Everything You Think About AI Is Wrong with Damien Benveniste

Welcome to episode 266 of the Business Development Podcast.

And today I'm joined by Damien Benvenist, a machine learning leader and former Meta tech lead who's now helping businesses and entrepreneurs understand and apply AI.

In this episode, we break down what AI and machine learning really are, how they've been shaping our world for decades and where the future is headed for business, innovation and everyday life.

Stick with us.

You're not going to want to miss this episode.

The great Mark Cuban once said, business happens over years and years.

Value is measured in the total upside of a business relationship, not by how much you squeezed out in any one deal.

And we couldn't agree more.

This is the Business of Development podcast, based in Edmonton, Alberta, Canada and broadcasting to the world.

You'll get X expert business development advice, tips and experiences and you'll hear interviews with business owners, CEOs and business development reps. You'll get actionable advice on how to grow business brought to you by Capital Business Development capitalbd ca.

Let's do it.

Welcome to the Business Development Podcast.

And now your expert host, Kelly Kennedy.

Hello.

Welcome to episode 26066 of the Business Development Podcast and today I bring you Damian Benvenist.

He is a distinguished leader in machine learning and data science with a career spanning over a decade and marked by groundbreaking projects across industries from ad tech to healthcare.

With a PhD in theoretical physics, Damian seamlessly transitioned into a world of applied machine learning, leveraging his expertise to develop transformative solutions in fields as diverse as online retail, energy valuation and credit score.

Modeling his work at leading tech firms like Meta, where he recently served as a machine learning tech lead highlights his role in scaling complex model optimization processes, showcasing his ability to innovate at the intersection of technology and real world application.

Now stepping into a bold entrepreneurial journey, Damian is harnessing his experience to build tech focused ventures and share his insights through his acclaimed publication the AI Edge.

Focusing on continuous learning in machine learning, system design and ML ops.

Damian's work is a beacon for those eager to advance in these fields as he moves forward.

Damian's relentless pursuit of impactful style scalable tech solutions promises to inspire and elevate the next wave of machine learning innovation.

Damian, it's an honor to have you on the show.

Wow, that is the best introduction that anybody has ever done of me, even myself.

So I need to live up to that now.

So thank you Kelly for having me.

That's great.

I'm excited.

I am extremely excited.

One because AI to me is incredibly amazing and honestly I have no idea how it works.

And two, you're the very first data scientist I've ever had on this show.

And dude, I don't even know what a data scientist is, but I'm excited to find out.

You know, I can try to give you a sense if you want.

You want to move to that or you want to.

We start by something else before.

Yeah, you know what?

Yeah, sure, why don't we start with that and then I want you to take us back to the beginning and how you ended up on this incredible journey.

But yeah, sure is a data scientist.

For everybody listening.

I would not actually describe myself anymore as a data scientist.

And it's funny because data science or data scientist has different definition depending on where and when.

So when I started my career, somebody that wanted to work in machine learning or somebody that wanted to train machine learning models and build product with machine learning models were data scientists.

So it was fashionable at the time to hire PhDs to trans.

I mean, to become a data scientist.

With this field being very hot in the 2000 and tens, it was like the job of the 21st century.

At the time, many people tried to transition in title to data science.

So many data analysts, many business intelligence analysts, bi people, even some data engineers would take on the name of the title of data scientist.

And what I've seen in the US at the very least, is that people that used to be data scientists back in the 2010s slowly transitioned to being titled machine learning engineers to push the idea that the people that were working with machine learning models were, you know, that they were working with machine learning models.

And data scientists became this label that could describe anything.

Now if you dealing with data from close to far, many companies will give you the title of data scientist.

And it can be very different things.

Now people that wanted to really specify the idea that they were machine learning experts slowly started to take on the title of machine learning engineers to make sure that they were distinguishing themselves from people that had other type of skill sets.

So now I would describe more, I mean, I transition titles and I would describe more myself as a machine learning engineer.

Although like you said earlier, now I'm more into the educational part of machine learning and an entrepreneur.

So I'm not anymore effectively actively working as a machine learning engineer, but it's a better description of who I would be as an engineer.

Yeah, yeah, okay.

Yeah, it's, it's just for, for people that are outside of that field, it's really hard to understand.

Right.

Like, like you mentioned, people have Been working on this for years, like really in theory.

And like that's what I've been learning.

I've interviewed a few people on AI and what they kind of said is, look like AI has been around forever.

Like, you know, your, for instance, they, they gave your spam folder in your email and they're like, how do you think the spam folder figured out what spam was?

It was AI.

But like, none of us, none of us understood that.

We just assumed that there was something in the system filtering things.

But like the background or the behind of how old AI is can sometimes be a little bit unbelievable.

Like in your experience, you'd mentioned, really since the 2010s is where essentially the AI that we're dealing with today came from the chatgpts of the world.

No, I would not say that.

Back in the 70s in computer science department, they were working on AI we, which was a very deterministic domain, something where we were trying to formalize intelligence, trying to understand what it means for a computer to be intelligent, to have some cognitive capabilities at some point.

There was a transition in the 80s, the 90s, where.

