so
1:02:26 this is weird
1:02:28 so it looks like you only gave me
1:02:33 dollar signs where actual numbers would
1:02:36 be period did you do that on purpose
1:02:40 question mark if not can you please give
1:02:44 me this list with actual numbers
1:02:48 from the plan
1:02:53 that's what I wanted a cost estimator
1:02:55 there you go okay cool
1:03:00 so we're getting there emilios this is
1:03:02 why you've come all these nights this
1:03:05 was all your plan this was your plan all
1:03:07 along show up for months every single
1:03:08 night so that I would help you figure
1:03:11 out
1:03:11 your plan you're absolutely right as an
1:03:15 AI assistant without access to specific
1:03:16 plan details okay if forgot it had the
1:03:19 plans okay so we're gonna re-upload the
1:03:21 plan
1:03:29 it looks like you may have
1:03:31 forgotten that I uploaded the plan
1:03:33 earlier period attached is the PDF that
1:03:38 has
1:03:39 a number of plans including Modas plans
1:03:43 one through seven
1:03:46 period
1:03:49 Moda
1:03:54 please give me
1:03:59 please give me a cost estimator like you
1:04:02 did before but now use the actual data
1:04:06 from this PDF
1:04:10 all right let's see
1:04:12 it finally paid off exactly
1:04:17 9 30 Denver time and 21 000 likes I mean
1:04:21 I'm just sitting here like banging my
1:04:23 head against the keyboard trying to get
1:04:25 these tools to behave in any way like
1:04:28 they can help us
1:04:31 but listen hey not for nothing it can
1:04:33 write you a country song about your
1:04:36 chubby dog or a turtle at the county
1:04:38 fair data can do really well
1:04:40 you're right my apologies now that
1:04:42 you've provided the health plans and
1:04:44 details okay assumptions so it did
1:04:46 remember so now it's still giving us the
1:04:48 weird ass
1:04:50 dollar signs only based on the plans
1:04:55 oh and I've only got three messages
1:04:57 remaining until 2 A.M so we get we get
1:05:00 three more shots to do this
1:05:03 again you didn't include any actual
1:05:06 numbers period please look at the PDF
1:05:10 and the actual numbers of the plans
1:05:12 before doing your cost estimators
1:05:19 so the I'm a librarian the the
1:05:23 um
1:05:23 what's it called rate limiting they call
1:05:26 it the rate limiting I'm hitting right
1:05:28 now is
1:05:29 um
1:05:30 it would be five times this if I were
1:05:33 paying the 20 bucks a month
1:05:42 ask it to do a cost analysis given
1:05:46 certain parameters okay we can try that
1:05:49 ah here we go
1:05:52 oh
1:05:54 this is actually pretty cool
1:06:07 so this is same assumptions Moda plan
1:06:09 one premiums 732 a month offensive
1:06:14 87 91 annually co-pays three grand total
1:06:18 11
1:06:19 000.
1:06:21 or 11 800 plan two less per month
1:06:29 3300 in copays but 11 000 total so
1:06:35 that's looking pretty good
1:06:45 all right so the overall annual it looks
1:06:48 like
1:06:49 it looks like plans one and two are the
1:06:51 most expensive by a lot
1:06:57 [Music]
1:06:57 um
1:07:03 co-insurance
1:07:05 forty three hundred uh
1:07:14 let's see
1:07:25 this is perfect can you now
1:07:31 put in some of the elements
1:07:34 of things like
1:07:38 prescription payments doctor visits Etc
1:07:41 so I can see what I get for my money
1:07:49 um
1:07:54 so so it seems to me Emilio's wife that
1:07:58 you you should take take this document
1:08:01 upload it to clutch seems like it can
1:08:03 actually do this
1:08:06 and and you know kind of more
1:08:08 specifically what you want to look at
1:08:09 like it can at least explain [ __ ] to you
1:08:11 let's see what it does with this next
1:08:12 I've got I've got two two uh prompts
1:08:15 left
1:08:18 you're right breaking out okay
1:08:25 oh I see what it was doing it it was
1:08:28 doing the the co-pays
1:08:33 oh that's interesting oh I see what it
1:08:36 was doing okay
1:08:37 so so it was sort of it was sort of
1:08:41 um in this in this analysis up here
1:08:43 here's your premiums you're paying and
1:08:45 then saying your co-pays are 3 700 I was
1:08:48 trying to figure out where that came
1:08:49 from and it it had six Primary Care
1:08:52 visits 240 bucks that's based on that
1:08:55 rate two Urgent Care visits I guess it's
1:08:57 40 bucks an Urgent Care visit two
1:08:59 specialist visits six prescription drugs
1:09:02 total copay
1:09:04 um so yeah this this could this could
1:09:07 probably give it to you let me know if
1:09:09 breaking out estimates for co-pays
1:09:11 um I think I've got one message left so
1:09:15 based on everything you see here
1:09:19 can you give me a recommendation of the
1:09:23 two plans that you feel represent the
1:09:26 best deal
1:09:28 for me and my spouse
1:09:38 [Music]
1:09:44 one of my meds is a hundred K and I need
1:09:47 it every six months unbelievable I know
1:09:50 I've got kids with chronic lyme disease
1:09:53 and [ __ ] nothing is covered nothing
1:09:56 the insurance companies basically go
1:09:59 after doctor's medical licenses if they
1:10:02 try to you know prescribe anything
1:10:04 outside of their horrific
1:10:07 oh it's uh it's so criminal it's so
1:10:10 criminal I'm so angry all the time about
1:10:12 this [ __ ]
1:10:13 um based on the plan details and cost
1:10:16 estimates my top two recommendations are
1:10:20 Moda plan one lower premiums at 8 700
1:10:25 annually co-pays for services mean you
1:10:28 pay a fixed dollar amount versus the
1:10:30 co-insurance percentage that's a good
1:10:33 point lower co-pays for office visits
1:10:36 your total estimated cost is 9 200 bucks
1:10:39 although it gave us a different
1:10:41 estimated cost before so so again some
1:10:44 of the math you're gonna have to sort of
1:10:46 check the math here
1:10:49 um
1:10:50 it's saying Moda plan 4 comes in at 6
1:10:53 500 that's not what it's set up here
1:10:55 before
1:10:56 said Moto plan for oh yeah it does
1:11:00 oh wait is that oh yeah
