0:00

Hello everybody . We have people joining

0:02

in , all

0:05

right , welcome . Welcome everybody

0:07

, as you're joining . Great

0:10

to see everybody joining . Some familiar faces

0:12

here . I

0:16

have captions turned on but unfortunately it's like

0:18

hiding some of the list of people so

0:20

I can't . I'm like , oh , somebody just joined

0:22

, who is it ? I can't tell because of the captions

0:25

. How's

0:30

it sound out there to everybody ? You guys

0:32

doing okay ? Can you give us thumbs up

0:34

that you have good audio quality or

0:36

are you having challenges with audio ?

0:39

Se e a lot of thumbs ups . That's good .

0:43

Two , I do . All

0:53

right , we'll give about a minute and get started here . I

0:58

see some people from the AI

1:01

garage . I see somebody I haven't

1:03

talked to in many years hey , adam , all right . All right , it is 10 am

1:05

, so we'll

1:07

get started . Hi everybody , thanks for joining

1:09

us on this inaugural

1:12

episode of the Boring AI Show

1:14

. We're really excited to

1:16

be chatting with you guys today , and

1:19

our focus is AI

1:21

literacy today . Joining

1:23

me is Tali . Thank

1:25

you , tali , for going on this adventure with

1:27

me . I appreciate it . Of

1:32

course . It's exciting , and

1:35

so our goal today is to help you get some baseline around AI literacy and

1:38

share some things that are happening in news , some

1:41

things that we're seeing winning , and we have

1:43

a special guest today who will be joining

1:45

us via speakerphone

1:48

, so we'll see how that audio

1:50

quality works and see what we can do from there

1:52

. But , to kick off

1:54

, you know , one of the things we want to start with is

1:56

what's happening in the news . What are Tali

1:59

and I watching , what do we think that you

2:01

guys should be watching , and how do you build

2:03

up your AI knowledge ? By seeing what's going

2:05

on out in the world . So , tali

2:07

, what's going on in the news ?

2:09

Yeah , so the news article that I was looking

2:12

at was with regards to Disney

2:14

and it was about how Disney is creating

2:16

an AI task force to

2:19

explore AI and cut costs , and

2:21

this is supported by 11

2:23

job openings that they have looking

2:25

for applicants who

2:27

have expertise in the area of AI and

2:29

machine learning . So I just

2:31

thought this was a really interesting news

2:34

article to bring up . I do want to note

2:36

that this task

2:38

force was launched prior to the Hollywood

2:40

writer's strike and that Disney's been

2:42

pretty careful about how it discusses

2:44

the Hollywood writer's strike and

2:46

that Disney's been pretty careful about how

2:48

it discusses AI publicly . But

2:55

obviously , you know , media companies really have to figure out , you know , how to go about adopting

2:57

AI or risk becoming obsolete , and obviously Disney has

2:59

a pretty strong history and

3:01

track record for investing in new

3:03

types of technological innovation . But

3:06

I thought this was a really interesting article

3:09

because it touched on a lot of really unique use

3:11

cases and , tim , I'd love to hear your

3:13

thoughts on some of these . The main use cases

3:15

they touched on were AI-powered

3:19

ad systems so using AI in their

3:22

advertising department , ai

3:24

to enhance customer service . So

3:26

they brought up an example of creating

3:29

novel interactions for the guests at

3:31

the park . The

3:34

specific example they used I don't know if

3:36

you're familiar with Baby Groot , but creating

3:39

that character that guests could

3:41

interact with in real time , which I think is really

3:43

fun . And then , of course , ai and

3:45

effects , not necessarily to replace

3:48

human actors but to augment

3:50

the digital effects . So I'm curious if

3:52

you have any thoughts on those use

3:54

cases or if you think that they would

3:57

probably use AI even beyond those

3:59

three .

4:01

I'm sure they've got a ton of stuff going

4:03

on in the AI world . I mean , disney

4:05

is like one of the world leaders in

4:07

software patents and technology patents

4:09

and , you know no

4:11

surprise their history has all they've

4:14

been very focused on the

4:17

latest , greatest technology to create

4:19

that magical experience . Disclaimer

4:22

I'm a Disney nerd . I love Disney

4:24

, I'm a big fan , so that

4:27

is no surprise to anyone who knows me on this

4:29

call . So you know

4:32

, the ad aspect like that's

4:34

a great example . We talk about boring AI

4:36

. That's a great example of a

4:38

use case in business , like you know

4:40

, using machine learning to optimize your

4:42

ads and optimize your communication

4:44

totally that is a standard

4:47

business use case . Creating

4:50

robotic systems to interact

4:52

with guests at your

4:54

theme park most businesses don't have

4:56

a theme park as far as I know , so

4:58

I think that's kind of a novel and

5:01

interesting thing . Novel

5:04

and interesting thing I would really like to see . One of the things

5:06

I think is done well by Microsoft

5:09

Research is they do a lot of the

5:11

publications around . This is what we're doing

5:13

and how we're doing it and what's going on . I

5:16

think that would be really cool to see Disney doing similar

5:19

things . I don't know how much they do of that

5:21

to kind of share their research

5:23

and what they're working on . Being

5:25

open and transparent with AI I think

5:28

is going to be a very big theme

5:30

in 2023 and 2024

5:32

. So curious

5:34

to see how that plays out , and

5:36

I do give them

5:38

kudos that it's not the $900,000

5:41

a year salary jobs that Netflix

5:44

had posted , which

5:46

kind of was a bad timing .

5:50

Right .

5:51

So very cool , yeah . So

5:53

this , you know I think , is a good , a good story

5:55

to watch , and this was on rudders , correct

5:57

? Disney creates a task , so

6:02

very good stuff . Encourage

6:04

everybody to keep an eye on . You know what Disney

6:06

and crew are doing , not just in

6:08

the entertainment industry , but like from

6:11

a business perspective . You know , I

6:13

would be certain that they're using AI

6:16

for their business processes . That's

6:18

just not what's super attractive to talk

6:20

about right now .

