Return to site

61. AI Revolution with Denis Rothman: Exploring BehaviorAnalysis, Recommendation Systems, and the Future of Large Language Models

· podcast,AI


broken image


Podcast with Denis Rothman Part 2



In Part 2 of our captivating podcast with AI expert DenisRothman, we delve into the fascinating realm of AI's future implications.
Discover how behaviour analysis and recommendation systems are already shaping
our daily lives through tech giants like Alibaba and Google. But are these
systems truly capable of understanding and predicting our behaviour? Join us as
we explore the fine line between human behaviour and AI algorithms, and uncover
the limitations of large language models in generating truly creative outcomes.
Plus, gain insights into the potential applications of AI in healthcare and the
delicate balance between human expertise and AI assistance. Don't miss out on
this thought-provoking discussion that challenges our understanding of the
future of AI. Click here to listen to Part 2 now!

InPart 2 of the podcast with Denis Rothman, the discussion revolves around the
future implications of AI and large language models. Rothman emphasizes the
importance of understanding the present and the existing capabilities of AI,
particularly in relation to behaviour analysis and recommendation systems used
by tech giants like Alibaba and Google. These systems track user behaviour to
make personalized suggestions and optimize user experiences. However, Rothman
also highlights the limitations of large language models, as they are based on
existing data and patterns, which restricts their ability to generate truly
creative or unpredictable outcomes. The conversation touches on the potential
applications of AI in healthcare, where classical algorithms and expert
analysis are still crucial, while AI can serve as a valuable assistant in data
interpretation and optimization. The episode concludes with a thought-provoking
exploration of the future of AI and the role of large language models in
shaping human behaviour and decision-making.

[00:00:00] Andrew Liew Weida: Can you share us what willthe future look like in this kind of context?

[00:00:04] Denis Rothman: Yeah, let's talk about thepresent. Yeah. Because it's, so there, there are two things.

[00:00:08] Denis Rothman: The future of what the publicknows and the present of what really exists. Okay? Yes, please. So the future
of what the public think is coming is already here in the present. . Let me
give you some example. You have Alibaba in China. You have Google in United
States, and they have transformers.

[00:00:29] Denis Rothman: These transformers areeverywhere. And what they do is they use 'em as recommenders re recommender. So
if you go on YouTube or you go on Facebook, or you go on TikTok, or you can go
wherever you want, or you can go on in China, you can go on the Chinese version
of TikTok. I don't know if you are aware of that, but there's no TikTok in CH

[00:00:51] Denis Rothman: TikTok is in United States, butin China they have a special version, which is very educational. You can't use [00:01:00] the general version. So in any case, bothsides are taking your behavior, they're using your behavior. How do you behave?
Okay? So your behavior in the United States. For example is how do you

[00:01:15] Denis Rothman: What do you do? What gender?What this, what that politic in China, what's your behavior as a social person?
All these behaviors are being, are noted, okay, but we don't need to know who
you are. We just need a lot of behaviors. So we need a lot of behaviors to
understand how this works. And once you have these behaviors, these are rules.

[00:01:36] Denis Rothman: So it's already there. They'resemantic, the way you behave. I'm not in, we're not interested in your private
data or who you are. We just want to know your behavior. If you click that,
then you go there. Ah, that's interesting. So maybe another person that clicks
like you, that's interesting. We can suggest, so these behaviors become rules.

[00:01:56] Denis Rothman: These rules become like expertsystems. So in fact, we. [00:02:00] we're justunknowingly feeding a lot of rules into the system, which then puts it in a
transformer and the transformer will say, ah, there's a sequence. It's not just
a sequence of words, it's a sequence of events of anything. So you go from A to
B to C to D, but he goes from D to F.

[00:02:16] Denis Rothman: Then you get these suggestionsand it's a treat. And it's so it's already there. The semantic part is already
in these transformers because they combine different things. They even still
use cosigned similarity, which is a very old algorithm. These algorithms are a
mixture of many other algorithms already.

[00:02:37] Denis Rothman: So it's going to go in thisdirection. We're going to see rules merge with probabilities knowledge events,
private data, but not in the old sense of who you are, your name. No one's
interested in fact in that. But your behavior is the key. How do you behave? ,
that is the key word for the future, to control behavior or to suggest

[00:03:00] Denis Rothman: So on one side you can control.On the other side you can advertise, but in any case, you're influencing people
with their behaviors. That's why people don't are always on the smartphone. You
can't leave it because since we know exactly what you want, give it to you
immediately. It's even better than sugar or nicotine or alcohol.