1:11:03 but wait is that what it's set up here
1:11:05 no see up here
1:11:09 it said it was 9600 and down here it's
1:11:12 saying 6500. so it's doing some
1:11:14 hallucinating here but
1:11:17 like I feel like I feel like at least
1:11:19 what you could do with Claude at least
1:11:21 here is have it give you sort of the the
1:11:24 like distill out you know the most
1:11:28 important pieces of each plan and then
1:11:30 you can sort of go back and forth and
1:11:32 check between what Claude did and what's
1:11:34 in that PDF and it'll just at least help
1:11:36 you make sense of it so anyway all right
1:11:40 Rabbit Hole done I'd pick plan five on
1:11:43 my own oh interesting
1:11:45 yeah what what did they say plan five
1:11:47 was plan five
1:11:49 yeah 9 500 so that's the one with
1:11:52 coinsurance so your risk there with
1:11:54 coinsurance is if you've got an
1:11:56 expensive office visit you're paying
1:11:59 higher than the flat the flat fee you'd
1:12:01 pay with plans one through four
1:12:03 all right well there you have it all
1:12:06 right I hope that helped uh
1:12:10 [Music]
1:12:10 [Laughter]
1:12:14 wow
1:12:17 wow
1:12:26 this is the [ __ ] that AI needs to fix
1:12:28 like like
1:12:30 let's assume that insurance companies
1:12:32 are gonna keep being [ __ ] and
1:12:33 they're going to keep doing this right
1:12:34 because you know they will until
1:12:37 someone else buys someone else a
1:12:38 politician
1:12:40 and but if AI can
1:12:45 take all that nonsense and just easily
1:12:48 with all this these complicated
1:12:50 decisions we have to make in our lives
1:12:52 just you know instead of us having to
1:12:55 struggle to figure out which tool to use
1:12:58 we've got some sort of agent that we can
1:13:00 just talk to and go go make [ __ ]
1:13:02 sense of this to me you know my
1:13:05 preferences you know go figure that [ __ ]
1:13:07 out and come back with two
1:13:08 recommendations and you know and they
1:13:11 don't hallucinate and they do actual
1:13:13 math and you know that's kind of the
1:13:15 three year out my you know three to five
1:13:17 years out
1:13:19 these things should be doing a lot of
1:13:21 this [ __ ] for us where
1:13:24 remember when we asked had to actually
1:13:25 look at PDFs with like big columns and
1:13:28 numbers in it and figure that [ __ ] out
1:13:30 that's what we'll be saying in five
1:13:31 years we ain't there now
1:13:36 oh my God what a nightmare that was all
1:13:39 right
1:13:40 let's get going here way to hijack alive
1:13:42 yeah Emilio's wife what are you thinking
1:13:44 here I bet I bet you were sitting there
1:13:46 the whole time just watching me do that
1:13:48 work and you weren't even tapping the
1:13:50 like screen you were letting all the
1:13:51 other other Irregulars get us up to 22
1:13:54 000 likes you're probably just sitting
1:13:55 there like oh that's curious I know how
1:13:57 this goes I got it that segment gave a
1:14:00 great power nap
1:14:03 all right welcome to the AI learning lab
1:14:06 let's get started tonight okay so if any
1:14:09 of you survived that
1:14:11 um
1:14:12 here's where I think you should go play
1:14:16 is
1:14:18 chat.openai.com you know what's funny
1:14:21 you know what's striking me all of a
1:14:23 sudden thank you share bear
1:14:26 um
1:14:27 for the past month or month and a half
1:14:31 a fair amount of people in here and a
1:14:33 fair amount of people and Tick Tock or
1:14:35 in in Twitter land and things like that
1:14:37 are saying that gpt4 is not programming
1:14:40 as well it's it's not as good
1:14:43 and then and then some people not all
1:14:45 but some people have been saying Claude
1:14:47 is better than gpt4
1:14:49 the fact that we couldn't get gpt4 to
1:14:52 [ __ ] figure out a relatively decently
1:14:55 structured PDF and Claude could figure
1:14:58 it out
1:14:59 something's going on man
1:15:01 open enrollment season at the AI
1:15:03 learning lab good Lord I know right oh
1:15:05 that was exhausting
1:15:08 um yeah that was weird that was weird so
1:15:11 again
1:15:13 this reinforces something that I've been
1:15:15 saying for a while that that what you
1:15:18 need to do with these tools is don't get
1:15:20 too precious with any of them
1:15:21 everything's changing all the time and
1:15:23 some of that [ __ ] that's changing is
1:15:24 behind the scenes like I don't know
1:15:26 what's going on with gpt4 but maybe
1:15:28 they're [ __ ] with the model to make
1:15:29 it I don't know less
1:15:32 um
1:15:33 less reliant on copywritten training
1:15:36 data or sorry I like I don't even think
1:15:38 that's possible but let's just say
1:15:39 they're doing something like that
1:15:41 um
1:15:43 the most important thing to do right now
1:15:45 is understand conceptually how these
1:15:48 tools work and and just use them enough
1:15:51 that you start to understand the
1:15:53 limitations of each one and and just be
1:15:55 able to bounce between them because
1:15:57 they're all basically the same [ __ ] it's
1:15:59 all basically a box
1:16:00 some of them are connected to the
1:16:02 internet some are not some you can
1:16:04 upload pictures some you can upload data
1:16:07 files
1:16:09 none of them are perfect all of them
1:16:11 hallucinate sometimes they're [ __ ]
1:16:13 brilliant to the point that you're like
1:16:15 holy [ __ ] and then other times like you
1:16:18 spent whatever we spent [ __ ] half an
1:16:20 hour with chat GPT or 20 minutes trying
1:16:23 to get it to just see the [ __ ]
1:16:25 document
1:16:27 all right so carrying on welcome hey I
1:16:31 don't know if you knew this name's Kyle
1:16:32 Shannon it's the home of smooth jazz and
1:16:37 huh fantastic Bob come on in the water's
1:16:41 nice you're gonna like it in here
1:16:43 artificial intelligence machine learning
1:16:46 we do it all
1:16:48 what are what are my qualifications
1:16:50 that's cute that's very cute anyway
1:16:53 carrying on all right