And it was helped with the similar transition that we saw in statistics department where the idea was instead of trying to formalize intelligence, we're going to try to extract the statistical patterns from the data of entities that generate intelligence to try to mimic this intelligence with some models that would be able to express those statistical patterns.

And that's when machine learning became a more prominent science in as a subset of AI.

So it is this science where we are trained to.

It's a subset of artificial intelligence where we are using data to extract the information instead of trying to guess it, instead of trying to build some kind of models by hand with human intelligence.

Instead of that, we are looking at the data and we are building models that have the flexibility to extract the information as complex as it can be.

So came the birth of machine learning that was much more.

That has a simpler goal because what machine learning was good at was to be able to learn the relationship between different variables.

So try to understand.

So for example, if you think about linear regression, that's the simplest machine learning model that you can build.

And the goal of it is very simple.

You regress a variable and trying to understand the relationship of a variable to another.

And there's a lot of this in machine learning where we're just trying to understand by using some input data, some input variables, how it's going to impact another variable.

So for example, when we look at the text of some emails.

What would be the effect of it when it comes to trying to classify this as a spam or not?

When we look at the history of a user on a platform like meta, like Facebook, trying to understand the user activities to try to infer what would be the affinity to that user to a specific ad.

And this is a way meta is making, for example, 95% of its revenue by placing personalizing ads on social media, on feeds to make sure that users are clicking on them.

And the more they are clicking on them, the more they are making money.

So they need very accurate models to make sure that we are providing to the users the most adapted ads for them.

And this is machine learning.

It's very, you know, like the goal of it.

There's nothing cognitive about it.

There's nothing intelligent.

Yeah, there's nothing, or at least there's nothing that is easy to describe as being intelligent.

But it has always been a subset of AI and we were fine to describe this as a subset of AI.

The people working in machine learning never had a sense of.

I never had the sense that I was doing AI.

I was building regression models, like classification models, simple models that were used to build software application.

There was no idea of, there was no sense of cognition, intelligence, reasoning.

That was not the point.

The point was really to build something that is effective business wise.

So most of the big tech companies are using machine learning to make most of their revenue.

When you are on the Amazon website, you are recommended objects items to buy.

The more we are recommending you the best items, the more you're going to buy stuff on Netflix.

We are recommending the best movies.

The more we recommend to use the best movies, the less attrition that there is on the platform.

Meta Google.

Google is making 70% of its revenue with ads.

It's again machine learning models making that revenue.

So most of the revenue of those big tech companies has been generated by machine learning models for decades, maybe too long, but a long time.

Long before ChatGPT and machine.

Every day before ChatGPT we used 20 different applications that are ML powered.

You know, like Google Map, Google Search, Netflix, Amazon, anything you're using every day, Snapchat, whatever.

You know, those are always ML powered and ML has been around us for a long time.

Wow, wow.

Okay, okay, see you.

You made a pretty clear differentiation.

And essentially that machine learning isn't necessarily AI or like a thinking machine, is that you're a thinking computer is kind of what you're getting at.

It's still pre programmed but it's it's learning from the inputs that we're giving it to recommend things based on if other people like this is what we should do them.

Does that sound.

Is that.

Am I getting this correctly?

And I'm not.

I'm not suggesting that AI is a thinking machine either.

Right, okay, okay.

But it appears to, yes.

Okay.

But when it comes to machine learning, at least on the applications that we have been doing with those, it's been difficult to perceive a sense of intelligence.

You may have used Google Translate, you may have used like those kind of tools that were obviously ML powered, and you could extract from that a sense of intelligence.

But I think before ChatGPT, people didn't see a sense, didn't imply that there was some kind of intelligence in those models.

So now in 2022, December 2022, came out, chatgpt in a way that was very public.

You know, that was not the first time that we had LLMs.

I know that six months prior, there was an employee in Google that got fired because he really thought that, that the language model he was working with was conscience.

You know, I forgot his name.

So, you know, it was not, it was not ChatGPT that really discovered that, but I think it was ChatGPT that made that kind of capabilities very public.

Even myself working in machine learning, I could only experience this type of models through papers.

By reading papers, you don't get a sense of how intelligent maybe that kind of machine can be if you are reading a paper.

But if suddenly you're in front of a UI with a chatbot and you're chatting with it and you see that thing responding to you in a way that would be the Turing Test very easily, you're suddenly surprised.

And there are many people working in the field like me, that were very surprised that this kind of things could be.

But ChatGPT is the result as well of machine learning.

The difference between these and maybe other type of machine learning models is that those machine learning models are specialized in generating text or generating natural language in a similar manner, by training on the natural language text that we are generating as humans.

So it's specializing into generating text.

So a large language model is a classifier.

In the same way that you classify ads, yes or no, that guy is going to click on it.

Or no, you classify words and you are generating words little by little.

By choosing, by using the probability that the model is outputting, you're choosing the word that will be the next two output into the sentence that the LLM is providing to you.

And it Was.

I think it was surprising because the difference is that despite the fact that on the basic aspect of those models, they were very similar to any of the models we were using before, at least conceptually in the application, they were seemingly displaying intelligence in a way that we rarely seen before.

Yes.

And it became suddenly a bit more relevant to talk about AI.