[00:03:20] Denis Rothman: You get it so quickly.Satisfaction. So it's here already. So the future, what people think the future
is more. More of this huge amounts of this. The next step is meta, means that
this meeting will not be with two pictures, but will be in the same room. We'll
be in another reality and then we'll need a psychologist because people won't
see the difference between that world and the real world anymore.

[00:03:44] Denis Rothman: They're gonna be a lot ofpsychological problems. .

[00:03:47] Andrew Liew Weida: Like it, it sounds when youtalk about the meta, like the way you describe it, it sounds like those I don't
remember a science fiction movie called Matrix where yeah. Like Neo, he
couldn't really differentiate between the [00:04:00]reality and machine world.

[00:04:02] Andrew Liew Weida: Yeah. In sense.

[00:04:03] Denis Rothman: Yeah. But the, but yes, butthat's the present. That's the past. Because Matrix, speaking about Matrix is a
very interesting movie. It's not just a action movie. If you go to Eastern
Philosophy and Religions, if you go to Hinduism Buddhism, you'll see that they
speak a lot about my the

[00:04:23] Denis Rothman: The way we represent the worldis an illusion. And if you go even to German and French philosophers like
Emmanuel Kent, which is one of my favorite you see the world through a
representation. So we already see the world through representation. . But we
don't realize it. That's what Matrix is trying to explain at the same, it's a
very interesting movie.

[00:04:45] Denis Rothman: The thing is, today we userepresentations without knowing it. For example, we don't know what's going on
anywhere. The only thing we can do is watch it through news, information that
is given to us, [00:05:00] and then werepresent the world. Any, we represent all the countries in the world through
news. But if you read newspapers around the world, you won't get the same

[00:05:10] Denis Rothman: I like to read all kinds ofnewspapers from all continents. I have no filter. I can read any newspaper,
African Chinese, Russian American German or England. I don't care. I just want
to see the different perspectives of the same event. And you can see that
everyone's living through a representation but it's strange that a same human
being, Would have two different representation just based on the place he's
looking at the information

[00:05:38] Andrew Liew Weida: Yeah, I mean like talkingabout that, like you're right in the sense that if we, we have a part of the
world that we are living, the information that has fit to us has a certain
level of representation depending on the local government and the local media
company or even technology company.

[00:05:54] Andrew Liew Weida: And the way that youactually get a better view of what is the truth is to [00:06:00]read multiple perspective or multiple sources on the same set of information.
So you get a much more holistic angle about, the classical one that right now
we are talking about like Ukraine versus Russia you get a sense of what's the
American are thinking, the European are thinking the Russian themselves, the
Ukrainian, the Chinese, and you get to see, you forget the Indian Oh yes.

[00:06:22] Andrew Liew Weida: The Indians. And so theinteresting part that like one of the use case I always wanted to ask the only
problem is the truth. We don't know the truth. Yeah. That, like how do we
interpret the, because as human beings, yes. As human beings in the world, we're
living in the parameters.

[00:06:38] Denis Rothman: There's so many parameters. Theworld is so large, so big with different cultures and everything. It's very
difficult to understand what the truth is. That's a lifelong quest to find the
truth. We don't know the truth. We're just a little human being. We can, we
under, we can get a holistic representation, but what the truth is, that's, we
have to be very [00:07:00] modest.

[00:07:00] Denis Rothman: So you're, you were talkingabout a use case?

[00:07:02] Andrew Liew Weida: Yeah. Because one of theuse case I always wanted to ask is let's talk about healthcare. Because it's
that singular truth that affect everybody regardless of how we, of course
there's always different interpretation, but let's use Western medicine as a
more as a specific context.

[00:07:20] Andrew Liew Weida: So if you think about it,how would the future of using large language model being used in healthcare
where the objective truth or the biological truth will actually help the
patient and the doctor in a sense? Yeah.

[00:07:34] Denis Rothman: When we're talking aboutmedical care, we can't just speak about large language models.

[00:07:40] Denis Rothman: We have to speak aboutclassical algorithms very classical, or two plus two equals four like
spaceships. You're not going to send a spaceship to the moon with
probabilities. Oh, maybe this is a good answer. Why don't you turn left. Why
don't you go right up. So you need. for medical purpose. [00:08:00] Very classical, old-fashioned math, whichis very solid.

[00:08:04] Denis Rothman: Okay. Then you can have somerule bases. Like you can have a rule if this, if you see something on the skin
and it's that big, then maybe it is something. But these images are extremely
difficult to read when you're reading imagery. Imagery is very difficult. So
you, for the imagery, you can use artificial intelligence, but you need pure
technology that will capture the image better so that a professional can
interpret it.