1:16:56 Bruno we're Claude fan girls up in here
1:16:58 right Silver Fox exactly
1:17:03 that's awesome
1:17:07 oh man yeah I I'm Claude is Claude is
1:17:11 listen I'm a big fan of anthropic I'm a
1:17:14 big fan of the Constitutional learning
1:17:16 model I want them to win what the weird
1:17:19 thing about about
1:17:22 anthropic is that their primary mission
1:17:26 is not to make
1:17:29 AGI
1:17:31 open ai's Prime imperative is to get to
1:17:35 AGI artificial general intelligence
1:17:37 basically a chat that's smarter than all
1:17:39 of us
1:17:40 and can reason better
1:17:43 anthropic's core mission is to make sure
1:17:45 that whoever gets the a AGI doesn't
1:17:48 destroy the world so anthropic's all
1:17:51 about safety so my only concern with
1:17:54 them is because their primary focus is
1:17:56 not the tool but the safety of the tool
1:18:00 um
1:18:01 they may not keep keep up but they're
1:18:04 doing pretty [ __ ] good right now and
1:18:06 their hundred thousand token context
1:18:08 Windows pretty flipping amazing gpt4 is
1:18:11 programmed much better for probability
1:18:13 statistics mathematics calculation or
1:18:15 calculus and Etc
1:18:17 yeah and I find it much better for for
1:18:19 General writing stuff as well like I I
1:18:22 think the I was talking to a reporter
1:18:24 friend of mine today that I knew from
1:18:26 New York days back in the 90s and and
1:18:29 she was saying that she liked what
1:18:32 Claude could process but when it
1:18:34 actually came to like writing articles
1:18:35 she was like boy was it boring it was
1:18:37 really bad so all again all of these
1:18:40 tools learning what they're good at what
1:18:42 they're not good at is is a big part of
1:18:44 the trick
1:18:47 all right let's see uh if you're new
1:18:50 here pop any questions you have down
1:18:52 below and I'll spend I'll spend 45
1:18:55 minutes trying to get these stupid tools
1:18:58 to look at a PDF file
1:19:01 that was the first time we Dove that
1:19:03 deep into something that was pretty cool
1:19:04 though that was that was actually a
1:19:06 really good
1:19:07 um like I feel like that was a good
1:19:08 learning exercise for me hopefully
1:19:10 hopefully you know you're paying
1:19:12 attention a little bit I know I wasn't
1:19:14 exactly uh being articulate in what I
1:19:17 was doing but I was just trying to
1:19:19 figure some [ __ ] out all right there's
1:19:21 talk about AI regulations yeah so so
1:19:24 um in fact I just the the talk that I
1:19:27 gave tonight
1:19:28 today D I cut together this video from
1:19:33 Elon Musk
1:19:35 where he was talking about so so he and
1:19:38 whatever 10 10 other AI CEOs
1:19:44 um met with Chuck Schumer in a in a in
1:19:47 an AI meeting how do you think they took
1:19:50 the message that you brought them well I
1:19:52 think it was it was very um civilized
1:19:54 discussion basically what he says is
1:19:56 everyone in there thinks that regulation
1:19:58 is important
1:20:00 um he feels that
1:20:02 um there will likely be some sort of
1:20:04 like FAA or or FCC kind of you know
1:20:08 regulatory body and he supports that and
1:20:11 you know
1:20:12 he said he thinks that the the meeting
1:20:15 with Schumer may go down as historic so
1:20:17 they didn't none of them really talked
1:20:19 about what they talked about in there
1:20:20 I realize I can get rid of all insurance
1:20:22 no wife no children and rather die than
1:20:26 look at that I know and and you know
1:20:28 what Winston what what is the really
1:20:30 [ __ ] up thing about that statement
1:20:31 about insurance is that's exactly that
1:20:35 that is by Design that's exactly the
1:20:37 response that they want if we make this
1:20:39 complicated enough
1:20:41 people will just give up and stop trying
1:20:44 to fight us to get us to pay for their
1:20:46 [ __ ]
1:20:47 It Is by Design and it's disgusting it
1:20:51 is Criminal
1:20:52 open enrollment season at AI learning
1:20:56 um
1:21:03 isn't that special now we know what what
1:21:06 they're not good at
1:21:09 aren't you special well we give you a
1:21:13 little simple Health Care exercise there
1:21:15 and you couldn't do it why was that why
1:21:18 do you think you couldn't do that Chachi
1:21:19 PT was it that you're not smart enough
1:21:21 no that's not it I wonder what it could
1:21:23 oh could be Satan
1:21:29 might have been Satan
1:21:31 uh well it was just okay for me that's
1:21:33 just mean I appreciated seeing all the
1:21:36 back and forth to make it work yeah a
1:21:38 little bit
1:21:39 um
1:21:40 it that that's one of the other things
1:21:42 about the
1:21:45 working with chat GPT is much more like
1:21:48 working with a really thick-headed
1:21:50 intern
1:21:52 then like a Google search it's it's much
1:21:56 more like wait how thick are you oh God
1:21:59 you're really thick okay let me spell it
1:22:01 out
1:22:03 um they just happen to be really well
1:22:05 educated
1:22:06 [Music]
1:22:08 guilty King true
1:22:12 um put Medicare into the equation oh my
1:22:14 God I'm in the middle of that right now
1:22:16 yeah
1:22:19 Claude feels like you're in a big empty
1:22:21 room sometimes Pie's couch is much more
1:22:23 comfortable
1:22:27 oh that's hilarious
1:22:31 uh I'm at the bottom of comments here
1:22:35 um
1:22:36 let me see you know what I'll do let me
1:22:38 um because there might be some folks in
1:22:40 here if you have any questions about a
1:22:42 AI pop them in the comments below
1:22:44 um I've Got Mods that'll that'll pop
1:22:46 them to the Top If it's anything
1:22:48 interesting I'm gonna show you right now
1:22:51 um
1:22:54 all these tools that I was kind of
1:22:56 talking about generically before just
1:22:57 just because if there's people in here
1:22:59 that are new so if you're brand new to
1:23:01 chat GPT go go right there if you
1:23:03 haven't played with chat EBT that's the
1:23:05 official chat GPT website Claude and