You know, like, I personally, I never considered myself to be an AI expert.

I considered myself to be ML expert.

There's a whole field of AI that is not ML, that is completely separated from ML, and that I'm not pretending to know.

But suddenly those models that were displaying a sense of intelligence, it made sense to suddenly start to talk about AI, you know, like much more than before.

So I think AI is not new, but the popularization, you know, the fact that it has been democratized is very new.

The fact that you can access AI in a way that was not accessible before.

You know, you needed to be an engineer before to be able to access AI, or engineer or researcher, and now you can be anybody and you can have access to the most powerful AI we ever created.

Yeah, obviously.

I mean, if you're in business, if you're in anything, if you're living with access to a computer, it's been the biggest, most massive, monumental shift since probably the invention of the iPhone.

Like, genuinely that big.

I would argue even in year two, we're already that big.

And.

And I had no idea, dude.

Like, I don't know whether I had my head in the sand or what, but I had no idea anything like this even existed.

And then suddenly it was like, oh, chat GPD's here.

And I, you know, at first I was like, oh, whatever, like, yeah, so people playing on a computer program.

And then I hopped on and started playing on it myself.

And I think, like, anybody who it's their first time, I was like, hobby.

Holy cow.

Like, this is going to change everything.

And obviously we've had quite a few updates since then.

It's now connected to the Internet and it's like, wow, like it's game changingly unbelievable.

And you're absolutely right.

Like, I could see how people could look at that and be like, are you sure there's that thing's not conscious?

Because it seems pretty damn smart.

Yeah.

So, you know, I mean, what's kind of cool is that you have a background in theoretical physics as well.

You know, what are your thoughts?

Do you think we'll ever get, you know, one of these machines to become conscious, like, or is that where we're going?

You Know, it's, I'm very, you know, like I could try to answer, but before answering or try to answer, I would say that I'm very uninterested by these kind of questions.

I'm very excited about the technology, things you can build with it.

You know, I'm, I'm product oriented.

I'm, I'm, I like the technology and I really, for me everything is demystified.

I'm not thinking about those models as being thinking machines.

I'm seeing a model that is mimicking the intelligence that we're able to display in natural language, which is great.

But the idea to try to project in the future and me trying to infer if we're going to have intelligent robots in the future, sure, why not?

But you know, like it, to me, it's, it's, it's, it's not really an interesting question.

You know, like, it's, it's fantasy, Fantasy, you know, like I don't see a good sense, a good reason for me to fantasize about what could happen.

There's a lot of things that are missing to maybe connect to what we tend to define as being intelligent being.

So, you know, from there, you know, I don't have even the knowledge I believe to understand what are the research that are done beyond LLMs and on the cognitive aspect to get to something that is intelligent.

And I only aware and I'm knowledgeable about the idea that we have that machine that is mimicking our level of intelligence by just reusing the data that it's seen during its training.

And I'm never as a machine learning person, the guy that was personalizing ads for ads ranking on the meta feed, Facebook feed.

I'm not the kind of person that fantasize about what will happen in the future when it comes to thinking robots.

You know, it's never been, never been my thing.

I find that not very productive even.

Yeah, you just, you know, you just end up to be in a debate with people that are on one side or the other of a specific, you know, opinion and you, you end up to be in a, in a very unproductive debate with no data to back up any claim.

No, that's fair.

I just watched way too much Terminator growing up.

So it's like, you know, hopefully the longer we are away from a conscious or sentient robot, doesn't matter.

It's, it's something that was, I got scared for a minute when it comes, you know, you get scared when you don't know what's happening.

Sure.

I remember reading the paper that came out in May 2023, I believe, by OpenAI, that was the paper that was describing GPT4 or April, maybe 2023, and it was describing GPT4.

And one section of the paper was how they actually tested the model for self replication.

And you know, like that paper specifically, they didn't describe the models, they didn't describe the architecture.

So we were not clear about what architecture they were using, what specific machine learning models they were using.

So there was a sense of fear because we were at the point where we were testing models for self replication.

Can they on their own start to replicate and you know, like becomes Kaynet, you know, Terminator.

So it really induced a sense of fear when I was reading this part of the paper.

But then, you know, like, as time went and more models came out and it became much more common, you know, we know exactly how those models are built.

You know, it's a fear that really is unfounded, you know, that is really.

Is there if you don't know what's happening.

Yeah.

You know, if you're just in front of a chatgpt thinking, oh, that could take over the world.

But.

And my argument would be, you actually understand how a large language model works behind just playing with it.

Sure.

And I, and I think if you get scared, imagine what the rest of us get when we just see a machine that's learning, that's answering questions that frankly, I'm not sure half the people on earth could answer.

Right.

Like, it is absolutely incredible.

And I think from that standpoint, from the outside looking in, from no knowledge, no background in computer science, no even understanding how coding, it looks amazing.

It looks unbelievable.

And it looks very terrifying for that very reason.

Like, it's just as.

It's just as terrifying as it is incredible because you look at it and you're like, look at all this amazing stuff we can do with it.

But then also look at all the horrible things that could be done with it too.

Like it's like nuclear energy.

Right.

Nuclear energy is incredible.

Use it.