6:22

Generally , it's all the rage attractive

6:25

to talk about right now , generate all the rage . Now

6:29

I do want to . You know , I know you made a joke there about . You know I don't know how many folks

6:31

organizations have a theme park where you would interact

6:33

with the character , but I think that application

6:36

could be applied to , you know , customer service

6:38

at large , for having

6:40

some sort of chat bot that's maybe not necessarily

6:42

animated like a Groot per se , All of that . You know

6:44

that could be really fun . But

6:47

I think there are ways to break down that use case

6:49

in terms of any

6:51

industry I can think of , there's some level of customer

6:53

service that's needed and

6:56

to have a customer service experience

6:59

where they can get maybe even

7:01

basic questions answered . I know we're

7:03

going to introduce Katie here in a bit and that's something that

7:05

we've leveraged almost like a character for Mind

7:08

Over Machines , but I do think there's ways that that

7:10

specific use case you

7:12

know can be leveraged in the customer service world as well

7:14

.

7:15

Oh , totally , and you know the

7:18

key . There is the

7:20

character , right . You want to make sure that

7:23

it's aligned to your brand . Chatbots

7:25

are very common . They really had

7:27

a day in the sun a few years

7:30

ago . That was a big thing and then it kind of

7:32

got quiet around that . And now , with ChatGPT

7:34

, we're seeing this resurgence of

7:37

chatbots . The

7:39

key things there is it is your

7:41

brand ambassador . So , just like Baby

7:43

Groot is a brand ambassador

7:45

for Disney and for Marvel , you

7:47

know , and for a specific demographic

7:50

, you know the I

7:52

Am Groot cartoon is

7:54

. You know it's cute , it's for children

7:57

, but also like people like me who want

7:59

to watch it with their kids , and so you

8:01

know when you're thinking about your characters

8:03

and that customer experience

8:05

. You know how do chatbots

8:08

enable that , what is the personality

8:10

of your brand and how does it manifest

8:12

in your chatbot ? These

8:14

are really important questions

8:16

to think about and this

8:18

is where AI needs to get out

8:21

of IT and IT

8:23

needs to be involved . But you need to have marketing

8:25

in there . You need to have your operations

8:28

team in that discussion , your legal team for

8:30

sure , making sure that

8:32

people are thinking holistically about AI

8:34

and not just as okay

8:37

, well , we can save a couple bucks if we

8:39

have a chat bot answering our call

8:41

center . You know where are the lines in the sand , where's

8:43

the process . You know where are the lines in the sand , where's the process . How

8:46

do you think about escalation

8:48

? How does the human get involved ? These

8:51

are all really important discussion

8:53

topics for people to think through and

8:56

using . I think you're spot on

8:58

saying hey , baby Groot is a good example of

9:00

using AI to

9:02

represent your brand . Good

9:05

stuff , cool , all

9:08

right . So , with that

9:10

said , let's jump into

9:12

our . We're

9:15

going to in these AI , boring AI

9:17

shows , we're going to have a guest come on , and

9:20

today's guest is our

9:22

. We didn't want to subject a human

9:25

to whatever might happen in

9:27

this first episode . You know technology

9:29

wise , so we are subjecting our

9:31

AI assistant , katie

9:33

, with answering

9:36

some questions . And so when

9:38

it came to AI literacy , one

9:40

of the questions we asked and

9:42

we use an array

9:44

of AI tools at Mind Over Machines , and so

9:46

we were working with Jasper , which

9:49

is an awesome tool that we use for

9:51

many things , and we asked Jasper

9:53

the following question we

9:55

said we are hosting a podcast episode

9:57

about AI literacy and would like to

9:59

have an AI system talk about what

10:01

is important in AI literacy . So

10:04

we wanted to get AI's perspective on

10:06

what's important . Jasper , what

10:08

do you think is important for non-technical

10:10

people to understand about AI literacy

10:12

? Jasper gives its

10:14

answer and , like all AI systems

10:16

, the humans now get involved . So Jasper

10:18

put out its content , tali

10:21

and I came into the content and said , okay

10:23

, yeah , okay , this is good . This

10:25

needs to be sharpened up . We need to change

10:27

this , and we used

10:29

Murph AI , which is another

10:31

AI tool that we use at Mind Over Machines , to

10:33

give a voice , and the voice

10:36

that we have is the standard Katie

10:38

voice for Mind Over Machines . We use this

10:40

for training videos . We use this for

10:42

content , like we're about

10:44

to do now . The goal

10:46

here is that , you know , katie has a

10:48

character , katie is a persona that

10:51

we work with , and now Katie is going

10:53

to tell us what she thinks about AI

10:55

literacy and we're going to talk about if

10:57

we agree or not .

10:59

All right .

10:59

Sound good .

11:01

Sounds good to me All right , let's go

11:03

.

11:30

So , this

11:32

is through my speakerphone

11:37

.

11:37

Sounds good to me . Okay . And the first area

11:39

should have a basic understanding of what

11:41

AI is and its potential impact

11:44

on various industries . They

11:46

should grasp the concept of machine learning

11:48

, which is a subset of AI that

11:50

enables systems to learn and improve

11:52

from data . I call this foundational

11:55

knowledge .

11:58

Allie , what do you think about starting with foundational

12:00

knowledge on the technical side of how machine

12:02

learning works ?