1:23:08 Bard and Bing are kind of the the top
1:23:12 four that I I sort of Bounce Around
1:23:14 between
1:23:15 woohoo thank you Kyle back oh Kyle back
1:23:19 to funny characters and interesting
1:23:21 voices what what darling I don't know
1:23:23 what you're talking about why would you
1:23:25 why would you say that my my characters
1:23:27 are funny
1:23:29 my characters are very very serious well
1:23:32 you're wrong Dr brusa I'm very proud to
1:23:35 be a woman name that movie name that
1:23:37 movie
1:23:39 all right
1:23:41 um
1:23:42 it's a classic it's a classic
1:23:45 you're awesome Kyle you completed my
1:23:47 thoughts exactly with the church lady
1:23:49 I'm good perfect that's why I'm here
1:23:53 um so Claude we just played with Claude
1:23:56 you can put really big documents in Bard
1:23:58 is connected to the internet so those
1:24:00 three are good
1:24:01 um
1:24:03 Bing is also connected to the internet
1:24:05 it can also make images even though
1:24:07 they're shitty it can make them it can
1:24:09 read images uh it can occasionally
1:24:11 answer a question but the safety guard
1:24:13 rails on it are so egregious that you
1:24:16 basically can't get anything done with
1:24:17 it they got to fix that [ __ ]
1:24:20 um
1:24:21 metas llama 2 models are being hosted at
1:24:24 labs.perplexity that's pretty
1:24:26 interesting Pi dot AI you can chat with
1:24:28 that's really good Poe is they've got
1:24:31 Bots and things like that oh
1:24:34 Winston coming in strong with the Llama
1:24:36 did you see what oh and Winston just hit
1:24:38 Level 18. take that Clint
1:24:46 did anyone get my movie reference you're
1:24:49 wrong Dr Brewster I'm very proud to be a
1:24:51 woman
1:24:56 [Laughter]
1:24:59 oh my God that was a good movie
1:25:02 um okay so that's those
1:25:04 then these there's a lot going on here
1:25:07 so screenshot screenshot right now and
1:25:09 if you're on a Samsung you can't not Mrs
1:25:11 Doubtfire close Silver Fox nope not Mrs
1:25:14 Doubtfire
1:25:16 not Mrs Doubtfire not Robin Williams
1:25:18 you're wrong Dr Brewster I'm very proud
1:25:21 to be a woman Jessica Lang was in it
1:25:26 um
1:25:29 um these are the image generation tools
1:25:31 except Tootsie I'm a librarian winner
1:25:35 winner chicken dinner
1:25:37 it was when the when the creepy soap
1:25:40 opera actor comes in for a kiss and she
1:25:42 slaps him
1:25:43 laughs
1:25:46 and then there's that line from the one
1:25:47 cameraman where where the the director
1:25:50 like the camera Zooms in on Tootsie and
1:25:53 the director goes he goes could you pull
1:25:56 back a little and someone goes how far
1:25:57 and the cameraman goes how about Toledo
1:26:00 oh my God that was a good movie
1:26:04 um
1:26:05 mid-journey is kind of the best of the
1:26:07 image generation tools in terms of just
1:26:09 quality of image but you have to make
1:26:12 images in Discord which is a pain in the
1:26:14 ass uh and
1:26:17 there's not a free version
1:26:19 and it can't do text ideogram is free
1:26:22 right now but my guess is it won't be
1:26:24 for long but it can do text although the
1:26:26 image generation part of it ain't nearly
1:26:29 as good as mid Journeys Leonardo is is
1:26:32 actually a really solid product
1:26:34 web-based image generation tool
1:26:38 um based on stable diffusion and then
1:26:41 heygen.com
1:26:43 go to labs.hn.com and click on the video
1:26:47 translation tool
1:26:49 and you can make
1:26:54 could not be reverted is this working
1:26:58 check this out so hey Jen is the video
1:27:00 translation tool there
1:27:03 allows for this if you haven't seen it
1:27:05 so we've heard from a number of clients
1:27:07 that modular content is important to
1:27:08 them getting more efficient with how we
1:27:10 create content and how we leverage
1:27:12 content is
1:27:21 so you'll notice it re-synced my lips it
1:27:24 synth my voice I'm speaking fluent
1:27:27 Italian with the right accent
1:27:30 um so so that's that's hey Jen
1:27:33 these little green boxes
1:27:36 that's hey Jen right there
1:27:42 [Music]
1:27:44 um and then Adobe Firefly and
1:27:46 photoshop's generative fill went live
1:27:49 two days ago so it's out of beta and
1:27:53 what that basically means is that
1:27:54 anything you make in the Adobe
1:27:56 generative AI stuff you can now use
1:27:58 commercially and their models have not
1:28:01 been trained on copywritten materials so
1:28:03 that's a big big deal
1:28:05 and then openart.ai prompt book will
1:28:08 teach you how to talk to those image
1:28:09 generation tools all right
1:28:12 that's dad stuff let's see if we got
1:28:14 some questions
1:28:16 somebody pinned a comment I don't see
1:28:18 any pinned comments right now with that
1:28:21 Robin Williams no that was Dustin
1:28:22 Hoffman
1:28:24 that was Dustin Hoffman as Tootsie
1:28:27 you're wrong Dr Bruce I'm very proud to
1:28:29 be a woman
1:28:31 such a good movie
1:28:35 maybe hey Jen could have translated the
1:28:38 Moda plans for us oh my God
1:28:42 it couldn't have done worse
1:28:46 wow how many languages eight languages
1:28:49 right now the I saw there was a um
1:28:53 there was a live session with the
1:28:55 co-founder of hey Jen today that a
1:28:57 number of a number of us from in here
1:28:59 were in it
1:29:01 and
1:29:03 the co-founder and chief product officer
1:29:07 Tick Tock won't allow screenshot it
1:29:09 blacks out so what you have to do I
1:29:12 think Clint is go out of the live to
1:29:15 your
1:29:16 um timeline where you can still see the
1:29:18 live I think you can screenshot it out
1:29:20 there or you can go to
1:29:23 um
1:29:24 open it up in a web browser and just let
1:29:27 me know when you do that and I'll just
1:29:29 pop it up on screen again
1:29:32 um
1:29:33 what's that
1:29:36 what that
1:29:38 well when I eat them out in my own
1:29:41 Walmart
1:29:43 um
1:29:45 best tool to help folks with resumes