[00:08:34] Denis Rothman: And then after that, I wouldsay artificial intelligence is today because you can always imagine the world
tomorrow. Today it's used already by, we're they in a Pfizer? They're already
using it to find combinations to see what about. , could this medication solve
that problem? Could this molecule solve it solved that one.

[00:08:57] Denis Rothman: Could it solve another one? Andthere's so many probabilities and [00:09:00]parameters. It's very interesting to use artificial intelligence to create
things we wouldn't think of, but then show it to an expert and the expert will
say, no, that's, ah, that's a good idea. Oh, that's be, ah, that's a good idea.
But it can't replace a human at all Today if you try to do that it's very

[00:09:19] Denis Rothman: But you can use a lot ofhelper, like when you're brain surgery, you can have enhanced vision. You can
have better images. Like today we're using Google. Yeah. And, but Google people
say, turn your camera on. Let me, let's use an example that we can use in the
medical field. Yes. Okay. If you're operating someone in the bring, you need a

[00:09:40] Denis Rothman: Okay. So people say, we're inGoogle Meet. Turn your camera on. The thing is, I'm on a smartphone. . And on
my smartphone, there is no camera. It doesn't exist now. Camera is something
where you have a shutter. There's no camera in here. It's just taking light, a
lot of light. And it's packed. I have a Google [00:10:00]phone.

[00:10:00] Denis Rothman: It is packed with ai. It ispacked with algorithms so powerful. You can't imagine it's not even our face.
We're looking at, we're looking at an enhanced image. It's making my face
better. It's taking little aspir charities off. It's making a nice face. So if
you're operating someone, it's good to have a good image.

[00:10:20] Denis Rothman: So this can help, but it's notgonna help make the decision. It's gonna get better information. So I would say
it's a it's a subject that will evolve. I was talking with someone in my family
was a doctor, in fact, yesterday, and I was saying, could this and in fact, an
explainable ai. Another book of mine in chapter one, I started it with a nephew
of mine who is a doctor too.

[00:10:45] Denis Rothman: So let's take both examples.The first example is yesterday. I'm talking about large language models with my
my brother-in-law, and I'm saying, could this help you? So I asked the
question, you're at home. We're in a few years and we have [00:11:00] a medical kit. The medical kit will takeevery vital information on us.

[00:11:06] Denis Rothman: The okay, the blood pluspressure, the heart all these basic things at home. Even these e c every a
little kit like this big that you put next to you, and it can monitor anything
you have on you. And I say, now I feed this information through a phone,
through a communication to you.

[00:11:27] Denis Rothman: You're a doctor, so you receiveall this in. And a nice piece of software with all the history of the person.
You can see the curves of all these indicators. You can even imagine for
diabetes a little the sugar levels and all that. And when you're a doctor and
you don't take, you don't take blood, you just analyze all these symptoms and
the person is coughing and all that.

[00:11:48] Denis Rothman: You can see it visually. If theperson, if I give you all this information and you have it very nicely
presented, could this bring your time with the patient from 30 minutes, for [00:12:00] example, down to 10 minutes withoutreplacing you? Cuz I'm for the moment that I would suggest that because same,
you'll see in my second example.

[00:12:08] Denis Rothman: So he says, yeah that's good.Because if I had all this information immediately without having to do it
physically, of course now he's a technical, he's a technical doctor, modern
doctor. So he says, yeah, I would get more information and I would go faster.
Okay. So that's one possibility.

[00:12:26] Denis Rothman: But I have my own doctor, whenI say that to him, says, yeah, I'm not so sure because when I have a patient,
I'm not just doing the tem talking to this patient that's not feeling good and
I'm doing some psychological work too. Oh, I'm taking your blood pressure. But
how is it today? Did you go to your, did you go somewhere?

[00:12:46] Denis Rothman: How are you feeling in general?So there's this psychological help that a doc, a good doctor will given, so the
person feels better even without medication. Oh, it's so nice for you to talk
to me cuz.[00:13:00] Okay. And the machine willwon't replace that . Okay. And then you have my second example is, and my other
book, hands on Explainable ai.

[00:13:09] Denis Rothman: We're there, artificialintelligence is almost necessary. To help the doctor find something. And it was
about this mosquito from Africa that brought a disease to the United States.
This is a, it's a real story, but with artificial intelligence, I managed, I
certainly could track the location of the patient everywhere he went.

[00:13:30] Denis Rothman: And this disease that thisperson got in the United States doesn't exist, for example, in France. So the
doctor's looking at him and he can't figure out what's going on. And this came,
it's strange because this was a case in January of the year Covid began, and I
phoned my nephew in January. It's just before all this happened.

