Natalia Burina, a prominent AI product leader and venture capitalist, shares her incredible journey from escaping war-torn Bosnia to becoming a leading figure in the AI field.

With a background that includes key roles at Meta and Salesforce, Natalia has a wealth of experience in product management, particularly in developing AI technologies that cater to a global audience.

The discussion delves into the challenges she faced in creating fair and unbiased AI systems for billions of users, emphasising the importance of understanding cultural nuances and the historical biases present in data.

Natalia highlights the need for transparency in AI and the ethical implications of technology, particularly how biases in data can lead to skewed outcomes in AI applications.

Her insights reveal how companies can proactively address these biases through tools like fairness flow, which helps engineers assess and mitigate potential biases in AI models before they go live.

Takeaways:

• Natalia emphasises the importance of addressing bias in AI systems to ensure fairness and representation across diverse populations.
• She highlights the transition from building AI products to investing in AI startups, focusing on both product and business aspects.
• Natalia discusses the challenges of working in a male-dominated tech environment and the necessity of finding supportive mentors and sponsors.
• In her experience, ethical considerations in AI involve transparency, particularly in ad targeting and user data usage.
• The future of AI is seen as promising, with potential applications that could revolutionise drug discovery and productivity in various sectors.
• She encourages individuals interested in AI to find their passion, learn through hands-on experience, and utilise available online resources.

Connect with Natalia Burina on LinkedIn

AI Studios by Natalia Burina podcast on Substack

Natalia Burina on X (Twitter)

MIT (Massachusetts Institute of Technology) - free online courses OpenCourseWare

Coursera online courses

DeepLearning AI courses

BBC article about how groups that support women in the tech industry are struggling to survive.

Companies mentioned in this episode:

• Meta
• Facebook
• Salesforce
• Microsoft
• Samsung
• Storm Ventures
• Nvidia

Hello and welcome to Women with AI, a podcast dedicated to amplifying the voices, insights and expertise of women in the ever expanding field of artificial intelligence while engaging in conversations about the impact of AI on gender equality, representation, and more.

I'm excited to welcome someone to the podcast who is an AI product leader, investor and entrepreneur who has previously worked for Meta, the social technology company that formerly operated under the name Facebook.

And she was actually recommended to me by one of our previous guests, Doctor Eva Agapaki.

So you can check out her on episode 13.

But before we get started, let me tell you a little bit more about who I'm speaking to today.

Natalia Barina is a seasoned AI product management leader with a proven track record of setting strategy and delivering business results.

Natalia invests in early stage b, two B AI companies at storm Ventures.

Prior to storm, Natalia led multiple teams at Meta AI, including video, understanding, privacy and transparency.

Before Meta, Natalia helped to define early versions of Salesforce Einstein and helped to launch Microsoft's bing search engine.

Natalia holds a degree in applied and computational math.

She's also got her own podcast, AI Studios, where she's had conversations about AI with the world's top AI builders, artists, researchers and thinkers.

So I'm very much looking forward to learning from her experience with that as well as all things AI.

Natalia Barina, welcome to women with AI.

Well, thank you for having me today.

I'm very excited about this conversation.

Great.

Well, before we start talking about everything that you are involved in, please can you tell us, well, tell me in the audience a little bit more about yourself, your journey to where you are now and how you got into working with AI.

Yeah, so I was born in Mostar, which is Bosnia Herzegovina, and as a child nearly escaped the war there.

And this was possible because my mom was able to get an engineering job in the US.

So those early experiences really meant that they made an impression on me.

And there were a couple of things that I took away.

So one, our world is very delicate and anything can happen at any time.

We saw this recently with COVID And two, engineering skills are a means of survival.

So after immigrating to the US, I was very driven to succeed and in particular to really get to a place where we had a safety and stability in our lives.

And so I studied applied and computational math at University of Washington, and I absolutely loved it.

Though honestly, I'm probably inherently more inclined towards more creative professions.

Initially, I gravitated towards literature.

I just read everything in sight.

I also love music and dance.

So somewhere in an alternate universe I'm probably a writer, a pianist or a ballet dancer.

But I ended up in tech and I managed to stay in tech for many years.

Now there were some periods of reluctance and doubt, but overall I feel very thankful and grateful that I was able to catch this wave of incredible tech growth and innovation.

And then secondly, I happened to be at the right place at the right time, so very lucky.

Now you ask how I ended up working with AI in particular.

And while it may be a little bit debatable, depending on who you ask, web search really is an AI application at enormous scale and it has a lot of the attributes that you really need to make AI work.

So to be specific, search engines use natural language processing to understand intent behind search queries.

Anytime you type something in the results that you see in a web search page are there because there are ranking algorithms that incorporate AI to determine the order of those results.

There's many, many other ways that AI is used in web search engines, but that was what got me started.

And from there on I just kept getting jobs that were related and I progressively grew in my career and built out that expertise.

Cool.

And so you've transitioned from, I guess, building AI products at companies like matter and Salesforce and to investing now in AI startups as a venture capitalist.

So kind of what motivated that shift?

And does your background influence your approach to the investments you're making?

Oh, absolutely, absolutely.

The shift was serendipitous.

I didn't stand out.

I didn't set out to become a venture capitalist.

I was super burnt out and I just took a little bit of time to recharge and focus on my health and family.

And as wonderful as the break was, I started a podcast and I wrote and I was tracking industry developments.

And I really like it.

Gave me the opportunity to look at the big picture.

When you're working at Meta or Salesforce, one of these companies, or when you're deep down in the trenches building AI products, these jobs are very consuming and you have to develop specific product strategy.