12:04

Yeah . So you

12:06

know , I think , depending

12:08

on your role , I don't personally know

12:10

the level of detail

12:13

that a non-technical person would need to know

12:15

with regards to machine learning , in particular

12:18

as a way to start off . I know that you

12:20

know you and I have talked endlessly

12:22

about human-centered

12:25

design thinking and making sure that before

12:27

you delve into anything , you're really understanding

12:30

the problems at hand

12:32

and the business use cases . I do

12:34

think that her call out on understanding

12:37

the impacts of AI

12:39

are really critical , especially as we think about ethical

12:41

AI , which I think you know we'll delve

12:43

into later this conversation , I hope

12:45

. But yeah

12:48

, I think in general it's important

12:50

to have foundational knowledge , but I don't

12:52

know if I would consider the ins and outs of machine learning

12:55

that she touched on necessary

12:57

to get started on that specific journey from a non-technical

12:59

perspective . What are your thoughts ?

13:01

Yeah , I totally agree

13:03

and you know , I think one of the big challenges

13:05

, one of the challenges I consistently encounter

13:07

, is the ability to articulate what

13:10

the problem actually is Like . What is

13:12

the problem people are trying to solve ? Why

13:14

do they think AI is the solution they

13:16

need to jump to ? There are many

13:18

, many business problems that do not

13:21

rise to the level of AI needing

13:23

to be the solution , problems

13:26

that do not rise to the level of AI needing to

13:28

be the solution . So

13:32

, with that said , I really would encourage people to start their AI literacy is actually to start

13:34

being able to articulate what is a business problem , who has

13:36

it , why do they have it , when do

13:38

they have it ? And really

13:40

think through . You know what's

13:42

the context of this problem and

13:45

that will help you understand how to

13:47

solve the problem , because if you just jump

13:49

in with , oh I want to put

13:51

some AI on that , it's not like a medicated

13:53

cream . You know it's AI as a technology

13:56

solution that you need to

13:58

make sure it fits the

14:01

problem . So I think I

14:03

would encourage people that's where you should start is

14:05

really understanding your problem .

14:08

Yeah , agreed .

14:11

So , with that said , this

14:13

is an episode about AI literacy . So

14:15

let's talk a little bit about what is

14:17

machine learning ? So

14:20

machine learning is basically using

14:22

algorithms , which

14:24

are basically code-based instructions , to

14:28

teach a AI

14:30

system how to react in

14:32

a situation . So you are providing

14:35

lots of training data , so data

14:37

that represents your

14:39

ideal scenarios , and here's what's going to happen

14:41

. And you know , just like AI , when

14:44

we are learning something new , we

14:46

are shown examples as humans

14:48

. This is what is . You know this is

14:51

an apple , this is an apple . You know

14:53

you're handed the fruit , the apple . You know this

14:55

is an apple , this is an apple , this is an apple . Eventually

14:57

, somebody puts a laptop in your hand and says this is

14:59

an apple . Well , you're learning

15:02

from all the data of what you're being

15:04

given to say oh , this is what an

15:06

apple is In machine

15:09

learning . That's how the computer

15:11

AI system is learning and

15:17

the fact you're being told it's an Apple , that's referred to as supervised learning . So you're being

15:19

told this is an Apple . The computer system is

15:21

told this is an Apple through

15:23

labeled data and , using

15:25

that labeled data , when this

15:27

computer is handed something it's never seen before

15:30

, it can compare it to its training data

15:32

, which was all about apples and say

15:34

is this an apple or not ? So

15:37

machine learning is

15:39

all about lots of data going

15:41

through algorithms to

15:43

determine what is

15:45

an apple , what is not . How do I learn from

15:47

this data ? There's other methods

15:50

of machine learning , like unsupervised

15:52

learning and reinforcement learning . We

15:54

won't go into those today , other than to say

15:57

you don't always have to have your data labeled

15:59

to be able to learn from it and react

16:01

from it .

16:04

Now Tim if somebody did want to go into that

16:06

and learn more about those . Do you have any recommended

16:09

resources or ways

16:11

to go about maybe learning some of that more

16:14

technical knowledge ?

16:16

Yeah . So if you want to dig

16:18

in , microsoft

16:20

has some really great free training

16:23

on generative AI

16:25

and artificial intelligence through learnmicrosoftcom

16:28

. I'm also a big

16:31

YouTube learning nerd . There's

16:34

tons of videos on this content that you

16:36

can learn from and really dig

16:39

in as deep as you want . But like

16:41

all internet knowledge

16:44

which kind of ties into something

16:46

we'll talk about , I think , later with generative AI , like

16:48

all internet knowledge , approach

16:50

it with a level of suspicion

16:53

. You know ?

16:54

does this person actually know what they're talking about

16:56

?

16:56

Are they giving good advice and

16:58

that's for all learning , no matter if it's internet or a book

17:01

or you know a lecture you're getting from a professor

17:03

, so definitely

17:06

I might say his

17:08

name wrong , but Andrew

17:10

Ng . He is an excellent

17:12

person to watch on LinkedIn

17:15

and see what he's talking about and his company

17:17

, Deep Learning AI , has

17:19

really excellent free classes to

17:22

dig into machine learning and . Ai

17:24

skills , cool , awesome

17:27

, all right you want to hear what

17:29

Katie says next . Let's do it All

17:32

right . Let's see what item number two

17:34

is .

17:36

Next , it's crucial to explore practical

17:38

use cases of AI in everyday

17:40

life and business scenarios . This

17:43

includes identifying areas where

17:45

AI can bring value , such as

17:47

automating repetitive tasks , improving

17:49

decision-making processes or enhancing

17:52

customer experiences . I

17:54

call this real-world applications

17:56

.

17:59

All right . So

18:01

, Tali , what do you think about that ?

18:04

I think that's huge . I think that understanding

18:06

the different possibilities is critical

18:09

, but making sure we're

18:11

not limited to that as well , I think I

18:13

know from my personal experience I'm learning

18:15

new applications and use cases almost every

18:17

day in

18:28

the world and getting creative with different solutions , but I think that

18:30

is a really helpful place to start in order to ground your thinking around AI , especially

18:32

in a specific industry . Yeah , I think understanding

18:34

use cases is really critical and could

18:36

hopefully lead to innovation within your

18:38

own organization or team . What are your

18:40

thoughts there ?