Right.

And it's horrible if you use it wrong or if it's damaged or whatever.

Right.

And I think, I think AI kind of falls in the same thing.

It's a great tool in the right hands and maybe a very dangerous one in the wrong hands.

Of course.

Of course I remember that.

So there's a friend of mine with who I did my PhD and now he's a mathematician, like a professor.

Yeah.

And he texted me once, he was like, I'm Terrified.

Like what's happening?

You like, you know, somebody that you may expect, you know, from, from far away tech person or a scientific person, you may expect that person maybe to be more connected to that type of technological novelty.

And yeah, he was, he was terrified.

And the scare, the fear was coming from the lack of understanding.

And that makes sense.

That makes sense.

You know, when it comes to badly using AI, I was talking to somebody that is, that is working in the Air Force, like in the British Air Force, and he's working in the US and he was telling me, and I may misquote because it's been a while, but he was telling me that we are now in the fifth generation of fighting jets.

Yes.

And the next generation, the sixth generation, the one that is now in progress to be made is the generation where data and potentially AI is going to play a much larger role in the way those jets are piloted.

So AI data or the fact to be able to utilize better the data that is available by sensors that could be placed on the jets.

So data is going to play a much bigger role in, or potentially AI in being able to have a fighting technology that may be superior than to, to the, to the enemy.

So AI is going to be part of the arsenal of, of tools that we can use to, to fight wars.

So that's happening.

And for sure.

Yeah, no, absolutely.

And you know, that was going to be kind of one of my questions that you see almost every new technology coming out now has aspect of it, right?

Whether it be your new smartphone or your new camera or whatever it is.

It's like, oh, look at all these touted new AI features.

Right.

I guess one of the questions for me to you is is it truly an, an advancement or are they just kind of using AI to tout some new features that maybe we've had in the past that were AI anyway and we just didn't even know, you know.

I, I actually hate that thing, you know, like I'm using a couple of tools in my day to day, like for my own work that are software based and they are useful for my work.

And there's a lot of inefficiency in the ways of softwares are implemented and there are things that I would like to improve, but instead of that they are focusing on AI features that are useless.

And we've seen this wave of AI features or AI companies or AI tools that came out that are useless that nobody needs.

And you know, companies now are hiring people to build product, AI product that nobody needs.

And it's Very annoying.

You know, like to see this hype around a real, real novelty when it comes to technology.

To see this hype that is giving a bad image to the, to the technology itself because of the way it's being marketed.

Yeah.

And you know, we've seen everything is AI powered now and every time there's something AI powered, there's an AI feature.

It's usually annoying.

This AI that does that, you usually don't want to use it.

I mean, I'm sure there's some exceptions where some features are useful, but I remember when I use Instagram and meta trained his llama model and now on the search bar you had access to the llama model as well as a way to, to kind of as have a meta like, I mean the chatgpt like from, for Instagram and that was annoying.

Like I don't want that when I use Instagram.

I, I mean I use Instagram to, to lose brain cells.

I am not there to, to, to, you know, to have to deal with that, that talking entity, you know.

Yeah.

So yeah, it's, it's been annoying to me.

Yeah.

And I am, it's part, you know, it's maybe counterintuitive because I'm part of the people that educate the engineers to become good at doing those things.

But still, you know, I'm, I'm very product oriented.

So I feel that, you know, it's important to be aware and to be educated about the technology as an engineer.

But that doesn't mean that you should apply that technology everywhere.

You know, I'm still very excited about what's not, you know, what's in the background, what has been less publicized.

And so, you know, like we have this hype around this technology but, but it's what the result we have, the results, the resulting products we have are not all, you know, that great.

Yeah, yeah.

I would agree specifically with like, you know, in this world, podcasting, audio processing, you're seeing go heavily, heavily, heavily in the eye and you know, probably me and you know that world well and have produced our own shows and everything and I still choose to self produce my own show.

I don't let AI do it.

I do all my editing on the back end.

It's Kelly Kennedy made and Kelly Kennedy listened to you because I want to make sure that it sounds good.

Because for instance, if I was to use like, I'm not going to name the company, but the company we're recording this software on today and then hit the button for the record and said, I want the audio to be AI analyzed.

It comes out sounding really bad and I don't know whether like I don't know who at their side the engineer that was like, yep, that's good.

Move forward with that because it does not sound good.

It sounds so much better to self produce.

Not like, you know, like you mentioned.

I think there's a lot of things out there that were maybe released too soon or just trying to capitalize on the moment.

Yeah, I'm using a tons of ton of tools that, that we are trying to, to bet on AI and it's annoying like not going to name it but I'm using this video recording tool that I use to make my courses and they are so into AI and everything is AI.

It's like I don't care.

I want you to be able to have a play button that works well, you know.

And it's you know, like sometime, you know, just don't focus on the AI.

Just focus on the basic things.

It's okay, you know.

Yeah, no, I know, I know.

I think we'll get back there.

Like I said, I think there's people that are just like it's the moment, everybody's excited about it, let's capitalize on it.

I think the part that I get really frustrated with is that specifically there's multiple programs that do the same thing.

And then one other thing may be good that I want to keep.