[00:13:51] Denis Rothman: And I said, can you give mesomething where one symptom can have several diseases? Say, yeah, I have this
patient that just solved the problem in two weeks. He had a fever [00:14:00] and I couldn't figure out where this feverwas coming from. And it's something we don't know about. And I even did these
blood tests and everything.

[00:14:06] Denis Rothman: I couldn't figure it out. So Ihad to go through all the locations he went to through with a, on a piece of
paper. And what did he do? What, ah, and then one time I figured it out, I
said, ah, you went to this animal farm and you caught a virus type a virus that
only goes from animal to men, but then not from men to another man.

[00:14:28] Denis Rothman: Again, you caught this thingthere and I found it through the location. So I got a, that's where I got my
idea for the first chapter of explainable ai. If the doctor has the trace of
all these locations, you can say, yeah, sure, you got that there, because with
this other medical statistics, here are the diseases that appear in this place.

[00:14:46] Denis Rothman: So that can be very helpful fora doctor, but it's not gonna replace his expertise. It's gonna, but it's gonna,
it's really gonna help. So I say artificial intelligence is a great helper.
Cool. Now the next interesting question, I want [00:15:00]to know what is your view on the future of ai? With reference to large language

[00:15:06] Denis Rothman: Now we get to another, this isa difficult subject because it's a painful subject for human beings. , why

[00:15:14] Andrew Liew Weida: painful ?

[00:15:15] Denis Rothman: Hu it's a, you're gonnaunderstand, first of all, large language models, as I explained earlier, will
take all the sentences people say, right? But it, you have friends, you have

[00:15:29] Denis Rothman: They always say, they alwayssaying the same thing. In fact, you can have a new event. You go to lunch or
dinner with your family and there's a new event, and you say, I know exactly
what this person's going to say. I know his political views. He's gonna say, If
I say this, he's gonna say that if you go to the United States today, you can
predict exactly what someone's gonna say based on his political affiliation.

[00:15:53] Denis Rothman: If he's Republican or Democrat,libertarian, whatever, and you know exactly what he's gonna say, give him a new
event, [00:16:00] boom. He's gonna repeat hispattern. So what happens with human beings, if we go to behavior, that's why
behavior is as important as expression, because w we have the same behavior.
We're like cats and dogs.

[00:16:14] Denis Rothman: We, we peak the same thing allthe time. Bababa, there's not much in, if you get up in the morning, think
about your light. You get up in the morning you maybe you wash. Maybe you have
your breakfast first, maybe you do. But think how many times in a year do you
do something different In that sequence of events, maybe you just invert

[00:16:34] Denis Rothman: Or maybe you don't want to takeyour shower this morning when take in the evening. Or maybe you want to eat
your breakfast in the morning, but not in, but maybe a bit later we're gonna be
eating mostly the same thing. You're not gonna change your habits. So habits.
So we develop these habits as human beings, and we stay in them, and we're
stuck in our behavior.

[00:16:54] Denis Rothman: And that's why it makes it so difficultbecause now we're uniting everything in our conversation. We're [00:17:00] speaking about different representationsof the world, because once you grew up in a culture, you're stuck. You have all
these no, that that's the way it is. But it, so you're like a dog.

[00:17:10] Denis Rothman: You just learn something, butyou can't learn something else.

[00:17:13] Andrew Liew Weida: I agree with you that humanbeing are predominantly driven by behavior. I call it deterministic thought
which is either is fixed or is a habit, but there's also an aspect of human
being whereby I call it enlightenment or figuring out something like, hey,

[00:17:28] Andrew Liew Weida: Every morning I've beeneating ice cream and suddenly I realize that, hey, eating ice cream is really
bad for my health. I'm gonna eat fruits instead ice. But

[00:17:35] Denis Rothman: that's not Yes. But then you'reout of the large language model. Ah, see what I mean? The large language model
will take all the habits, all the patterns that repeat themselves in

[00:17:48] Denis Rothman: The ones that are not probable,we're not gonna take him. So I'm gonna say he's probably going to eat a
pancake. He's prob, but then human beings are so unpredictable that he might
just have someone [00:18:00] that changes hiseye like genetic modification. Someone living in one country and says, I don't
agree with my country.

[00:18:07] Denis Rothman: Maybe I'm going to leave mycountry maybe. And all of a sudden the behavior doesn't fit anymore. It's not.
So the large language model will go all the way to the border of repetition and
will start exactly where you're talking. What you're saying. Or maybe we don't
want to go on forces anymore. Maybe we have to invent a car.

