Subscribe
Share
Share
Embed
Join us on the Disruption Now podcast as we challenge the status quo and advocate for digital equity, ownership, and responsible technology.
doing the right thing it's this moral compass one has
of knowing where you are headed
your North Pole
you can hide behind this cloak of opacity that says
you know if I don't understand
if nobody can understand what I'm doing
I can get away with it as long as I make a buck
we uh break everything down into rules
which can be easily read and understood
and those rules that level of transparency
in a way exposes you for good or for bad
but if you start breaking bones
and you do it in a way without an IRB process
Silicon Valley is using human beings as test
Guinea pigs
she did not opt or sign a waiver on her life correct
you claim you're a safe driver
prove it
when technology really matters and it is important
and you want to scale that up
you have to have that element of trust
AI would need to be insured
irresponsible behavior is not a good policy
for return of investment it is bad business
everybody has access to the same tools
yes now you ask yourself
what would make you a differentiator
that's it
a human has to be accountable for the decision yes
welcome to disruption now
I'm your host and moderator Rob Richardson
disruptors always love to have you on
but before we get started with our guest
and I promise you you're gonna wanna listen in
please make sure you like and subscribe
if you're watching us here on YouTube
if you're listening to us online
go ahead and subscribe as well
so you can know when the next disruption comes out
we need to spread the words about
spread the word about how we are empowering the future
and making sure that technology is accessible to all
and so with me today is a mentor of mine
a friend of mine Doctor Kelly Cohen
he is a researcher he is a entrepreneur
he he has been leading
leading an AI for a long time before it became
became the thing to do he's been doing it for decades
he invented a concept called fuzzy logic um
so he's not only smart uh
he's kind I'm
I'm proud to call him a friend
and more than anything he is a teacher and a mentor
he's helped so many students
help so many people and he's opened so many doors
and so
it's an honor to have Dr Cohen on here on this show
and we're gonna be able to really talk about uh
explainable AI and why that should matter to you
and I promise
I promise you stick around because it does matter
Doctor Cohen welcome to the show brother
thank you good to see you
Rob and for hosting me and for having me here
and for all the kind words
so
Fuzzy's been around since it was invented by a lot of
fizzled in Berkeley in 1965
before you get to fuzzy we'll go down that rabbit hole
yes but you have a
ultimate since I've known you
you've had a sense of responsibility and a sense of
this is not just about me
where did you get that like
who shaped your values the most
and how does that show up in your work
so my parents my teachers
my background I believe that has shaped
I grew up in an educational institution
and my aunt was principal
my father was vice principal
it was a family business in a way
the school but it was one of the best schools in Mumbai
India
and I literally grew up on the premises of the school
for 17 years I was there yeah
until I graduated from high school
and so uh
I called my father sir
most of the time because he was
I saw him more in school than privately at home and uh
I got a lot from him he died about uh
you know four years ago
you said you got a lot from him
what's one lesson you can remember from your father
that sticks out doing the right thing
it's this moral compass one has
of knowing where you are headed
your North Pole and you've known me for a while
you've known that
there is a direction in which I follow
I'm very consistent with that and looking at the impact
so whatever you do
you make sure you leave behind a positive footprint
in the hearts of people yes
well you've definitely done that
so let's talk a little bit about your work right
and I think it's people see AI and there's a lot of
there's a trepidation
there's some fear and I think most of all
a lot of people don't know what they can
what they can trust with AI
what they see and of course
we're gonna talk more about this
but you talk a lot about explainable AI
and why that's important
and behind your work with fuzzy logic
if you if you can without getting too technical
what what is explainability when it comes to AI
and why should that matter to any
to any of us right now like when you when
when people say explainability and it's important
and some people say like well
maybe it's not why do you say it's important
and what is explainability
so there is a lot of grey areas in business
people feel that the goal make money
that you could somehow
not get into what is right and what is wrong
as long as you pursue your goal
and then you can hide behind this cloak of opacity
that says you know
if I don't understand
if nobody can understand what I'm doing
I can get away with it as long as I make a buck right
now what we do is we provide one with a
uh
transparency
to the extent that if you're doing something unethical
it is open to the world to see we uh
break everything down into rules
which can be easily read and understood
and those rules that level of transparency
in a way exposes you for good or for bad
and then if you have a set of values
because very often companies
big ones feel that they have a set of values
that they promote on their website
but you can see their actions
don't always align with their set of values right
yeah happens a lot
and so
you can have this big charter about responsible AI
but what they're doing behind the scenes