1:29:50 there may be some specific resume tools
1:29:53 I would say
1:29:56 check out futurepedia.io and just search
1:29:59 for resumes and see if there's like I I
1:30:01 would imagine that in a cool in a tool
1:30:05 like copy AI or or uh what's that other
1:30:08 one called
1:30:10 um
1:30:11 uh what the [ __ ] I can't
1:30:14 um
1:30:15 this is ones that are like pre-prompted
1:30:18 like lots of different widgets that are
1:30:19 all written on top of GPT there might be
1:30:21 some that specifically do resumes
1:30:24 um
1:30:25 what I would do is I would probably use
1:30:28 Claude for this I would upload an old
1:30:32 resume
1:30:34 um copy as much of your data out of your
1:30:37 LinkedIn profile as you can and either
1:30:39 save that as a like a text file or just
1:30:42 paste it into Claude
1:30:44 and then I would I would also put in
1:30:46 either a job description or or describe
1:30:49 the kind of job you want to get and then
1:30:51 just have it write a resume based on a
1:30:53 previous resume or based on an example
1:30:56 resume you like you could go
1:30:58 you could you could go get examples of
1:31:01 good resume formats or even have Claude
1:31:03 or chat GPT tell you that
1:31:06 and then just have it match that format
1:31:08 pulling all your data out of your
1:31:11 um
1:31:13 out of your LinkedIn profile I I would
1:31:15 do something like that just because
1:31:17 again I think that I think that learning
1:31:20 how to interact with these things and
1:31:22 just kind of troubleshoot with them as
1:31:24 much of a pain in the ass as it is every
1:31:26 time I do that like it sort of locks
1:31:28 something away in my brain of like okay
1:31:30 now I know how to get around that thing
1:31:33 now now I know how to problem solve that
1:31:35 because I you know the next two years
1:31:38 are just going to be lots and lots of
1:31:40 tools resume help might be through
1:31:42 gpt4's plug-in ambition great so I don't
1:31:46 know that plug-in so I'm a librarian
1:31:50 um
1:31:50 uh Firefly just made that
1:31:54 um
1:31:55 recommendation ambition is a I guess
1:31:57 it's a resume building tool that's
1:32:00 really cool that's really smart I like
1:32:02 it it's solid solid
1:32:05 tools to do text to speech with Avatar
1:32:07 and input to hey Jen
1:32:10 well there's a hey gen
1:32:12 um
1:32:13 there's a hey gen plug-in
1:32:15 which you can just tell it to go make
1:32:18 you a video and it will sort of pick its
1:32:20 own it'll pick its own avatar and voice
1:32:23 thing you can go to hey Jen you can go
1:32:25 to dash ID you can go to Synthesia
1:32:29 you can go to uh what's the other one
1:32:33 and I can't remember it
1:32:36 does I'm helping patrons with theirs and
1:32:39 sometimes they don't have digital skills
1:32:41 at all yeah yeah that's that's rough
1:32:45 um in in that case
1:32:53 I assume canva has resumes in it doesn't
1:32:56 it I I would say go go look at the
1:32:58 plugin that Firefly just talked about
1:32:59 ambition
1:33:02 I would say go there and then let me go
1:33:05 see if canva has resumes I bet they do
1:33:09 canva.com
1:33:13 resume
1:33:15 [Music]
1:33:20 yeah so canva has oh they've got piles
1:33:25 of resume templates so
1:33:32 yeah
1:33:35 so
1:33:38 I'd go find a template that looks good
1:33:40 here
1:33:42 and then you could probably
1:33:45 a couple of ways to do it but I would
1:33:48 basically just sort of copy copy the
1:33:50 text out of one of these templates
1:33:52 and pop it into
1:33:55 um
1:33:57 Chachi be to your Claude have it rewrite
1:33:59 it with
1:34:01 I think this is going to be a clod thing
1:34:03 because it's you got you got you got too
1:34:05 many too many documents here you take
1:34:07 the text from
1:34:09 one of the resumes from canva as the
1:34:12 format
1:34:13 or like here's the sections that I need
1:34:16 you take your LinkedIn information or
1:34:19 just whatever job information that they
1:34:22 type in
1:34:23 then you take a job description and then
1:34:26 you have it rewrite their resume kind of
1:34:28 based on that format
1:34:30 using their data
1:34:36 there's probably a tool well you know
1:34:38 there's a oh wait hang on
1:34:42 I don't need to be a dumb dumb about
1:34:44 this I'm a little Punchy because I I
1:34:46 give a talk tonight so I'm a little I'm
1:34:48 a little like uh whatever you call it
1:34:50 well
1:34:51 look at my hair uh there's there's a lot
1:34:55 going on here tonight
1:34:57 um
1:34:58 let me try something there's a canva
1:35:00 plug-in let's go try the canva plugin
1:35:07 [Music]
1:35:12 plug-in store canva
1:35:17 canva install okay so I'm installing
1:35:21 the canva plugin
1:35:23 and now it's turned on and I'm going to
1:35:26 say
1:35:27 um let me go to LinkedIn
1:35:30 [Music]
1:35:31 linked and
1:35:40 Kyle Shannon that's me you did it yay
1:35:46 experience can I show all experiences
1:35:49 show all experiences
1:35:52 he's a chief generator officer yes he is
1:35:56 he's a Founder he's a Content publisher
1:35:59 he's an EVP creative he's all of those
1:36:02 things
1:36:04 all right
1:36:05 there's all my [ __ ]
1:36:07 so I'm gonna grab all my [ __ ] and then
1:36:10 I'm gonna go in here to chat GPT and I'm
1:36:12 gonna go
1:36:16 based on
1:36:18 this LinkedIn profile
1:36:27 I'm gonna paste all that crap in I hope
1:36:29 it's not too long
1:36:32 um
1:36:37 write me a resume
1:36:41 that
1:36:43 amplifies my experience with AI
1:36:47 so that I can get a nine hundred
1:36:50 thousand dollar prompt engineer job
1:36:59 let's see if let's see if it sends it
1:37:01 off and does anything on canva
1:37:06 um objective dedicated experienced
1:37:08 professional with a profound
1:37:10 understanding of generative Technologies
1:37:12 you hear that people I have a profound
1:37:14 understanding it's not an understanding
1:37:17 it's not an understanding it's a
1:37:19 profound understanding
1:37:22 seeking a position as a prompt engineer