18:42

Yeah , absolutely Build out your use case

18:44

library . That really

18:46

being aware of what are other people

18:48

doing in your industry , what are other

18:50

people doing across industries and

18:52

this is where you know this

18:55

is . I know Tali and I watch this

18:57

obsessively what are people doing ? How

18:59

are ? How is AI being applied ? And

19:02

I just actually got a text message , you

19:05

know , from Adam Dreesen . Thanks

19:07

Adam for sending this over . But his

19:09

question slash comment is around

19:12

thinking about data and

19:14

the challenges you could get with

19:17

how your data is applied in AI

19:19

and , as we all

19:21

know , anyone who's worked in tech knows that

19:24

data man , that can be a mess .

19:27

It really can be a challenge to come

19:29

into an environment and you know Adam's

19:31

question here around insights you

19:34

know , can I get insights from my data ?

19:36

Well , it depends how good your data is

19:38

, but this is also where the real-world

19:40

use cases come into play . Is

19:42

you want insights on your data

19:45

? What kind of insights

19:47

? And I've often encountered people

19:49

that are like , well , I just want the AI to tell

19:51

me stuff . It's like , yeah , that's not

19:54

how this works . You have to have

19:56

a really you

19:59

have to have a question that is answering . You

20:01

can't just kind of wander around

20:03

in the data . Even when

20:05

you're doing clustering and unstructured

20:07

content and unstructured work

20:09

, you're really trying to answer a

20:11

question still . And so

20:13

you know , making sure that people understand

20:16

, have the use case upfront

20:18

. And now let's take a look at your data . Does

20:21

your data actually support you solving

20:23

that use case and , if not , what

20:25

data ? Would Tally

20:27

thoughts on that ?

20:30

Completely agree . I think data

20:32

is critical there and

20:34

, yeah , I think , understanding your purpose

20:37

I think we touched on this with the , you know , the very

20:39

first recommendation that Katie had for

20:41

us , you know , making sure you

20:43

understand a problem at hand that you're actually trying to

20:45

solve , rather than maybe creating problems that don't

20:47

actually exist or just seeking . I

20:50

know folks use the analogy a lot . You know , seeing

20:53

everything as a nail , you

20:55

know , and AI is the hammer just trying

20:57

to hammer anything that's available . Really

20:59

being strategic about what you're doing and how

21:01

you're going about it , especially with the data you're using

21:03

. Yeah , but with

21:05

that being said , I do think there's a lot of really great use

21:08

cases out there . I know , you know , at Miner

21:10

Machines , we use AI . You

21:12

know you mentioned Jasper earlier . We use

21:15

that from everything from , you

21:17

know , marketing to vast amount of knowledge or data . You know

21:19

, if you don't have a full

21:30

data cleanse , leveraging

21:32

an existing system like a chat GPT

21:34

, you know can be used obviously cautiously

21:38

, because that data is similar

21:42

to your advice around searching

21:44

on the interweb for general

21:46

knowledge . You want to question everything , but

21:49

I think there's ways to use tools

21:54

that can be really beneficial

21:56

just in day-to-day life , depending

21:59

on , again , what your actual problem is

22:01

. I don't know if there's specific use cases you've seen

22:03

lately that you want to touch on here but

22:05

I think there are really good use cases , again

22:07

dependent on the reliability

22:10

of the data and the specific use case . Obviously

22:12

, that is something we want to consider , but

22:14

there are different levels of quality

22:18

of data or the types of data needed , depending

22:20

on your specific use case .

22:22

Yeah , one use case that comes

22:24

to mind is actually

22:26

using computer vision

22:28

for reading invoices . This is a really

22:31

common business use case for

22:33

AI systems , whether it's

22:35

robotic process automation or , you

22:38

know , backend system

22:40

to system automations . You

22:42

know the quality of that

22:44

image can really dictate

22:46

a lot around what you

22:48

can do with it . As well as

22:50

helping people understand that

22:52

, you know AI systems are not

22:54

infallible . Like you need

22:56

to be able to have

22:59

your error handling and

23:01

recognizing . It's not going to work perfectly

23:03

every time , and so

23:06

part of your real world use case is

23:08

where do we insert humans and

23:10

this for everybody listening here . This

23:12

is a question you should always be asking

23:14

your AI team members and your AI vendors

23:17

. Where do we insert humans to

23:20

ensure the accuracy and quality

23:22

of the AI system ? And that

23:24

is so critical because the

23:26

one use case that I'm thinking of and it's

23:28

always my go-to when I talk about computer vision

23:31

was a project we were working on

23:33

where there

23:35

were truck drivers taking pictures of their invoices

23:38

and you know they would stretch the

23:40

paper out over the steering wheel and it'd be

23:42

all crumbled up and it'd have grease , stains

23:44

and stuff on it from like you know the work

23:46

that the team was doing , stains

23:49

and stuff on it from like you know the work that the team is doing and you know the computer

23:51

vision model . Sometimes it would be inaccurate

23:54

because you're dealing with a crumpled

23:56

up piece of paper being taken a picture

23:58

of in a truck and so

24:00

what you know we work to understand like

24:02

, okay , well , where are those inaccuracies

24:04

happening ? How can we work with

24:06

the computer systems to get better quality pictures ? How can , with the

24:08

computer systems to get better quality pictures ? How can we coach the

24:11

drivers to take better quality pictures

24:13

? And how do we get a human in the middle to

24:15

make sure that there's nothing bad slipping

24:17

through here ? And it

24:19

turned out to be a really great engagement point

24:21

to talk with the truck drivers . It also

24:23

was a great moment

24:26

for us to take and say , okay , well , we can take

24:28

better pictures . Just from a technology

24:31

standpoint , we can take higher quality scale images

24:33

and store those . So it

24:35

was a very engaging and

24:37

interesting conversation all around , but

24:40

it started out with pretty crumbled

24:42

up paper invoices .