is very irresponsible yes
and the reason for that is lack of transparency
so when you are transparent
when you speak openly as to why the decisions are made
from my background I was told to do things
my father was an educator
he explained to me
why it is that you have to behave in a certain manner
why is it that your decisions
need to be in a certain manner
it is the path of righteousness
it's the path of doing what is good for your community
doing what is good and so you empower your students
so in my class which I teach every spring
I dedicate the first four lectures
not on the how to mechanics
of getting something done with machine learning
I dedicate that to talking about what it is to be
responsible with your AI yes
I talk about the essence of responsibility
looking into the impact and then giving them examples
and you've seen some of those examples
that I bring to bear with AI
but we haven't seen it
so you can talk to some of the listeners about it
yeah you know
there are some cities that
feel that it is okay to go fast and break things
yes and that's the meta model
and they're still breaking things to this day
breaking things is one thing
but when you start breaking bones
and you do it in a way without an IRB process
what's an IRB process is what we do here
when we look into the safeguard of rats
when we do a biomedical test on them right
so at the university of Cincinnati
you want to do a test on animals
let alone people you have to go through a process
a process that looks into the ethics of things
are you behaving in the right manner
what are you doing
what would be the impact of your experiment
even if it is a
asking people their opinion on certain matters
you have to go through a process
so what I hear you saying is essentially right now
Silicon Valley is using human beings as test
Guinea pigs and Guinea pigs
and and the best example is GM's where a cruise
self driving taxis
where they ran over a woman and sent her to hospital
now she is a pedestrian
she did not opt or sign a waiver on her life correct
she did not say OK
I'm in the city and because I'm in the city
I'm OK with right
people experimenting with me with these uh autonomes uh
cars moving around
so you can elect to drive in an autonomous car
you can elect to buy one
but you can't necessarily provide that to a pedestrian
walking on the streets he has not given you a waiver
I would like to and I
and I and I hear you on that point
and you know
as this disruption we like to hear all sides
and just just so we have a
a conversation I'm sure that like
you know Waymo and others
the technology behind that would say that well
our accidents are are
are substantially less than a human being behind
the wheel and that we're at the point where we can show
that is the whole thing they don't have proper testing
to prove to the public that they are safer
they have an agenda yes
they've been hiding their data
they've not been very transparent
yes about what actually happens
and they haven't gone through a proper certification
process I'm a driver yes
I wanna drive a car my car can impact and hurt people
I have to go through I think
a license I have to get one
I have to go through a driving test okay
can you imagine somebody who just uh
wakes up from the jungle one day
and comes into the city and says
I want to drive a car without a license
no
you're going to jail especially if you hurt somebody
it appears that Cruz was lying to GM
and so they shut Cruz down as a result
and then the city invoked the license
so there was something wrong
right that was corrected
but the question is in the first place
why were they allowed to on what basis
so I go and drive on the streets
I first get a license then I'm allowed to drive right
and same with the autonomous cars
you you're safer
you claim you're a safe driver
prove it that's fair
I think it's fair and that's why I have a problem
I'm not against AI I believe in AI
obviously
one of my AI products is now in the market for aircraft
is it safe yes
why because it is certifiable
certifiable is going through that so called testing
and we advocate for extensive testing
providing the user
providing the community with the proof that says
not only will we develop to do a certain job
but we've gone through testing
now what does that take
that takes time and money
yes that takes being cautious
but the outcome is not no progress
the outcome is progress right
in a responsible manner and that's what we want to see
responsible progress
so Doctor Cohen is the leader of the
let's say you got a chance to set policy for let's just
we'll take it in a a
a micro example of autonomous vehicles
but it could apply to algorithms across the board
what would that look like as a high level basis
you have a certification process
that they would go through
and then be approved for using the actual algorithm
is that what would that look like practically
we demonstrate tests are
that prove that the risks are below certain thresholds
so every product has uncertainties
uncertainties in the way it operates environment
so if it's a car you have uncertainties in the weather
you have uncertainty in the quality of the roads
the ice on the roads you got uncertain to be the
how the passengers are an old woman with a stick
you can't always guarantee that
each and every future
scenario would be part of your learning data
so in reality you need to adapt
and that's where a lot of the systems
that are traditionally so called AI fail right
and AI can do a few things
quite a few and quite well yes
but they can't do everything
and so we are rushing to take immature technology
whose foundations have these flaws in it
and we try to extend that under the guise of oh
we've got to work fast and break things right right