[00:18:27] Denis Rothman: Maybe we would. As soon as itbecomes very creative, unpredictable. Hold on. That's the end of large language
models. They can't, you can't invent anything cuz it's based on existing data.
So as soon as the human being becomes creative, then that's the end of the
large language model.

[00:18:43] Andrew Liew Weida: So you're saying that intothe near future this kind of large language model is helping us to predict like
recurring behavior, but don't you, do you think that will also help us to free
out a lot of our connective mind to think about it?

[00:18:58] Andrew Liew Weida: Maybe we should change [00:19:00] certain habits. I don't know. I'm justthinking a lot here.

[00:19:02] Denis Rothman: So now what can be the effect?How can you use a large language model at best? So this one I can say, through
my experience in corporations. Yes, please. Because I would go see someone and
I say, tell me out of your day how much work is created.

[00:19:19] Denis Rothman: Not always in the sense ofcreating something new, but let's say you're just someone getting parts for an
airplane. You're just specialized in getting screws and all that stuff, so your
job doesn't look interesting. So all day you're spending your time on all of
this, but all of a sudden I give you an algorithm and I give you an algorithm
that will optimize because I didn't only do large line, I was doing a lot of
optimizing the resources.

[00:19:46] Denis Rothman: You won't need so many screwsbecause you're wasting a lot of screws because they're lost in boxes because of
they break or on the airplane, I shouldn't say this online, but airplanes lose
a lot of [00:20:00] screws when it's live,people.

[00:20:01] Andrew Liew Weida: Be too scared to fly.

[00:20:03] Denis Rothman: Now they're not screws, they'rerivets that you put under the plane.

[00:20:06] Denis Rothman: And sometimes they just theyjust fly. They just, this is, so you look at some big airplanes, you have 50 of
them missing. So we try to find ways maybe in maintenance to optimize this so
that they don't fall so quickly. But okay, let's pretend people live. Hear
this. And I'm speaking about big airplane so you optimize this and I say, what
if I optimize your day?

[00:20:27] Denis Rothman: So the time you're wasting, Ibring it from seven hours down to one hour. What will you do at the rest of
your time? He says I will have time now creative in the sense not creating
something new, but to talk to my suppliers. Maybe go see them, maybe find
better ways to better trucks or maybe cargo or better ways to work and we can

[00:20:49] Denis Rothman: So yes large these largelanguage models. can help you go faster. Maybe in communication I can send some
customized messages to [00:21:00] my to mypeers person. They'll be nice. And also I use chatbot a lot for myself as a
search engine. You have Bing, the new Bing. Yeah. I use it a lot to get the
information faster.

[00:21:13] Denis Rothman: I know, because since I knowthe algorithm I know the ones. That's why you need to understand. I don't waste
my time on questions. It can't answer. It can't answer. I would say 40% of the
que what I'm looking for, it can't answer. So I'm just, I know it's, I know
where it can answer. So I'm looking for where it can enter and it's for
example, I did a LinkedIn post this morning.

[00:21:34] Denis Rothman: Yes. And I was thinking of howto express what a multi-model I said, why don't I just ask chatbot instead of
trying to craft a nice sentence for 15 minutes, cuz it takes time to write a
good post. It, it doesn't take time to write buzz or hype, but it takes quite
some time to, with a good post and you wanna put a program with it and people
don't realize.

[00:21:56] Denis Rothman: But if you want to do goodwork, it

[00:21:58] Andrew Liew Weida: takes time. Yes. So [00:22:00] I said, why don't I ask Chad

[00:22:01] Denis Rothman: g p t? And it gave me this niceanswer and I just copied it in my LinkedIn post and I put it in quotes of
course. I said, this is chat G P T because open, and I says, you have to do
that. If you're using it, you can use it. No copyright, but you say it's chat,
G p T.

[00:22:16] Denis Rothman: So I say chat, G P T told methis and chat. G p t told me that dli, so I, it saved me, I don't know, 20
minutes cuz sometimes I'm trying to find the best sentence so that in a post
where you don't have any words, it looks good. So instead of spending maybe one
hour on my post, I spent a half an hour.

[00:22:34] Denis Rothman: And then after that I used itfrom other things I'm working on. and at one point it was like an hour and a
half before our podcast and I said, gee, I thought I wouldn't have enough time
before the podcaster finish my morning. Now I can even take a break. So I took
a break and I walked around and I said, gee, I'm going, I'm gonna have find
other things to do if I work like that.

[00:22:55] Denis Rothman: So cuz they're gonna help youwith program certain, be extremely helpful if you [00:23:00]use it as you use a scientific calculator.