24:46

So yeah , and

24:49

pretty fun project overall

24:51

and quality of images of invoices , but

25:09

then using AI to structure

25:11

that data in a way that

25:14

makes it more reliable to use for another

25:16

purpose . So , yeah

25:18

, I think leveraging AI in that way can be really useful

25:21

too .

25:21

Awesome , you

25:24

want to go into Katie's advice number

25:26

three here ? Let's do it . Let's

25:28

see what we have to say . All

25:31

right , here we go

25:33

.

25:35

Non-technical individuals should be aware

25:38

of the ethical implications surrounding

25:40

AI . This includes topics

25:42

like bias in algorithmic decision-making

25:45

, data privacy and the responsible

25:47

use of AI to ensure fairness , transparency

25:50

and accountability in its applications

25:52

. I call this ethical

25:54

considerations .

25:58

I think we could have a whole hour just

26:00

on that topic , but

26:02

Tali , where would you like to start on that one ?

26:04

No , absolutely . I think , clearly , ethics is

26:07

huge . I'm sure you know , with

26:09

the news nowadays , there's

26:11

been a lot of obviously use cases

26:13

of , you

26:16

know , ai implementations that have

26:18

gone really wrong and have had

26:20

really , you know

26:22

, poor side effects

26:24

that have really impacted people . So I

26:26

think understanding how

26:29

we can mitigate those risks is really critical

26:31

. I know , tim , in the past you've

26:33

used the example of the

26:37

experiment of using AI to

26:39

sift through resumes and

26:41

how that led to weeding

26:44

out candidates that would have been a great

26:46

potential match for a specific opportunity

26:49

but who were sifted out by

26:51

the algorithm because of biased

26:53

data . I don't know if you want to share that , but I think that's

26:56

a really good example of yeah

26:58

, just the ways in which things can go wrong if we're

27:00

not being considerate .

27:36

And then I guess , beyond that I don't know if you have a top

27:38

three things that not only an example of the ways in which

27:40

you know these situations can have detrimental implications

27:42

, but also ways to mitigate that risk . Yeah , there are so many , because

27:45

, keeping in mind , your data is going to

27:47

drive your AI system , and

27:51

so the historical nature of your data could create

27:54

lots of problems for your system . And so one of the steps

27:56

that I would encourage everyone to

27:59

as you all know , I'm an avid

28:01

reader . I talk about reading all kinds

28:03

of books . I feel like every event

28:05

or class or whatever I'm always like , add

28:08

this to your reading list Ethical

28:11

Machines by Reid Blackman

28:13

. It was recommended to me by Tom Loveland

28:15

, who's on the call here . Thank you , tom . That was an excellent

28:18

read and I would highly recommend it for

28:20

everybody To

28:22

really get a good ethical

28:24

mindset towards your AI

28:27

approach . You got to start in your company

28:29

culture and values . You know and

28:31

really understanding what are they , how

28:34

do they impact what you are choosing

28:36

to do with your AI systems

28:39

, because if

28:41

you have a culture , that's you know , win at any cost

28:43

and you know it doesn't

28:45

matter the human impact , you

28:47

know churn and burn , let's go get this

28:49

done . You're going to have a very different

28:52

compass from an

28:54

organization who is focused

28:56

on we are providing

28:58

healthcare solutions for this population

29:01

and we deeply care about this population

29:03

. We are passionate about this population or

29:06

school , you know , and so , looking

29:08

at you know we are here to grow these

29:11

students and grow their knowledge . These

29:13

are very different cultures that

29:16

I just described . So , making

29:19

sure that you're aware what is your cultural

29:21

compass and be

29:24

clear on that for your , for your AI

29:27

ethics programs . You

29:29

had mentioned tally transparency man

29:31

. I cannot hammer that enough , but

29:33

there is kind of a problem with that of

29:36

. The more advanced and sophisticated

29:38

an AI system gets , the less

29:40

transparent it can be , because

29:43

it doesn't think like humans , and so

29:45

it really gets really hard to explain

29:47

why some decisions were made

29:49

when the system is

29:51

not thinking like humans , not performing

29:53

like humans and would say well

29:55

, on this 200,000

29:59

dimension array , I

30:01

connected these dots and that's why

30:03

we got this . This output and

30:06

those dots are , you know , millions

30:08

of data points . You know it

30:10

gets complex real fast , so you

30:13

know . You

30:15

had also mentioned about a use case . You

30:17

know the hiring and resume use case

30:20

is a good one , but I'm going to pivot

30:22

on that a little bit . A recent use case

30:24

that I encountered was video

30:27

interviews . Use

30:32

case that I encountered was video interviews and so recruiters using video interviewing to screen

30:34

candidates . An AI system reviews the video , gives

30:37

the candidate a score and

30:39

progresses from there

30:41

. This , on the surface , is

30:43

like oh , this is a great for saving time

30:45

, because any hiring manager knows , man

30:47

, when you get into that interview slug

30:49

, it can be brutal , and

30:52

so I can get some help with making

30:54

this easier and faster . Awesome . But

30:57

you got to think about your training data

30:59

. If you trained

31:01

on interviews with neurotypical

31:04

people , they have very

31:06

different behavior patterns and very different

31:08

speech patterns and very different words that

31:10

they would use from a neurodiverse

31:12

community , and so if

31:15

the system is scoring with

31:17

the expectation of a neurotypical population

31:19

because you have data bias , well

31:22

, your neurodiverse candidates are going to get

31:24

penalized for that . So

31:26

really thinking through

31:28

how your historical

31:31

training data is going to impact

31:34

your use case is

31:36

so critical .

31:38

Interesting , yeah

31:41

, really good call out .