where do we draw a line
in our responsibility towards the human
how do we balance that out when we're in a global
competition state with China
that is also as they have even less regard
because they don't even have a democracy
they can move when they decide
and I'm sure you probably have some agreement in this
that there's a lot of power for those who
are able to develop AI and essentially
you know reach a d I or something like that
so how do you balance that question
let's take a look at my industry
which is the aerospace industry it is highly regulated
do you agree absolutely okay
nobody would fly an aircraft
which is not certified for airworthiness
how many Chinese aircraft are being sold
around the world versus American aircraft
who's regulated we are regulated
America
has that come in the way of us being world leader in
aerospace definitely not
so it goes together why do airlines prefer us to China
cause they trust us yes
they trust our certification process
and they know that if they want their passengers
to move from point a to point B safely
you buy an American product and
and that's the way or in Airbus
which is European
but we have the same safety standards
and so it is not the fact that when it matters
when technology really matters
and it is important and you want to scale that up
you have to have that element of trust
if you don't have that element of trust
nobody's gonna buy us stuff and
and if we look at
what's happening with the aerospace industry
with the AI is gonna come once you start getting um
uh products that are not tested that are not safe
people will say I'm not gonna use this right
no
the government can either wait for uh
20,000 people to die before they regulated
how do you think uh
it happened with the aerospace industry
19 0 3 Kitty Hawk new North Carolina
they took this bicycle shop from Dayton
right brothers yes
and they flew for the first time it was 20 years
it was the Wild West everybody flew
but 20 years of flying and crashing
created the need for regulations of aerospace
so it took time but once we started the regulations
we didn't look back
and then the progress was coupled with regulations
okay now we have our outliers
like what happened with the 7 3
7 Max and Boeing and all that but on the whole
the record is wonderful that's why
the underwriter doesn't mind insuring your life
when you take an aircraft flight right right
yeah um
we would reach something similar
does that answer your question because no no
it answers my it answers the question well
so you've we've talked a lot about autonomous vehicles
aircraft
I feel like that the listeners can understand the
the gravity of those situations but
but let's let's bring it down to a little more uh
specific level
like dealing with these chatbots that give advice
that can sometimes get
into counseling that can tell our kids and adults
you know
it could it could help I think even form more biases
it can give wrong results
how does one navigate themselves individually
when there is no regulation right now
that's an excellent question
there comes the insurability of the AI
if you're a chatbot take Canada Air
so they want to save money yeah
people are expensive let's use a chatbot
I get a license for
for 15 dollars a month or enterprise level is 20
25 bucks a month very cheap
I saved so many extra number of dollars
so they employed a chatbot for this
a Canada Air it gave it hallucinated
gave a customer a passenger wrongful advice
the passenger took a snapshot of the chatbot
and then sued Canada Air what do you think it won yeah
the refund because there is no such disclaimer saying
hey don't trust me
I bullshit right so if you don't have that disclaimer
you win your cases
now what would you want as an alternative
I have a patent by the way
for a anti hallucination filter
with Pete Blackshaw's brand
rank as chief scientific advisor there
and we got
funding from the state to develop this filter
so the filter would basically be a add on to a chatbot
and it would guarantee in a way
and it would guarantee as in if I make an error to me
because I've got insurance coverage
so when I go to the market
the idea is to save your brand
you'd go to an insurable AI chatbot because interesting
I can get a guarantee because it's
it's like uh
my car getting stolen what do I do
cars get stolen they get
but if there is a balance between my premium
and the fact that the insurance company
wants to make money and they look at the statistics
and they figure out the risks as such
that they still make money
even though cars are being stolen from time to time
and broken into and they would insure your car for you
right so and that is
you know not necessarily life or death
it's just your car insurance
there is accident insurance
but I'm keeping that aside right now
I'm talking about you know
your car getting stolen your house getting broken into
right we have insurance for a wide range of things
AI would need to be insured
now that is going to be the differentiator
which AI is insurable and not
and I don't know if you've heard of this concept
insurable AI as often no
I haven't
this is the first time I've been hearing about it yeah
there you go
you heard it here insurable AI Doctor Cohen there
I feel that you know
so we have standards for airworthiness
when it comes to insurable AI
a clever person after being hit
now Canada has been hit right right
they were shamed around the world
their brand was hurt and they
whatever savings they had on the chatbot
I'm sure they lost most of it because of uh
you know the story going out
that they were irresponsible in the
using a chatbot that hallucinated
now if you go back to them and you say look
there are 10 products