Like there's an AI feature I actually like but then I end up having to buy like eight programs to do one task that one program should be able to do because each one tries to do everything but only does one or two things Great.

You know what I mean?

I see completely what you mean.

I have a bunch of product, bunch of tools as well like that.

But I pay and I don't need most of it.

I know it's super, super frustrating.

Anyways, we got on a long tangent here and I really wanted to narrow in on your life, you know, take me back.

How did you end up on this incredible path?

You've had an incredible career, you're still doing incredible things.

Walk me through it.

How did you end up on this journey?

It's funny that you say that because at every step of the way I felt I was failing.

I being you know like later three times in my career and I have a short term jobs because of that and, and on paper, you know, like at the time didn't look good.

Now I feel very confident into my, with my skills, you know.

But at every step of the way I felt I was like in the wrong place or it didn't work out as expected, you know.

But in the end, in the long run, you know, if you look back, you know, things are smoother, you know, like so they seem, they seem more, they seem maybe, you know, the good things, the, the positive points accumulate and they start to shine.

But that was not the case when it happened.

So I did a PhD in theoretical physics.

I was specializing in, so I was in between the applied mathematics department and physics department specializing in the mathematics of turbulent flows and I was doing a lot of data analysis.

So I was analyzing data that was coming from petabytes databases.

I was somebody that was strong in mathematics, strong in computer methods and that were doing data analysis on a daily basis.

For me to move to machine learning, that is a science where you are using computer computational methods and from, you know, by using data, it was a very obvious switch.

I always prepared myself actually to go on Wall street to be a quantitative trader.

I ended up to go on the west coast to follow my wife and my fiance at the time and I ended up to be in the Silicon Valley.

And I was a data scientist.

It was an obvious transition from being a theoretical physicist that were used to do data analysis.

The mathematical models were very similar.

Actually the mathematical models used in machine learning were extremely simple compared to physics.

So for me it was an extremely simple transition and the goals were different.

I was feeling the same about quantitative trading where it was very similar to physics on the mathematical aspect and the computational aspect, but the goal was to make money instead of publishing papers.

And here with machine learning I felt I was finding very similar goal.

So I, I, I moved from places to places being a data scientist and at some point I switched to being titled a machine learning engineer.

Really, you know, every company was for me an opportunity to learn a bit more about how businesses are making money and how machine learning can help with that.

It was very, a great learning experience because when you're academic person like I was, you have no clue about how money is made and you have no clue how mathematics or mathematical models can help to do that.

So I, you know, that, that learning was great and it was a long learning curve for me.

And my last real job, my last 9 to 5 was at Meta, which, which I hated, I hated working there.

I really hated the culture, I really hated the, the teams or the, you know, like I didn't like the, the way we were working there was not enjoyable to me.

But you know, I was doing some of the most advanced machine Learning that there was to do in the world.

I was working the ADS ranking so the team that was generating 95% of the revenue at Meta.

So there was a lot of optics on, I mean there was a lot of eyes on our work.

Our models were the ones that if you have a percent improvement it translates into hundreds of millions of dollars.

So yeah, it's, it was so many people with the same title than me trying to really improve on those models, trying to make more money and then being in, in that place where people consider to be the best place to work as a tech person, you know, and me hating it.

I, I thought that the only logical transition was to become an entrepreneur because I could not anymore work for somebody else and started to partner with a couple of people to try to find a startup.

It didn't work.

You know, I didn't maybe push as much as I could, but at the time I transitioned to be more of a simpler entrepreneur.

So you know, there's like I think this new wave of entrepreneurs that are not startup people.

So there's usually this dichotomy between startup and, and non startup big tech.

But as an entrepreneur there's also this additional option which is to choose a simpler, smaller, less scalable option where I'm trying to monetize my, my skills on the education side on the consultancy, you know, like being consultant, these kind of things which is not scalable but which allows to have a very enjoyable life without targeting billions of valuation when it comes to the company that you are building.

And that's the type of entrepreneurship that I'm currently pursuing, which is fine.

I'm still itching for potentially build a scalable product that I could sell within startup entity.

But right now I'm having a lot of fun, you know, like growing as an engineer, helping people grow as engineers and trying to really educate people.

For me it's as well an amazing learning experience because having to be on top of everything to upskill people require for me to learn about everything 10 times faster than anybody else.

And this is something that I really enjoy.

I became what many people would describe as a unicorn because I had to for being an educator.

I had to be the guy that knows everything well to educate other people.

And I love this feeling of, of being on top of things.

And at the same time I think this knowledge makes me realize that there's so many things I still need to learn.

Yeah, to, to be, to, you know, to, to, to continue in this, in this path of being an educator.

So that's, that's for the moment.

I really love it.

Yeah.

I was going to say, do you get a lot of passion and enjoyment from teaching?

I know that for me that really, I didn't realize how much I was born to be a teacher until I started doing teaching and coaching on business development.

I was like, oh my gosh.

Like, this is what I, this is what I was meant to do the whole time.

I had no idea how much enjoyment would come from it.

But it's.

Do you feel the same way about your teaching and coaching?

I do love teaching.

I've been teaching for a long time.

Actually I was doing my PhD.