1:37:24 with a rich history and creative
1:37:26 strategy content Evolution and AI driven
1:37:29 initiatives I aim to leverage my
1:37:32 expertise okay
1:37:34 so
1:37:36 it did that stuff pretty good
1:37:41 Okay so
1:37:45 can you find a template on canva and put
1:37:49 all of that information in
1:37:53 the template so it looks pretty and
1:37:56 professional and so I don't have to do
1:37:57 any technical work
1:38:01 let's see if it does it I'm a librarian
1:38:03 let's see if we can help your people
1:38:06 do this
1:38:08 come on chat gbt you failed you failed
1:38:10 Emilio's wife it was it was really not
1:38:12 good it was really you were really bad
1:38:15 tonight
1:38:15 Emilio's wife oh
1:38:19 Marianna Anderson hello what
1:38:24 oh
1:38:26 I found some professional templates on
1:38:28 canva for you
1:38:30 here are some options so there's black
1:38:33 and white minimalist
1:38:35 there's minimalist white and gray
1:38:40 and there's blue professional
1:38:43 that one looks like some sort of Romper
1:38:45 Room thing
1:38:48 that one's nice and professional I think
1:38:51 that one's more classic so if I were if
1:38:53 I were applying for a Fortune 500 job
1:38:56 I'd go down but let's go with let's go
1:38:58 with black and white minimalist oh wait
1:39:00 oh there's more
1:39:03 pink simple
1:39:05 all right select one of the templates
1:39:07 above once you've made your choice I'll
1:39:09 help populate it with your resume
1:39:10 details I'm a librarian look how cool
1:39:13 this is this is
1:39:15 this is like this is like my night to
1:39:17 learn I'm like this is the AI learning
1:39:20 lab it's going the other direction
1:39:21 tonight I'm digging it look at all this
1:39:24 [ __ ] I'm learning did good very good
1:39:27 deal I'll tell you what I like them
1:39:29 taters
1:39:31 all right Taylor's tasty it was a sling
1:39:34 blade
1:39:37 um I'm still going black and white
1:39:39 minimalist black and white minimalist
1:39:42 BW
1:39:44 minimal list
1:39:48 all the
1:39:51 way
1:39:55 great choice I'll populate it
1:40:04 [Music]
1:40:05 wait however I can't directly edit the
1:40:09 template for you what
1:40:14 that's horse hockey
1:40:21 edit your personal details replace the
1:40:24 existing text
1:40:41 [Music]
1:40:41 um
1:40:42 camera allows for a lot of customization
1:40:46 um
1:40:47 if I ask really nicely and say pretty
1:40:50 please could you put my data into the
1:40:52 template
1:40:56 I know right right bubbles it was so
1:40:58 close it was so close to not sucking
1:41:04 I truly appreciate your patience and
1:41:06 politeness however due to the current
1:41:09 limitations of my integration with canva
1:41:11 I'm unable to directly well I mean
1:41:15 listen here's the deal
1:41:16 it it at least did pick us some nice you
1:41:20 know professional templates and we've
1:41:22 got all of the all of the data here so
1:41:25 so even without technical still skills
1:41:29 you could sort of copy and paste this
1:41:30 stuff into the appropriate sections of
1:41:32 the other resumes so it's not totally
1:41:34 useless I I you know there you go
1:41:36 hopefully that helps
1:41:39 I'm a librarian if pretty please works I
1:41:42 tried pretty please
1:41:44 you see I'm I'm willing to work with
1:41:48 these AI things they're just not willing
1:41:49 to work with me tonight
1:41:51 listen
1:41:53 a poor a poor carpenter blames his tools
1:41:59 it's all chatty PT's fault I'm really
1:42:02 smart people listen listen I know some
1:42:05 people in here came in and they said
1:42:07 that I was not qualified to teach this
1:42:09 class
1:42:11 but but
1:42:13 let me tell you I
1:42:17 I have a certificate
1:42:21 I worked with George Santos and I have a
1:42:24 certificate and it's a it's a it's a
1:42:28 university it's a it's a degree it's got
1:42:32 a degree it's got I'm
1:42:34 I'm qualified
1:42:36 also three AI generated resume building
1:42:39 sites okay grammarly LinkedIn Resume
1:42:41 Builder and Hemingway editor awesome
1:42:45 Firefly thank you yeah I figured there
1:42:46 were tools out there I just I just don't
1:42:48 know that world at all so yes awesome
1:42:52 Emilio's wife unreal incompetence thank
1:42:54 you appreciate it
1:43:09 whoa
1:43:10 they can't make it that easy I know well
1:43:14 some of those some of the more dedicated
1:43:15 tools I'm sure are actually really good
1:43:17 at this because all you need to do is is
1:43:19 is basically
1:43:22 generate your
1:43:24 um
1:43:25 your chat GPT with with structured like
1:43:28 Json or or HTML and then you can just
1:43:32 pop it in a template right so
1:43:34 technically it's it's not that hard to
1:43:36 do so I would I would imagine that any
1:43:39 of the resume building
1:43:41 sites
1:43:43 um
1:43:46 have added AI [ __ ] so this is relatively
1:43:48 straightforward I'm asleep can't stay
1:43:50 awake good night my favorite people
1:43:51 night share bear sleep tight sleep tight
1:43:54 good night good night thank you for
1:43:57 joining tonight
1:44:03 sense of identity because of her capture
1:44:06 Emilio
1:44:08 yeah we'll never know her name right
1:44:13 but he's very kind to me
1:44:17 foreign
1:44:21 I just almost shot water out my face
1:44:25 I you did the best you could I did do
1:44:28 the best I could I'm a librarian
1:44:30 hopefully that helped a little bit
1:44:32 um the the tools that Firefly told you
1:44:35 about are probably a much better
1:44:36 solution than my my Approach but you
1:44:39 know I just do this to entertain myself
1:44:41 quite frankly at this point because you
1:44:44 know I I am wholly unqualified to this
1:44:47 was a night of failures and and you know
1:44:50 we celebrate our failures here at the AI
1:44:52 learning lab you can make money with
1:44:54 Chachi mate I don't know what you're
1:44:56 talking about
1:44:57 oh
1:45:02 oh my Lord Kyle you're an awesome
1:45:06 Professor well
1:45:09 I profess a lot
1:45:14 oh