but in this product which is insurable
you pay a small premium
but we can guarantee you if a mess happens hey
we'll take care of the customer
we'll fund him make sure that he is you know
yeah he's okay
and then the insurance company works out
if you pay me insurance
for so many use cases and for so many customers
in the end
the hallucination rate of this product is so low that
you know I'll still make money on it right
and that's how what we would do
so we run these millions of scenarios in a lab
literally millions
they're called Monte Carlo simulations
where we look at all the uncertainties
and what we're doing with the um
brand ranks their product
which is the anti hallucination filter
we're running these Monte Carlo simulations
where we're looking at you're a customer
there are so many variations
in which you can rewrite a prompt
we want to guarantee that after writing 10,000 prompts
all of them lead to the right answer
yes right because that's what happens
we different people would interact with the check
so it's insurable is it an audit too
what what would what
what would happen is
are you just insured that because you have
your product is already gone through the audit
we test it okay with a wide range of simulations
so it's testing right
and we demonstrate the risks associated with it
so we cover all the usable you know
now a chatbot for an airline shouldn't answer
a question on fashion design
it shouldn't answer a question on the weather
so the question
the only questions you need to answer are
relevant ones yes
now within the field of relevant questions
you have to ensure
that the answer is in your rule book right
so uh any uh
person in a behind a chatbot
if it's a human also he has to follow the rules
he can't talk about a refund
just because he feels like being nice to the customer
he's got a set of rules we ensure
with the anti hallucination filter
that every answer is guaranteed
with the letters of what is in the rule book
that makes sense so I get it from that
from that perspective of being able to
ensure the work product of the AI chatbot
or algorithm that you have
but let's
let's walk it through just a more pragmatic scenario
you're meeting with a CEO of a company right
and they wanna implement I
and they're trying to figure out
how to just implement it as
as effectively as possible
what's the first thing you advise them to do
now we have to get into the details
what are they looking for
what type of decision making
what is the impact of the decision
whether
the decision would cause people to come in harm's way
like with autonomous driving
or whether it would cause a liability
in terms of money right
so we look at the impact and based on the impact
we'll come up with a procedure for development
and a procedure for testing
something in the end would demonstrate risks
and those demonstrated risks could create a package for
the underwriter that will be able to say
I can if you meet all these capabilities
and you demonstrate it in test
and you're transparent about the type of testing
you have the good
the bad and the ugly right
we will then insure your product yes okay
because that's how insurance companies make money
now those insurance companies today
nobody's gonna say
AI is not gonna be part of our lives
so let me not do AI it's like a company uh
who said
I don't want to deal with aircraft exactly for travel
I'll just stick with my uh
uh horse driven carriages or I'll stick with my trains
so let's so let me
let me clarify the question a little bit
you're a leader right I get the
the route of we're doing insurance and all that
but let's not take it to that specific level
let's go a little bit higher if people are saying
I we're gonna implement AI and transform AI right
and let's just I'll try to give an example of a exact
let's go with a hospital right
a hospital wants to figure out how they can use AI
to do better billing and medical billing be
be more efficient I've seen that use case a lot
what are the three questions that the leaders must ask
before they even think about implementing AI
what are the risks associated with with a hallucination
how often do you hallucinate
and what is the potential impact
so what you do is
you run 100,000 scenarios with the current system
and then you'd get a result saying okay
there's a chance you'd lose X number of dollars
and you'll be liability
and you'll hurt your brand and so on
and so we would go ahead and say
what is the alternative
the alternative is to create an insurable AI
you pay a little extra you develop it
you test in a certain manner
and now you'll get a better return for your investment
without having to be that liable
in the future
we would translate that into a detailed program
where somebody
would be able to look at the facts and make
data driven decisions now a CEO might say
you know what
as long as in the next two weeks there is no problem
I'm okay but if you have like a Germans style
you know and and you've seen that happen yes
where people rush to get a product
which is not mature to the market
and then whether they say oh
I'm sorry I should have tested it more often yeah
now you ask yourself
why couldn't Google do the enough testing
well they lacked money to test their own product
yeah of course not
they make poor decisions about
the maturity of the product
and when to enter the market
why people are ding ding
ding ding
counting the money oh
we need to we have that race with this organization
or with this country that's right
we need to go forward but in the end
irresponsible behavior is not a good policy
for return of investment it is bad business
and we've seen that happen time and again
yeah I'm curious
do you feel like any of the large language models have