I taught the whole time I was a TA and I actually taught as well.

When I was in France and I was teaching, I had a, for half, half a year I taught at the high school level.

Oh, wow.

And I, I also taught here a course in, as a local university at the master level for data science.

So I had a lot of opportunities to, to teach as an expert in the field and somebody that needed to do it for, you know, when you're a PhD student.

I was a theoretical physicist.

I mean theoretical.

On the theoretical side, there's not a lot of fundings to found PhD programs.

I mean, on the surgical side, you know, there's not as much funding.

So people tend to be TAs to pay for this kind of program.

So that's what I did and I learned to become good at it and now I really enjoy it.

And I found, you know, it's not a passion as much as I spent years thinking about it, to become better at it.

And it's not a passion as much as I've understood some of the logic on how to teach to make it better for people.

And I kind of enjoy getting better myself at that.

I kind of enjoy using those tricks that I've learned.

It's a lot more about doing what I'm good at than doing something that I'm passionate about.

Yeah.

And pursuing that what I'm good at.

I'm good on the technical side on what I'm teaching and I'm good at teaching because I've been doing it for years.

And so merging the two became much more of a way to express myself on something I'm good at.

Yeah.

Yeah.

One of the things when I was kind of digging into you was all the courses that you actually have and you have quite a few of them.

Walk me through what was that process like, like in creating these courses.

I imagine that must have had its moments.

The logic is simple.

As somebody that was in this entrepreneur mindset, trying to monetize.

So you know, like I have a mortgage to pay, I have two kids and I had to make money.

So I was really convinced I didn't want to work for somebody else and I didn't want to, to go back to work for somebody else.

So I had to find ways to make money.

And when you have this need, you find so many ways or you explore so many ways to, to monetize your skills.

And it's something that when you're an employee, you, you don't understand very well.

You know, you don't understand what it means to try to find ways to make money.

I could make money by being, by doing, by being a speaker.

I could make money by being consultant.

I could make money by being, being a coach or by being a mentor, you know, like helping for interviews.

And I ended up to, to run a newsletter.

I got actually some investment.

An investor was helping me to, to get through the first six months to make sure I could get the newsletter running and get enough subscribers to monetize the newsletter.

And I realized that if I were to continue this way, I, I would have difficulty to pay my mortgage.

So I started to try to find additional ways to make more to make money.

And, and one thing that came naturally was let's try to do some courses.

They, there's some, there's Udemy.

So I tried to put a course on Udemy and I did a course on LangChain if you're familiar with it, which is Orchestrator framework for LLM Pipelines.

And then I did another course on Introduction to Transformers or Large Language Model with Transformers.

And then I did another course on machine learning fundamentals.

And so I moved from courses, things that you can find for cheap online to cohort based courses.

So boot camps where I was having this live interaction with people and I was providing projects for people to solve.

And I really wanted to bring people, you know, at the level of the job itself.

I wanted to give them a sense of the difficulty of the job itself.

So it was important for me to give them projects that were hard and, and similar to the ones we find on, on the job.

So and I, I, the last bootcamp I did, I taught was a bootcamp on large language models to learn to train, fine tune and deploy large language models.

And I'm preparing another bootcamp on Introduction to Data Science and Machine Learning, another bootcamp also on Orchestrator pipe frameworks and also agents to build LLM applications.

There's a couple of, there's a sense of boot camps that I'm thinking about building.

I realized that there's a demand for this.

I have the skill to make something of good quality.

So I felt it was a good fit for me to try to provide this service, something that I can do well and people are happy to receive.

So I found it's a good fit so far.

And like I said, I've been very freed in my entrepreneurial, entrepreneurial journey.

So this is true now.

It may, it may not be true six months from now.

Sure.

Yeah.

That's fair.

Man.

It's changing so quickly.

I, you can't expect anybody to commit to anything longer than about six months at a time.

For sure.

For sure.

Do you, do you consult with like large organizations who might want to create their own internal large language models?

So I had a couple of consulting gigs, but every time the deal was you need me for a short amount of time every week, that's fine.

But I didn't want to derail my current business model to move towards something that is more consultant based.

And I didn't want also to be a contractor.

I didn't want to be the technical guy that is actually doing the thing.

I wanted to be potentially working on the strategic aspect, you know, like how, you know, maybe helping to build the teams, maybe to lead the teams, but not to actually implement things myself because I don't want to, to be in that, that employee situation that I tried to flee a few years ago.

Yeah, yeah, you're absolutely right.

You can definitely end up trapped right back in it, even as a consultant.

And you know, like there was a point where I was doing a, I was trying to become more of a consultant.

Yeah.

And I ended up to find myself doing interviews to be hired as a consultant all week long.

And I hated that.

I hated that idea that I was trying to flee being, you know, I was trying to flee renting my time to other people.

And then I ended up to spend my time in interviews to see if people were going to hire me, which.

And it didn't, it didn't really make sense.

So I, I can do it if there's a good fit.

But right now I'm, I'm, I'm trying to make sure that it's not taking place too much of my time.

Yeah, well, you know, and that was going to be one of my questions that ultimately you've actually created quite a bit of a personal brand around yourself.