1:45:17 dude
1:45:20 [Music]
1:45:23 are we glad we can't pin Kyle's singing
1:45:27 that's rude that's rude I like it
1:45:36 boom boom
1:45:38 Kyle you were not incompetent the people
1:45:41 who don't wait the people who don't
1:45:43 replace your name here with their name
1:45:45 are yeah exactly exactly
1:45:50 I made assumptions that their tool would
1:45:52 not be a
1:45:55 janky piece of [ __ ] and they're all
1:45:58 janky pieces of [ __ ] right now I've said
1:46:00 it before I'll say it again
1:46:02 tell us more about that hey gen meeting
1:46:05 um okay
1:46:07 so they started seven minutes late
1:46:08 because their social media manager
1:46:11 couldn't get YouTube running
1:46:13 [Laughter]
1:46:17 so there's that
1:46:19 then Wayne the co-founder and CPO which
1:46:25 I assume means Chief product officer
1:46:29 explained
1:46:32 kind of like I explained [ __ ] he was
1:46:34 like well you upload a file and then it
1:46:37 turns into a video and then
1:46:39 and then you pick a language and it
1:46:41 makes it in a different language and
1:46:42 people are like show us the [ __ ] show us
1:46:45 the [ __ ] and then other people are like
1:46:47 we already know what it is what is this
1:46:49 meeting for it was it was the the crowd
1:46:53 was a little hostile
1:46:55 um
1:46:56 and then
1:46:57 um
1:46:59 and then he showed it and he said he
1:47:03 said uh
1:47:04 so so they they launched hn video
1:47:07 translate they didn't launch it they
1:47:10 they had it in some sort of functional
1:47:12 form on August 7th
1:47:15 and it wasn't until like I don't know
1:47:17 two weeks or so ago that they put it out
1:47:20 but like I think it had been up on the
1:47:22 website in just kind of a non-announced
1:47:24 way for like a month and basically no
1:47:26 one used it no one noticed it
1:47:29 and then a couple of videos got posted
1:47:33 on
1:47:34 um
1:47:35 on Tick Tock and I think their social
1:47:37 media manager posted another one on Tick
1:47:39 Tock or or they they boosted one
1:47:43 um and then all of a sudden he said in
1:47:45 the past week like they're they're
1:47:47 totally blowing up on social media and
1:47:50 he was gonna he was gonna
1:47:53 um translate one of his videos live but
1:47:55 the queue was like
1:47:58 I can't remember if it was fifty seven
1:47:59 thousand or five hundred and seventy
1:48:01 thousand there was like a really long
1:48:03 queue of videos that had been you know
1:48:05 uploaded to the service
1:48:08 um
1:48:10 he said a couple of really interesting
1:48:12 things one of the things that he said
1:48:14 that absolutely blew me [ __ ] away was
1:48:18 he he was sort of talking about hey Jen
1:48:20 is this thing where you can do voice
1:48:22 synthesis and you can type in text and
1:48:25 it'll make an avatar and do that that's
1:48:27 their primary business he kind of said
1:48:29 video translates was like a side project
1:48:32 of one of their Engineers it was like a
1:48:34 throwaway project
1:48:36 and he doesn't understand why people
1:48:38 think it's such a big deal effectively
1:48:40 he didn't say exactly those words but he
1:48:42 was like we're a little caught off guard
1:48:44 here that so many people want to use
1:48:46 this
1:48:47 but he said they're working really hard
1:48:49 to solve the scalability problem because
1:48:51 a lot of people are using it now
1:48:54 they have pricing for both
1:48:57 60 bucks a month
1:48:59 for like 30 credits and then 179 bucks a
1:49:03 month for
1:49:04 60 credits that are up to 20 minutes
1:49:06 each and there's an API
1:49:08 he said on the call that the API doesn't
1:49:11 exist yet so like this is this is a
1:49:15 afterthought project
1:49:17 that I think is going to become their
1:49:20 primary product very quickly and I don't
1:49:22 know if they're going to be able to
1:49:23 scale it up so so one of the risks right
1:49:26 now is
1:49:27 um I don't think they I this is sort of
1:49:29 catching a tiger by the tail I don't
1:49:31 think they know what they have on their
1:49:32 hands
1:49:34 um
1:49:35 some of the other stuff that he said
1:49:36 that was really interesting was
1:49:39 um
1:49:41 that it's going to get much faster soon
1:49:46 and it'll probably be real time
1:49:50 in not too long like he kind of implied
1:49:53 in less than a year I don't know if
1:49:55 that's a reality again I don't think
1:49:57 they know what they have on their hands
1:49:58 here
1:49:59 but when this thing's real time
1:50:02 holy crap
1:50:09 all right let's see
1:50:12 I'm trying to think what else there was
1:50:14 one other thing they said
1:50:16 it was the real time thing
1:50:19 oh
1:50:22 he knows they need Japanese lots of
1:50:24 people are asking for Japanese they've
1:50:26 got eight languages right now
1:50:28 they're gonna go to
1:50:30 20 and then he said a lot after that
1:50:36 I saw it blew up on the Asia social
1:50:38 media circuit that's interesting Crystal
1:50:40 yeah interesting
1:50:42 um yeah clear clearly it caught fire
1:50:44 somewhere because I I mean I've been
1:50:45 talking about a lot but I don't I don't
1:50:47 like every time I show someone no one's
1:50:49 heard of it
1:50:51 um
1:50:52 but every time I show people they're
1:50:53 like uh holy [ __ ] oh my God
1:50:57 um so he said they're gonna they're
1:50:59 gonna go to 20 languages pretty quick so
1:51:01 I assume Asian Asian languages they've
1:51:04 got Hindi in there right now but that's
1:51:05 the only thing that's you know sort of
1:51:08 Beyond Western Europe
1:51:09 and then
1:51:11 um
1:51:12 he said they will be going to dialects
1:51:14 so you'll be able to do English with an
1:51:17 American accent with an English accent
1:51:19 with an Australian accent so so they're
1:51:21 gonna start to have different dialects
1:51:24 it was really hard to understand from
1:51:28 him
1:51:29 wear any any of this stuff was on the
1:51:31 road map
1:51:32 like I couldn't tell if it was just like