getting it right or they all just on
on the race to who could uh
innovate the most
if I want to create a little greeting card for Hanukkah
and personalize it I would use uh
you know a large language model
is there anyone there are telltale signs
but I don't care is there anyone better than the other
is anthropic better
in terms of ethics and responsibility transparency
I I did a little test OK
I asked Entropic Claude Perplexity
co pilot and chat GPT1 question
and they got it all right
I said of all the academic research groups in the world
which is the No. 1 on fuzzy based
aerospace critical systems
what do you think
they said Kelly Corn you know and so I said
oh they've
they've they've come along a long way
and then I asked them for justification
and then they justify based on actual publications
leadership roles in the world of uh
products that have come out
and then they compare you with others
and they say what your strengths and weaknesses are
I look at all the evidence and I said yeah
this is making sense now
US
News and World Report doesn't give you ranking per lab
right they give you ranking per institution
and so while as a college
for example we're happy being No. 87 okay
but you know you go far bragging about it
you're No. 87 in the United States
No. 300 in the world
but my lab number one in the world in what I do yes
I don't claim I do anything other than fuzzy
I'm that cat that has a thousand dreams
but they're all about mice
my dreams about fuzzy
that's what I do you want something not fuzzy
go to somebody else right
you want to come to me and graduate with me
I've got 22 A 20 two grad students right now in my life
all of them are doing fuzzy
if they don't want to do fuzzy
go somewhere else well
let's talk about teaching in the age of AI
there are opportunities and there are challenges
how do you see the role of teaching changing or
or how does it need to change
and what should never change
having to be bold enough to say that your AI
is the best in the world
you have to understand what's happening in there
to understand what's happening in there
you've got to develop these AI systems
in my class from scratch
now once you know how to develop them
so the other day I was
it takes me about three days to develop a good
fuzzy system for a new use case
and I was asked to do something
for one of the companies I'm working with
and so I had only two hours
I got up at 4:30
I had to come up with something by 6:30
so I used Chat GPT to build a model
it took about 3 4 iterations yes
but I could debug it to get what I wanted and tested
because I know the inner workings of the system
so
it is okay to use
a large language model to help you develop code
but you need to know how to debug it
you need to know and to know how to debug it
that's what I teach them
so I'm not against saving time or shortcuts
but you have to master
the work in order to be able to develop
on steroids good
uh
you know uh outcomes yeah
makes sense I mean
you it's like you can be great at basketball
it doesn't matter if you're able to shoot a 3
if you haven't mastered the mechanics of actually
how to take the shot because
you know you might make a 3 every now and then
but you're gonna miss more than you make
and you're gonna miss it under
under extreme conditions the same thing with AI right
it could help you come up with the right answer
but how do you know it's the right answer
how do you evaluate it how do you debug it
how do you how do you come to the point of knowing uh
when you need to iterate when you need to add more
you need to know the principles
there's no there's no getting around of
and you need to know what way to get into the system
and the tools you're using in order to make changes
so what's changed with with
what's changed with teaching when it comes to AI now
the fundamentals haven't changed
having to understand how it works
why it works what it is good for
how do you scale that up
that if I look at my curriculum no
but I don't go against saying hey
don't use chat GPT for your final project
you've got to be careful and honest
about how you've used it
you don't do it for writing papers
because conferences have a policy against it
you shouldn't go against policy you shouldn't lie
but if it if you're like there are some
supposing I have a concept that I want to explain
and it helps me instead of working out all the details
of the articulation and the graphics
I write a prompt to chat GPT
help me create a slide for my class
and this is what I want the slide to look like
now I know I've been working in this area for 31 years
so I know more or less what I want to bring about
so I give that description
and then it helps me work out the
the details in the slide
and then in the bottom I attribute credit
I said this slide was developed based on the
by chat GBT
based on a prompt developed by Doctor Kelly Cohen true
yeah it is true and I'm not ashamed about it
does that make me a more effective teacher
because I communicate better using the tool definitely
absolutely and so I'm
I'm I'm good
because any tool like in the past when I was in school
we had to use a log table to do calculations
instead of a calculator calculator was god forbid
going to create uh idiots of all our students right
so in high school I use log tables to do simple
multiplication and division yes
because when you do a log
a multiplication is an addition
which is simpler
then a multiplication and a division is a subtraction
and so we would do the log
do the calculation and do the anti log
the same with trigonometric functions
it was simpler to use that
so we were very good at using it
and that was what we did till the grade 12 yes
no calculators afterwards
when I went to the university
they introduced calculators