And I wanted to chat about that because 2024, so many people have chatted about the Power of personal branding.

And so every once in a while, when I have someone like you on my show who has an incredible LinkedIn following, I have to ask, talk to me about that.

How did you go from, you know, AI expert to building such an incredible personal brand?

Well, you know, like, it's relative, right?

You could say it's big, but I'm still comparing myself to bigger people and.

Yeah, well, for me it became, you know, like I was working at Meta and I was so bored at Meta, you know, like I was not doing what I enjoyed.

So I felt I was talking more about machine learning on social media than I was at Meta itself.

So I ended up to, you know, like, I never been a social media person.

I never was good at it, I never liked it.

So I remember, like seeing a couple of people on LinkedIn advising how to write as a, you know, as a.

On.

On LinkedIn and on social media.

And I didn't even understand the point at the time, you know, why, why would you do that?

Why would you waste your time doing this?

And I started to let myself go and write on LinkedIn, giving advice about machine learning and trying to opinionated advice.

And I was at the time a meta engineer, which had its own hello of fame.

And people were looking at me as being a representant of Meta.

I was talking for myself, but people were looking at me as being somebody that achieved something in, in the career because he was there, which was fake because it's just, you know, I just passed the interviews and I was there.

That's it.

But for some reason, you know, there was a.

Such a shiny aspect to being an engineer there at the time.

It was more true than now, I believe.

And every time I was talking on social media, people were very, you know, in an opinionated manner.

People were agreeing with me or disagreeing strongly with me because I was a representative of Meta, a company that had, you know, many people didn't like.

Everything I was saying was things that were meant to be debated and fought.

And so I started to get this following on social media because people were, you know, like, I had this image of the Facebook engineer and fine, you know, like, I did that for a couple of months.

I grew quite a bit and then I stopped because I tried to build a startup and I was focused on that, you know, 80 hours a week and 80 hours a week.

I didn't want to waste my time on social media.

It didn't make sense for my business model to spend my time on social media, really.

I was on social media Prior to that, because I didn't like my job and that was a way for me to talk about something I liked.

And I was talking specifically about machine learning and you know, and then, you know, when I realized that the startup I was trying to build was, was not going to work, I, I needed to pay my mortgage, find, find to, to a way to pay my, my bills.

And I thought that trying to continue my social media adventure and trying to monetize this influence that I was having on some people could be a good way for me to make a living without having to rely on renting my time to an employer.

So I continued with this business model.

I met a few other influencers.

It's funny because we tend to gather together.

I mean, we tend to meet each other and we tend to be colleagues of the same work.

Even then sometimes we don't even talk.

We never meet, but we feel like, it feels like some of those guys are my colleagues.

But I met a few of those people and from there we tried to find different ways to partner and to try to find a way to make a living out of being experts that could present ourselves, could build an image on social media.

Social media for us is really a marketing channel.

It's also something that we enjoy doing.

We enjoy being able to talk about the craft as well as being able to use it as a way to showcase ourselves, display ourselves as experts.

So, you know, that's, that's has been the story about how social media became part of my business model.

Yeah.

And, but obviously, you know, it's not the end of it.

It's just a really a marketing channel for, for all of us and we are trying to find additional ways to monetize this and there are many different ways to try to do it.

Yeah, well, you've done an incredible job and I just want to, I just want to say that you've done an absolutely incredible job.

And you know, I know I have a lot of listeners right now who are maybe trying to build their own brands.

They just launched their own company and they're trying to do this stuff, but they're not really sure how.

What advice would you give them to build their own, their own personal brands on LinkedIn?

I'm not sure, you know, like, I've seen people teaching other people with less followers than me teaching how to be to build their own brand on social media.

And I was very surprised on how convinced or how educated they seemed to be.

Even then they, they did less good than me.

And I had this feeling like I Had no clue how to do it.

Something that I, I.

So I have this feeling, I have no clue on how to do it.

But something that I try to, to hold on to is stay true to myself, you know, like try to keep on voice.

And something that I believe is true is that on social media, in my specific niche, there are very, very few people that can compete with me.

Yeah.

Because I have my own style, I have my own way to, to talk, my own way to write, you know, on social media, present myself.

And also I have my own, I have an expertise that is difficult to match.

And so I'm able to talk in my own way.

I'm able to have this, I have this expertise that allows me to be very, to have a very different image.

I mean, very specific image that is difficult to match on the expertise side.

And also I'm specializing in trying to take something that is complex, like anything you find in machine learning, and try to present it in a simple manner on social media.

The idea of social media is like, especially on LinkedIn, you have 3,000 characters to present a subject.

And it has to be entertaining, it has to be educative.

And I use for me, because social media has been so much part of my business model that I've learned to try, I mean, I've learned to condense information in a way that is entertaining, that is educative, that is expert.

You know, I'm presenting something that needs an expertise to be presented, but also that seem demystified and simpler because of the way I'm trying to present it.

The advice, it's hard.

I see I'm having difficulties now growing, so it's not like I know exactly how to do it.

But the advice, at least the advice that I'm following myself, is that I'm trying to have my own voice and I'm trying to be different than other people.