31:44

So yeah , so I can't emphasize

31:46

enough . There are wonderful activities

31:48

that you can use . There's the

31:50

in the AI garage session

31:52

that we did with Howard Community College , which

31:55

I see some of our graduates here you know

31:57

, hi guys . So the

31:59

ethics wheel , where you

32:01

take your AI solution and you plot

32:03

it against . You know what is our risk

32:06

for bias , what is our risk for impacting

32:09

human autonomy , what is our

32:11

historical data , what

32:15

is our behaviors that we're driving . It's

32:18

the ethics wheel which

32:20

we can share . We'll share as

32:22

a kind of a leave behind here for the

32:24

event . Plot your solution

32:26

there , ask those questions and go

32:28

in eyes wide open so that you can

32:30

say okay , we know we have these bias

32:33

items in our data . How

32:35

do we mitigate those items ? So

32:38

very useful tool , all right

32:41

. Do we want to go to item four

32:43

? Let's do it . Do we want to go to

32:45

item four ?

32:47

Let's do it .

32:49

Okay , here we go loading

32:51

it up . Let's see what Katie has

32:53

to say .

32:54

While non-technical individuals may not

32:56

require in-depth technical knowledge

32:58

, it's important for them to understand

33:01

the importance of collaboration with technical

33:03

experts . Building a cross-functional

33:06

team with individuals who possess both

33:08

business acumen and technical expertise

33:10

fosters effective communication and

33:12

ensures that AI initiatives align

33:15

with business goals .

33:19

I call this collaboration with

33:22

technical experts . Yes , please don't just go

33:24

do something on your own without IT knowing

33:26

about it . Like I can't emphasize

33:28

that enough , Tali , what do you think

33:30

about collaboration here ?

33:31

Yeah , no , I think I mean even beyond

33:33

AI collaboration is critical . Making

33:37

sure you're tying business objectives and strategy with the actual delivery or implementation

33:39

of any sort of system

33:42

or solution is obviously critical

33:45

. I know in the past we've

33:47

talked about the idea of an AI task

33:49

force and putting together a team that's

33:52

strategically in place to

33:54

move toward the business objectives

33:56

of an organization . So making

33:58

sure we understand the problems , like we touched on

34:01

earlier , but then also what's capable

34:03

, and making sure we're thinking about the ins and outs

34:05

of a certain technical solution with those more

34:08

technical folks . So really bringing everybody together

34:10

to have those conversations is

34:13

critical , and I don't know if you've anything

34:15

to add on from a , you know , ai strategy

34:17

perspective or having this AI task force

34:19

, but I think that that's yeah , that goes without

34:22

saying . That's , that's very important .

34:24

Yeah , absolutely Like . You

34:27

know the age of software as a

34:29

service , um , anybody

34:31

can set up any system with just a credit card

34:33

, and so you get instances

34:35

. Yeah , we've seen I don't even know

34:38

how many times at mind over machines where the

34:40

marketing team has a crm and

34:42

it's totally different than the sales team's

34:44

crm and it's totally different than

34:46

this other team's CRM . And

34:51

so you know , you have a credit card and this team likes HubSpot and that team

34:53

likes Salesforce and that team likes Dynamics

34:55

, and all of a sudden you have all these

34:58

silos and people are like IT

35:00

didn't know about two of those three , you

35:02

know , and so it's not

35:05

a fun conversation to have from

35:08

IT's perspective or the

35:10

business owner's perspective . Chatgpt

35:14

, man , that was such a

35:16

thing with people and still is Like

35:18

well , we can just use the free account

35:20

, it doesn't cost anything . And yeah , I'll

35:22

just load in . You know , here's this client

35:24

project . Can you help me , you know , define

35:27

the client project and communicate to the client

35:29

about this project . Well , now you're

35:31

giving all this information to a third party about

35:34

your client and so , just

35:36

being mindful of what

35:38

are the risks , and as

35:40

we all all of us we do our jobs . I

35:43

mean , we don't think about our jobs , we just want to get it

35:45

done , like , and we hustle to

35:47

get it done . We want to like OK , I got

35:49

to get this knocked out so I can move on to the next thing

35:51

and next thing . And

35:57

then we start making mistakes , like telling chat GPT things that it shouldn't know about

35:59

private company information . So , really working as a team and

36:01

that team IT , legal

36:04

HR that is what

36:06

we see as common , you know , from a strategy

36:08

perspective of where to start Get your IT

36:11

red team , your forward team

36:13

, whatever you want to call them . These

36:15

are the task force , these are the people

36:17

to get engaged . Hr to help you with your

36:19

human problems . It

36:21

to help you with your security risk posture , as well

36:24

as your technology solutions and consistency

36:26

and technology solutions and then

36:28

legal , from a governance and privacy

36:30

perspective , to make sure you're

36:33

not stepping on any toes that you shouldn't

36:35

be stepping on Makes

36:37

sense

36:39

. All right , let's go to the last

36:41

thing that Katie has to say , and

36:47

this one

36:49

I can't agree with more ai

36:52

is a rapidly evolving field and

36:54

keeping

36:55

up with the latest advancements and trends is

36:57

crucial . Non-technical individuals

37:00

should embrace a mindset of continuous

37:02

learning , staying informed about new

37:04

developments , attending relevant workshops

37:06

or webinars , and and engaging with AI

37:09

communities to stay updated on

37:11

best practices and emerging opportunities

37:13

. I call this continuous

37:15

learning .

37:19

Yeah , tali .