there was no internet then okay
so we've made progress
and every time there was a step in technology
there is the question of adapting to it yes
but now nobody would want to go back to lock no
absolutely not it's absolutely stupid
the amount of time I spent
and then there is more errors there
because you look at the wrong low row
you know you make an error yeah
now in a calculator
also can make an error if you hit on the wrong
the tabs right
so yep so I
I would say I
I would say
there's gonna be some fundamentals remain the same
I feel like there's gonna be
there are some differences
like in being in the field a lot
and working with the workforce quite a bit
you know there will be differences and I
I think we'll have to challenge
some of our assumptions like
I agree completely with your concept
that you need to be able to be honest
and show your work
you need to be able to describe how you debugged
working with AI what alternative approaches you took
what were and and you be
you should be able to show your trail of work
and you should be able to explain your thought process
your work to another human in an individual
because if you can't explain it
you didn't do it AI did that's how I feel
No. 2
how we process information is gonna change completely
like we have to kind of challenge the notion
and the funny thing is I use uh LLM
which is non transparent
to develop transparent AI fuzzy models
cause you can do that yeah
because you want to
and then once I see the transparency
it is easy for me to pick the errors
so by the end of it
I can certify that AI is something correct
transparent and good correct
but I got LLM to save me time to build that system well
and not only that you I say this often
we we we dance with the algorithm
so we have to
how we retrieve information is gonna be challenged too
is it necessary for us to memorize the level of things
we've had to I would say no
that doesn't that doesn't mean we stop using our brain
we use our brain differently like human beings
sapiens used to be able to go to an environment
a wild environment and be able to smell a lot stronger
and know predators coming from a distance
we would know how to memorize where the stars were
and orient ourselves like very few people
have those skill sets now right
so we'll have different types of knowledge
I think in terms of we don't
we won't necessarily be holders of information
but we will be have to know how to critically think
and connect many points and actually think more
so I tell people
if AI is not making you think and do more
you're using it wrong it is not a Google default
it is a and it is a way to help you
what also happened Rob
is that everybody has the access to the same tools yes
now you ask yourself
what would make you a differentiator
that's it
and and that's where my team my lab has stood out yes
we develop these unique set of certifiable AI
tools that today are being sold
and you can trace that directly back to my lab yes
and they're being sold around the world
as certifiable AI but go back to your students
I wanna hear that like a little bit start
start interrupting there
but I wanna hear about your students
how do you talk to your students about
how do you differentiate yourself in a world with AI
what does that look like
when you are talking to a potential employer
founder and investor
how do you differentiate yourselves
in this new world of how information is processed
and and um
and and uh learnings are are complicated and built upon
I do my best not only to teach them
techniques my students
but also to build their self confidence
and that self confidence you build them gradually
and then you have them present their ideas
to a top forum in the world
and get feedback and then learn from that
and then create products that
once we go through our initial testing
we know that there's nothing in the world
that can compete with it
and so also by having those initial successes
that took quite some time in my career to get there
the other students learn from that
so I have a network of my family of alumni in my lab
who then come back
and they inspire the younger generation yeah
how do I complete Professor Cohen's uh
lab and become a millionaire
they're examples you know that they are right
and then you can see that they want to come back
they want to hire more of my team
the entire Thales Cincinnati team is my grad students
yeah
and now they want to increase that times three and a/2
right and
they want to resettle into the Cincinnati Innovation
District so
this is the first time
this large company has a footprint in Ohio
so we are creating jobs
we are creating jobs not because of the low taxes here
but because of a very unique set of expertise
so they've been following what Chat
GPT says about being the premier lab in the world
and so they want to recruit more of my students
so my students now come to me and say hey
can I do an internship with Thales during the summer
can I do an internship with your network
so that I can grow more so
what I promise my students is
techniques that are effective
marketable and 2 connections and experiences
experiences with my network across the world
and you know I've got collaborations in Japan
in Belgium in France in uh
Italy
and it is those collaborations that further empower my
my students and the success of my students
what should change about education and universities
in terms of how they approach learning
what should that look like
there is this level of uh
rigor and honesty that one needs
which is the foundations itself
you need to ensure that you communicate
if it is at a conference level
if it is to a future employer
and if it is importantly to yourself
that what you're doing is making sense