That means that when you see other people educating other people on social media to teach them how to gain a following, I'm very comfortable with the idea that I do anything but what they're advising because anything that would be a pattern that everybody would follow would end up to be something that is not unique.

So I think in social media it's important to keep your own unique voice, to, to keep your own thing that is making you special in the eyes of the followers.

Yeah, no, it is, it's very interesting.

It's, you know, I mean, I've talked with quite a few people with massive followings.

Liz Ryan is one of them.

Lou Adler is another and yeah, just.

It seems to be.

It seems to be.

You're absolutely right.

It seems to be.

The advice is, speak from your heart, be your own person, be unique, be individual, be honest, be truthful.

You will build a personal brand.

And that really does seem to be like the advice from everyone that I've talked to, including yourself.

It's, it has worked for me.

And I've seen people being very bland, being very, you know, in the way they were writing on social media, being very boring.

So following the tricks and doesn't work.

And I mean, the, the more, the more unique you are, the more, you know, the more people want to hear about you.

Yeah, no, absolutely.

Damien, this has been absolutely incredible.

Thank you for coming on today and teaching us all about, all about AI and machine learning.

Honestly, dude, like, haven't had anybody like you.

You are, you are one of a kind, my friend.

Well, thank you.

Thank you for that.

It was, you know, I love talking about machine learning.

You know, like I said, I'm an educator, so the opportunity for me to be able to have these kind of conversations, really fun.

So, you know, if you need me again, you know, I'm happy to come back and.

Yeah, as much as you want.

Well, I'm sure as time goes on and more, I'll have more questions, no question.

And I'll be like, okay, I got to bring them back.

I need my expert.

For sure, for sure.

Before we do that, though, before we close out today, I know there's a lot of business owners listening who are struggling with machine learning.

They may need help.

They may need a machine learning consultant.

Can you please go over all of the services that you offer and how people get a hold of you?

If you're asking me, you know the services I offer on the consulting aspect, on the consulting side, I can help build a team.

I can help build a strategy around what should.

What maybe is needed.

I can help maybe about designing product that could be ML powered.

I can help maybe into guiding teams, leading teams.

I can help about how to get them started.

But again, I'm trying to limit the time I would spend on doing that.

Although those services are, I don't close them, I don't make them unavailable.

I'm open to the conversation.

I'm trying to not advertise those services too much because it's not something that I want people to.

I don't want to find 20 different emails about people wanting me to be a consultant on some ML product because I won't have the time.

Sure.

But, you know, I Love to have the conversation in and if it's a good fit, if I find people that need my help and I love what they're doing, I think I would be willing to really carve time to, to actually work on that because that can be, that can be for even my own growth.

You know, being, you know, to keep working on products that users would use is something that I enjoy and it would help me also continue to be connected to the business aspect of machine learning.

So that could be, that could be something that I would be quite tempted to do on a limited basis.

Amazing.

Amazing.

Okay, can you specify maybe what types of passion projects those might be?

What are you passionate about?

That way maybe we can really narrow down who reaches out.

I love to be able to help on LLM projects.

I love to be able to help on LLM pipelines, which is very different.

People are confusing the two.

So orchestrators, orchestration around LLM pipelines, building agentic workflows.

This is something that I enjoy.

This is something that I'm pretty good at and it is something that also would love to have a bit more product experience building an actual product that people like actual users would use.

Because things are moving so fast, it is easy to be lost on the theoretical aspect of things and to forget about the users.

And I always want to be able to learn things and educate things in a way that are product oriented and not theoretically biased.

Okay.

Okay.

Amazing.

And just before we wrap it up, bring us in quickly to the AI edge and how people sign up.

Well, it's the URL is Zai Edge, the AI edge IO and but to get to the newsletter, you will need to add the newsletter.

Zaih IO zaidge URL will direct you toward my bootcamp courses, you know, which may not be for everybody.

So it's very much for engineers and you know, I'm targeting those people.

The newsletter itself might be more friendly to non technical people.

Although I'm not making things easy.

I'm not the kind of guy that you go to if you want to see hype and prompt engineering.

I'm not that kind of guy.

I will make things difficult.

I will try to make it look easy, but I will dig into the details.

And I'm targeting people that want to be on the engineering side or want to understand the engineering side, want to understand the in and out, the details.

So that would be my focus.

So newsletter theiedge IO would be for the people that are ready to be active in their learning.

Amazing.

And for those of you listening, it will be in the show notes for this episode.

So if you're wondering where to find it, I will make it very easy.

It'll be in the show notes on this episode, wherever you guys listen.

And yeah, if you want the technical deep dive into machine learning, this is where you find it.

Damian, it has been an absolute honor chatting with you.

Thanks for joining me today.

Well, it was a pleasure, so thank you for inviting me.

Until next time, this has been episode 266 of the Business Development Podcast and we will catch you on the flip side.

This has been the Business Development Podcast with Kelly Kennedy.

Kelly has 15 years in sales and business development experience within the Alberta oil and gas industry and founded his own business development firm in 2020.

His his passion and his specialization is in customer relationship generation and business development.

The show is brought to you by Capital Business Development, your business development specialists.

For more we invite you to the website at www.capitalbd.ca.

see you next time on the Business Development Podcast.