37:22

Like you said . I mean , this is huge , obviously

37:24

beyond AI , I think you know we should all strive

37:27

to continuously , especially in the

37:29

tech field , learn what's possible

37:32

. But I think with AI in particular because it's

37:34

transforming honestly

37:36

, it feels like daily we really

37:38

want to make sure we stay on top of the

37:41

latest and greatest what's possible . What

37:43

are the risks that we can learn from ? Yeah

37:47

, it's really pertinent

37:50

that that's something we all do , especially

37:53

just to mitigate risk and , to

37:55

you know , increase

37:57

any potential revenue

38:00

or efficiency within your organization . The only way

38:02

to do that is to really stay ahead of the game here

38:04

, and I know that's something you

38:07

and I , just from an standpoint , try

38:09

to stay involved in . I don't know if you

38:11

have any tips , practical

38:13

tips in terms of how to best go about that for

38:16

the audience , but I think that's a really

38:18

critical piece here , absolutely

38:20

.

38:22

Yeah , linkedin , linkedin , linkedin

38:24

, follow the people

38:26

who are keeping

38:28

track of what's going on , but don't

38:30

try to follow everybody . You

38:33

know , like , pick out who are the people in your industry or that are focused

38:36

on your topics , and

38:38

how do you find those people ? You look for them , you

38:40

know , and so I would encourage

38:42

everybody and

38:45

I might say his name wrong , but Jan LeCun

38:47

, who is the head of Meta's

38:50

AI research group . He's

38:52

a great person to see and follow

38:54

because he comments

38:56

on a lot of other people's projects

38:58

. The team at Hugging

39:00

Face is another group to

39:02

keep an eye on and see what they're talking

39:04

about . And again , narrow and

39:06

focus . Don't try to follow everything

39:09

. There's too much going on to follow

39:11

everything . Focus

39:13

, don't try to follow everything . There's too much going

39:15

on to follow everything . That's

39:17

what Tali and I are here for is to be insane and follow everything

39:19

. So , but yeah , like , focus on your industry or

39:22

an industry you're interested in

39:24

. You know I have

39:27

no ambitions of working in the video game

39:30

or entertainment , like gaming industry , but

39:32

I keep a close eye on what's happening there because

39:34

that's kind of a leading indicator

39:37

of what we'll see in the commercial space

39:39

. So , you know , find

39:41

your industries , keep a close eye

39:43

on them , so .

39:46

Definitely .

39:47

Cool , all right , okay

39:49

, so from a , that's

39:52

all Katie had to say about the topic

39:54

and you know , kind of in summarization

39:56

there , what

39:58

I would really encourage

40:00

you all to take away from this is you know

40:03

AI literacy is a continuous

40:05

thing , you know . Know the basics

40:07

, which are . You know your use

40:09

cases for your industry . Dig into

40:11

those use cases , dig into the problems

40:14

. Understand

40:16

that data is key to your

40:18

success . Understand that

40:20

machine learning is how you're going to harness

40:22

that data and work with it , and there's lots

40:24

of different ways you can dig deeper into machine

40:27

learning . If you would like us to

40:29

dig deeper into machine learning

40:31

in a future boring AI

40:33

show , let us know in the comments . We

40:35

really want to make this useful

40:38

for everybody here . So you

40:40

guys tell us , what do you want to see us talking

40:42

about and working on ? As

40:44

far as the continuous learning , we

40:47

think that upskilling is critical

40:49

. You know that's a key component Mind

40:51

Over Machines has been working on literally for years

40:54

, and so we're going to have

40:56

for the next Boring AI Show

40:58

. We've been working

41:00

with a guest to come in and talk about upskilling

41:03

and how we are building out a

41:05

AI education program for

41:08

high school students in Baltimore

41:10

County , so very excited about that

41:12

episode coming up in two weeks

41:14

and Tali , anything you

41:16

want to add to wrap up our discussion with Katie

41:18

.

41:19

Yeah , I think also just really empowering

41:21

your workforce , not only to feel confident

41:23

in you know using these tools in ethical

41:25

ways , but also to feel confident enough to

41:28

speak up and ask questions and

41:30

be skeptical and , hopefully

41:34

, you know , highlight potential risks that maybe others aren't

41:36

seeing . I think if you're in a space where folks feel

41:38

confident to ask , I

41:41

think that's really what can help

41:43

organizations

41:45

adopt a more , yeah , ethical

41:48

, responsible AI strategy

41:50

. So definitely , you know , doing what you

41:52

can to empower your workforce in that way is

41:54

huge .

41:55

Totally , totally , all

41:58

right . Well , the next part

42:00

of the Boring Age iShow is Tali

42:02

and I both brought a win

42:05

that we've seen out in the AI world

42:07

and we're going to chat about and

42:09

share with you guys . If

42:11

you have a win that you want us to

42:13

talk about and explore , or news , just make

42:16

sure to share it in the comments section . We'll pick it up

42:18

and we'll go from

42:20

there . But , tali , do you want to kick us off ? What's your

42:22

win this week ?