so how do you trust and verify what you're saying
so there is something how do you trust yourself
how do you know that you're not a bullshit artist
yourself
and those are discussions that I have in my lab
meetings and also when I take my team
so I took my team when I say I took my team
I took about 15 of my students to Banff
Canada in August this year
we came back from the conference with the best
paper award all around
of the entire conference and this is a conference
that takes place in North America
once every six years
every two years it's either in Europe or in Japan
we came across with the best student paper award
three outstanding paper awards um
honorary mention the best PhD thesis award
the Early Investigator Award for one of my alumni
who I nominated and the Technical Achievement Award
across six continents yes
given to Nick my student and so if you look at the
the award ceremony about 85
90% of the awards came to my team
and the judges were from uh Canada
Italy uh Belgium and Japan right
not my people in a way right
but uh they
you know
did it because they felt that among those candidates
and the finalists they were the best
well Doctor Cohen
I think my question is less about what you teach
cause I know you're an exceptional teacher
and you are an innovator and you
cause I've worked with you
my question is more about the university model
really to get to that
what do you think needs to change
in the university as a whole model right
that's really has been the same that for the last
every couple hundred years
that was really focused on the era of what I call
intelligent scarcity
where we had to build complex organizations
to keep information
and now information can't be held by AI
that's there's no question about that
so how do we teach in this new paradigm
where it's not about
the old model of memorizing information
but actually learning how to apply that information
even more with AI and other advanced technology
what needs to change in the model
and how do we get there
so before I got to UC 18 years ago
I interviewed and I was a finalist with Texas A&M
and with Syracuse
I was first on the list of Florida Tech
but Texas A&M which is highly ranked
they had an issue with me with fuzzy logic
they didn't believe in it
whereas with UC hey
why not go prove yourself
I was never restricted you know
I had this meeting of IFSA
which is the International Fuzzy Association
it is an international group
you know
how many universities restrict their professors
from teaching fuzzy
and I tell them hey
one month before I teach my class
it is full my fuzzy classes are full
I am an engineering professor right
not math not computer science
but I'm not restricted in teaching what I'm good at
because the university sees that
by giving this academic freedom to the professors
you can now go and explore
other universities restrict their professors
because they have these biases
on what to teach and what not to teach
so MIT or Stanford may not teach a class in fuzzy
but you see never was impacted by them
Kelly you're good in this stuff
go ahead teach
we'll see your results and the results are good
but these other universities a priori said
we don't want you and somebody else edge you out
even though you're a finalist
because we don't feel your technology is up to par well
I proved them all wrong yes right
but the the reason I could prove them wrong
was that I was given that opportunity by UC
so I want UC to continue allowing professors they hire
so once you get hired do what you're good at yes
let's give you freedom
as long as you don't break any rules
which I don't with my right right
and then let's see the outcome OK
and let's see whether your class is marketable right
so if your class is full we can get you a bigger class
but you know that's it right
I can go as should we shorten the education cycle times
no no
no you need to mature
you need this amount of time
I took 165 credits for my undergrad
now we've cut that down to 130
so there's already been shortened
to shorten it even further would be irresponsible
you need to have certain foundations in place
in order to get a degree and I feel that we are
more or less at the bare minimum okay
you're right we're at the bare minimum
it's okay that's interesting
because there's been this tension back and forth
you probably hear it online
it's and I don't agree with it
but I think it's worth mentioning
that
people feel like education is not serving them as well
as they as it used to
that simply getting education
used to be a guarantee for a job um
and now that's no longer the case
and not in engineering yes
in aerospace engineering if you want to get into it
well I agree a company you
we have a very good rate of our graduates
you don't feel like AI is gonna replace that
not non aerospace no
yeah because of the
so you hear go go
go sign up for Doctor Cohen's classes
because of the requirements for airworthiness
and certifiability
as long as we're dealing with people lives safety
critical systems one has to
get trained
to be able to make sure that air travel is good
are you gonna fly on a uh
an aircraft
which is not got airworthy in its certification
of course not Doctor Cohen
I you make such a great point that I wanna emphasize
and then get to a few rapid fire questions here
as we wrap up
no matter who and I want everybody to listen to me
right if you're running an organization um
and you're using AI and you should be right
you have to do these things
cause you have no choice but to compete
but know that at the end of the day
a human has to be accountable for the decisions
yes definitely
a human has to be accountable for the decisions
that doesn't mean you don't use AI
that means
if you have a chatbot that gives bad medical advice