1:51:34 oh we could probably do more than 20
1:51:37 languages like is he just sort of making
1:51:39 it up as he goes or you know again back
1:51:42 to this idea like like these things are
1:51:44 kind of janky pieces of [ __ ]
1:51:47 um this one feels like it was really
1:51:50 um you know kind of just tossed together
1:51:52 and put out there so
1:51:55 um
1:51:56 they have a hit on their hands if they
1:51:58 can keep delivering it and apparently
1:52:00 their Engineers not sleeping trying to
1:52:02 scale this thing can you imagine
1:52:05 because I'm I'm sure that's it's got to
1:52:08 be taking a lot of gpus because they're
1:52:09 doing multiple passes right they're
1:52:11 doing they're doing the transcription
1:52:13 pass they're doing the voice synthesis
1:52:16 pass they're doing the translation pass
1:52:19 they're doing the the the uh the lip
1:52:25 um
1:52:26 modeling
1:52:27 then they're doing the lip
1:52:29 synchronization then they're doing the
1:52:30 layering like there's there's I don't
1:52:32 know seven or eight or nine operations
1:52:35 that have to happen to do one of those
1:52:36 translations if you don't know what I'm
1:52:38 talking about the uh
1:52:40 it's it's this thing called hey Jen and
1:52:43 it
1:52:44 um this is me speaking Italian which I
1:52:46 don't speak that's how we think about
1:52:48 stories well that's English break
1:52:50 stories down into little atoms right
1:52:52 little modules
1:53:01 so that's that's the output it's it's
1:53:03 remarkable it's got one visual glitch
1:53:06 which I think is kind of minor which is
1:53:07 if you pass your hand over your face
1:53:09 since since they're they're sensing like
1:53:13 your lower mask when you pass your hand
1:53:15 over your face your mouth is actually in
1:53:17 front of your hand it's pretty weird
1:53:18 looking
1:53:19 more languages to be added yeah I think
1:53:21 he said he said up to 20. listen yeah I
1:53:24 thought Kyle would never return from the
1:53:25 insurance procedure he was locked into
1:53:27 it was ugly man I I know right there
1:53:30 with you I'm it drained me it drained me
1:53:34 it could have been a champ I suppose
1:53:36 with a Canadian accent a yeah apparently
1:53:39 the the uh the uh French
1:53:43 the French accent is a Quebec accent
1:53:46 it's not French
1:53:49 um so that's one of those things where
1:53:51 again I like
1:53:53 this was put together by an engineer so
1:53:56 I again I don't I don't know that they
1:53:58 know what they have on their hands here
1:53:59 this is this is a very big deal product
1:54:01 and it I I think the result of this
1:54:04 product will likely be that two or three
1:54:07 other or four
1:54:09 companies that have similar technology
1:54:12 are going to come out
1:54:13 strong with competitors to this so I
1:54:16 think that over the next six months
1:54:18 we'll probably see three or four of
1:54:20 these tools that that make it really
1:54:22 really easy to do
1:54:25 um translation into many many languages
1:54:27 so it's crazy
1:54:29 it's absolutely nuts
1:54:32 I sent hey Jen my viscom team we work in
1:54:36 20 primary languages in 70 countries
1:54:39 very cool
1:54:43 what's it called hey Jen yep everyone's
1:54:45 giving you that down below Firefly right
1:54:47 and business
1:54:50 can you imagine how this will help
1:54:51 teachers wow I know
1:54:54 just anyone really I mean teachers
1:54:56 anyone with a like a YouTube channel can
1:54:59 now just like enter another Market
1:55:02 speak in that language fluently
1:55:05 [ __ ] crazy all right listen listen
1:55:08 everybody listen we have to have a talk
1:55:11 okay come here come here we're leaving
1:55:14 close listen live in
1:55:16 I'm a little sleepy what time is it 10
1:55:20 21 here in Denver Colorado
1:55:22 I'm not sleepy so I'm gonna go
1:55:26 I think y'all are wonderful
1:55:28 uh I'm glad we did the insurance deep
1:55:31 dive except it did make me crankier than
1:55:33 [ __ ] because I gotta go deal with that
1:55:35 with my actual insurance
1:55:37 I don't have as many choices as Emilio's
1:55:39 wife did and I don't have to look at a
1:55:41 sheet that's nearly that depressing but
1:55:43 they're all that depressing
1:55:45 I mean actually what was quite
1:55:47 remarkable about the seven choices that
1:55:50 she had is there really wasn't a ton of
1:55:53 difference between them they're all
1:55:54 between 9 500 and 11 800. like
1:56:01 all of them are going to frustrate the
1:56:03 [ __ ] out of her for the next year
1:56:05 and then she gets to do it again the
1:56:07 following year and they get to reduce
1:56:09 even more what they give her
1:56:13 welcome to the AI learning lab where we
1:56:15 talk about The Bleak reality of
1:56:17 capitalism in America in 2023
1:56:21 where where if you can't afford your
1:56:24 health insurance
1:56:28 to be able to pay for more health
1:56:29 insurance so yeah
1:56:32 that's why we're here
1:56:33 uh it's been Gracie and y'all thank you
1:56:36 all for hanging out thanks for tapping
1:56:38 the like button so much you all are
1:56:39 blessed to be with I swear to God
1:56:42 um I'm at work I don't have many choices
1:56:43 either good night Kyle and everyone good
1:56:45 night great live thank you everyone I
1:56:47 know it got a little got a little geeky
1:56:48 there but I appreciate it 100 bucks a
1:56:50 month for Primary Care no insurance oh
1:56:53 interesting
1:56:54 um way to tie up the night in a bow
1:56:55 thank you very much thanks cabruno Joker
1:56:58 Firefly peace out Emilio's wife
1:57:00 goodnight goodnight goodnight Stevo
1:57:03 thank you Tobias great to see you okay
1:57:06 farmer hope you're doing well hope your
1:57:08 trip was good
1:57:09 peace peace peace Emilio's wife owes me
1:57:12 a beer or a whiskey I'll take it I will
1:57:15 take it hair growth geek peace out
1:57:18 everybody all right good night Winston
1:57:20 bye everyone