42:23

Sure , yeah . So I think you know

42:25

and really the reason we're going for wins is I think there's

42:27

a lot of news and hype

42:29

around the Doomsday AI

42:32

kind

42:34

of messaging and I'm sure folks have heard

42:36

about the letter that

42:39

many folks signed

42:41

around

42:43

halting , pausing all AI development

42:45

for

42:48

risk of kind of a doomsday scenario

42:50

. Well , in brighter news , or when

42:52

that came out recently , bbc News

42:54

published an article on the 18th of July

42:57

showing that , more than you

42:59

know , 1300

43:01

experts you know

43:03

, calling AI a force for good . So

43:06

this was kind of in response

43:09

to that doom letter . Doomsday letter which I said

43:11

you know came out in March . That included

43:13

tech leaders like Elon Musk , who actually ironically

43:15

just launched an AI business of his own

43:17

, which is something to just ponder

43:20

, called the XAI . But

43:22

this article

43:24

really focused on the positives

43:26

of AI , you know . So BCS

43:29

, the British Computer Society , organized

43:31

an open letter countering that

43:33

AI doom letter that

43:35

many experts signed , saying that AI

43:38

is a force for good

43:40

. Bcs really sees the situation in a

43:42

more positive light . Obviously it's still supporting

43:44

the need for rules and regulations around AI

43:47

, which I think is critical , and focusing

43:49

on AI ethics

43:51

. But

43:57

it really shows and it highlighted a couple really positive

43:59

use cases from some of the signers that AI could help human and society

44:01

advancement . So some of the ones they called out were

44:04

you know , ai can enable

44:06

healthcare professionals to spot really

44:09

serious illnesses faster , you know , potentially

44:11

saving lives . It could accelerate the testing

44:13

of new drugs . Another

44:16

use case was AI could

44:18

identify and target weeds

44:21

across you know fields of crops , which

44:23

could prevent the need for mass spraying

44:25

of harmful chemicals . So there's a lot

44:27

of really great and incredibly

44:29

positive use cases for AI

44:32

and obviously

44:34

we want to caution that with . You

44:37

know the need for regulation and

44:39

thoughtful implementation , but

44:41

I think we really need to understand that

44:43

there are such positive use

44:45

cases for AI that folks are using

44:47

and as AI continues

44:49

to advance you know it's

44:51

not stopping any time that I see so we

44:53

just want to make sure that we're researching and

44:56

staying aware of what's happening and

44:59

I think that this was just a really cool way to highlight

45:02

yeah .

45:03

AI for good .

45:04

I don't know if you have any thoughts on that Tim ?

45:07

I do actually . So that's awesome

45:09

, as

45:11

you were sharing that um , I actually have

45:13

not seen this um . So in our

45:16

pre-show uh discussion , tally

45:18

and I were she was like oh , you've probably seen this , I

45:20

haven't seen that . That's awesome . Um , there's

45:22

a community called ai for good

45:24

. Um is all

45:27

about like okay , here are use cases of

45:30

AI to support the

45:32

United Nations Sustainable Development Goals

45:34

, and so that is

45:37

super awesome to kind of hear this and

45:39

for everybody listening , like I

45:41

really encourage you to go and see

45:43

what's happening in the AI for Good community

45:45

. It is some amazing

45:48

use cases , amazing use cases . So

45:50

you know , really , really

45:53

, really fascinating stuff that

45:56

I think you guys would . You

45:58

know , really , really see yourself

46:00

in and see the potential as

46:02

well as see things that are like holy

46:05

crap , that's amazing and feel good

46:07

about it . It's not just all doom and gloom , so

46:09

awesome

46:12

.

46:14

That's a cool win . Yeah , it's exciting to see I'll

46:16

share the article .

46:17

Awesome , uh , so just post it in the comments here for people to

46:19

pick up . Absolutely , um , and

46:21

then , uh , my win for

46:24

today um , I love , I love

46:26

toys , I love the toys . Uh

46:28

, in the technology world , and

46:30

I was going to talk about

46:32

something called chat

46:35

USD , which USD

46:37

is a 3D file format that

46:40

you can , but it's text-based

46:42

, and so what the chat USD

46:44

allows you to do is like , build out a 3D

46:46

model of a car and then , using a chat

46:49

interface , you can say , okay , make the car green , make the car red

46:51

, change the tires to look like this chat interface . You can say , okay , make the car green , make the car

46:53

red , change the tires to look like this , and

46:55

, and you can basically use text

46:57

to 3d objects , which

47:00

I was like oh , that's so cool until this

47:02

morning when I saw a very

47:04

awesome demo of kadem

47:07

is the technology . All

47:09

of this is through nvidia ai , which

47:12

, again , if you want to see some of the really

47:15

powerful use cases in the world of AI , NVIDIA

47:18

AI is a great group to follow on LinkedIn

47:20

. So

47:22

the KDEM technology it

47:24

takes a 2D image and

47:27

extrapolates a 3D object

47:29

in from that image . Except

47:32

if you've seen this before , it's kind of like , eh

47:34

, that's okay , it's not all that great

47:36

. No , this KTM system . It was

47:38

amazing to see what was coming

47:40

out of this . So big win for

47:43

2D to

47:45

3D as well as text to

47:47

3D . I think these are really

47:49

cool technologies for the world

47:51

of construction , manufacturing

47:54

, wind

47:56

. You could be using this to rapidly

47:58

prototype different modules

48:02

for wind turbines and

48:05

explore what's breaking . Why is it breaking

48:07

? How do we make this better ? Super exciting

48:10

stuff and all ties into NVIDIA's Omniverse

48:12

, of course , which another

48:14

thing , if you're not keeping an eye on the

48:16

Omniverse . Pretty interesting

48:19

thing if you're really into being a nerd

48:21

about 3D technology and AI

48:23

.

48:25

Very cool , huh , interesting

48:27

.

48:29

So cool . Well

48:31

, that is all we had for you guys

48:33

today . You know

48:35

the next , the next

48:37

boring episode we have it currently

48:40

planned for August 24th

48:42

and our

48:45

guest right now we

48:47

want to get final confirmation before announcing

48:49

, but we'll be talking about AI upskilling

48:52

and the importance of upskilling

48:54

, as well as how to get started in your upskilling

48:56

journey , so we

48:59

hope to see you there . Please like

49:02

, comment , share . Tell us what you'd

49:04

like to hear . I've already been getting messages

49:06

of people saying can you do an episode on

49:08

this ? Can we talk about that ? Can we go

49:10

from it ? Yeah , absolutely Like . As

49:13

long as this is useful for you all , we'll

49:15

keep going . So tell

49:17

us what you'd like to hear and let's build it

49:19

out , and please feel free

49:21

to make this a conversation . We'll

49:24

meet with you guys on the 24th . Tali

49:26

. Anything I missed today ?

49:28

No , this was super fun . Excited

49:33

to meet again in the upcoming few weeks and read all your comments that you guys

49:35

have following the show . So thank you so much . This

49:38

is awesome .

49:38

All right , thank you , everybody . Talk to you later .

49:41

See you guys .