that doctor
and that hospital are gonna have to be liable
and be accountable if you are a airline
and you rely on a system
and it gives the customer wrong information
you have to be held accountable
so this is an opportunity
just to be able to see that you can use AI
and I think you should use AI
I think for processing information
you don't need a lot of humans to do that
so
there's a lot of people that used to hold information
I think that's going away
if you're a middle manager
that just was a keeper of information
but couldn't do anything else other than that
there's a lot of people that I'm describing right now
I think those jobs are a threat
but the next part when you get into the decision
the critical decision making right
you're gonna
still need people to make critical decisions
on things that matter health decisions
whether a customer gets a refund or not
these things matter
and people should be able to make more decisions
now maybe they can make next decisions because of AI
but some human and
humans are gonna have to be involved with this
critical integration of decisions
and then you can get to decision velocity
and make more decisions
and have that compound interest
but if you don't and I hear what
you're saying if you don't include policy and ethics
not as an afterthought but at the front end
you're gonna pay for it later
and in the area of trust and in this environment
nothing's more valuable than trust right
so alright
let's get to a few rapid fire questions alright
alright Doctor Cohen uh
we meet a year from now
what are we gonna be celebrating
we're gonna celebrate uh
Talis
officially being part of the Cincinnati Innovation
District we've gonna be celebrating
success of a few students who will graduate
we'll be celebrating shortly
another couple of years in 2029
the hundredth birthday of U C Aerospace
the second oldest department in uh um
the United States um
and uh we celebrate uh new innovations
new spin off companies
that some of my students are thinking of currently
you know it's just the success giving back
giving back through my students through their success
if the last five years was a chapter in a book
what would be the title of that chapter
and what would it be about
briefly
the last five years
so I've been working on AI now for 31 years
the last five years it is uh
it is defining our leadership in the world
and what we do which was not that clear five years ago
five years ago we had potential
we've now transitioned from potential
to actual credibility
by making products and which are being sold worldwide
finish this sentence
what's the one thing AI will change X for sure
it will
make the world a smaller place
and also
there will be new developments in the area of medicine
medical human health related
there will be new advances made
so I'm positive about what we all
the good we can do with AI
I'm sorry about that that's okay
um finish the sentence
AI will never change and you can just
just repeat that back AI will never change
AI will never change
certain elements like the importance of being kind
the importance of mentorship
the importance of showing empathy
so there are these human elements
that cannot be replaced with AI
AI is not a feelings living system
it is an assistance to a human
so we I feel that the world will be more human centered
and AI will play a role in empowering the human okay
so you think AI will never make the
AI will never replace empathy or AI
will make us more empathetic
what's your what's
what's your what's your thought
no it can never
it can never replace
it will not be able to replace because
you know you've got students
students have lives they've got good days and bad days
you don't wanna fire somebody
because he's had a bad day
you empathize you understand
you see and also
certain folks
struggles during different times of the year
yes with certain things
and so as a mentor you need to see things through
their point of view and then give them an assist
with what they're struggling with
I don't see AI doing that level of mentorship yes
they cannot
place themselves in the point of view
of each and every individual student
and see what struggles they're going through
on the personal side on the emotional side
and then take the right action
something a good mentor does by knowing his people
by knowing how to motivate them
AI is not going to be a good motivator
so chat GPT wants to be likable and agreeable
so it can flatter you
but you know that that flattery is empty yes
it is not coming from a good necessary
it's coming from some algorithm out there
which is as cold as ice right okay
it's different when you put your arm around
I mean student
I'm not talking about in a positive way
give a positive I know what you mean that yeah
yeah yeah
you know because it could be misinterpreted
you don't want to go there
no but
but what I'm saying is that that support
you provide somebody a good word
an encouragement right
goes a long way I agree
and my my hope is that that's the case
and as we end here you know
studies have shown though um
many of many youth are using AI for therapy
yeah right
and and they are getting that reinforcement there
so yes it can never replace humans
and I think that's
that's an area we always need to remain vigilant on
because it can end up eroding some of our empathy
which is what we have to be very careful of
which is why we always talk about making technology
human centric here at disruption
now Doctor Cohen
it's always an honor to work with you
we we work uh
together a lot I'm looking forward to the future
and building more together
thank you for all you do
thank you for having me on your show
thank you it's been a pleasure
thank you very much keep disrupting