You have to execute on specific goals.

You have to keep your teams healthy, but you don't have much time to think about what's happening from a big picture from a point of view, from an industry point of view, having taken that break, I was able to take some time and to think about that.

I met the team.

We knew people in common.

We started talking.

They liked my track record and background as an operator building AI products.

And I wasn't sure if investing was for me.

Again, it's not something that I had sought out.

And maybe, I think there's probably a bunch of people who don't, maybe they're not that familiar with venture.

So I thought it would be helpful to break down what exactly a venture capitalist does.

And there's really four things.

One, you source deals, so you find potential investment opportunities.

Two, picking deals.

I think this one is probably the hardest, deciding which companies to invest in.

Three, there's the process of closing a deal, so you have to agree on the terms with the companies.

And four, you have to support your portfolio companies, where you connect them with candidate customers, employees, you give feedback, etcetera.

And so for me, the biggest reason for the shift was really the opportunity to learn what successful companies look like.

If you're building product, you focus on the product, but you rarely think about the business unless you're super.

You're like an executive or a very high level executive in a company the size of meta.

One of the lessons I learned earlier in my product career, it's actually not enough to build a good product, you have to also build the business.

And so vc for me is the opportunity to learn things from a business side and get this unique view where I get to see a lot of companies in different markets and at different stages, and also learn how to assess a business, what does good look like?

It was also important to me that the team was amazing.

And then to address the question of how my background influences my approach to investments, I always think about investments from the product of a builder.

I know I've seen what works, I've learned hard way what doesn't work, and investing in b two B is particularly interesting because b two B in a lot of ways as well behind an AI when compared to what big consumer companies have built out, you may wonder, like why is this the case?

Consumer products are built more in a bottom up way.

New features are derived from looking at large volumes of data and user feedback.

And b two B companies rarely have this kind of infrastructure.

So I feel like b two B is an enormous opportunity and b two B has yet to kind of get there.

It's moving there and so from an investment perspective, it's very interesting.

Yeah.

Because I guess, yeah.

You can help shift them in the right direction with all your experience.

Exactly.

And we know that, you know, b two B, building products in b two B is an entirely, it's very different from consumer.

A lot of it is very customer driven as opposed to data driven.

And with AI coming up, all of those companies will have to become more and more metrics and data driven.

And that's where I excel.

And so I feel like I can add a lot of value to potential companies who come for investment in our portfolio companies.

Yeah, like I said, well, at meta, you worked on AI technologies for a global user base of around three, 3 billion people.

And that, that just seems mind boggling.

So like, I mean, can you share some of the challenges you faced in creating AI that can be fair and unbiased across such a diverse range of populations?

Yeah.

So, you know, when you have a product that touches 3 billion people, it almost feels like you could, you could sneeze and hit millions of people.

It's just like insane.

You make small changes and you still impact billions of people, which are entire countries.

A funny anecdote about this one is Facebook.

A lot of the times before they would launch something entirely to production, they would roll it out 100%, they would test it out on smaller countries.

So New Zealand was a good country where a lot of things were tested, but the reality is New Zealand still has a population of 5 million people.

So you test it out, you put it there, but you still impact 5 million people.

And so you have to be very careful and mindful of what you do in the massive scale because there's such a massive user base.

It really means you're spanning different cultures, different languages, different backgrounds, they all have their own context and operate in a very different way.

We all know that certain cultural practices are very different from country to country.

And so there's so many ways that things can impact people in a way that's not fair or is biased and so on, but I'll just name a couple.

So one is you have to really, there's a lot of bias that comes from data.

AI runs on data.

Data is its lifeblood, but data a lot of times has historical bias.

So what you use to train AI model often reflects bad things from the past.

I'll give you an example.

In the past it used to be the case that most doctors were men.

Now it's like 50 50% actually think there might be more women.

But if you use a lot of that historical data.

AI learns from it and might predict that doctors should be men and are men.

And this is one classic example.

The second one, data.

It's hard to get data that represents literally the entire user base, so AI can perform poorly for certain groups.

An example of this was meta used to have an AI powered smart camera.

I actually don't know if they still have it.

There was a product called portal, and this AI camera would prioritize zooming in on white men because that's where it had most training data, so it wouldn't be able to recognize people who had brown skin or darker skin tones.

And this is something that the company had to go correct.

There's data bias.

There's bias that can happen in algorithms.

AI can pick up ways that people engage with the products that may not be, that may reinforce existing biases.

Again, cultural nuances around languages.

There's concerns around privacy, and then there's concerns around ethical consideration.

And each one of these, we could probably do a separate podcast on them.

There's so much complexity.

It's when you're dealing with the population of literally that spans the entire world, there's just so many different flavors of problems and things that can go wrong.

Yeah, because that's what I've been learning.

So each time I speak to someone new or talking about AI or just reading around about it, it's that bias fact that everyone's going, oh, it's great, or it's done this or it's not done that, or it's doing this and it's like, but hang on, we need to be feeding it the correct data and making sure it's.

And correct as in kind of making sure it's not just all the bias from previous times that have.

That are leading up to now, because your right times are different now.

But if it's just, you know, learning from everything that's come before because it, you know, I mean, it's like picking up on.

Yeah, it's a white mandev.

Because I spoke to someone previously on the podcast that was something to do with eye tracking, and it was only recognized blue eyes.

So the eye tracking for the, you know, it doesn't now and it's been changed, but it's the same kind of thing.

It's just whoever's invented it just kind of models it on themselves.

Yeah.

And I get, you know, we're, I mean, this is women with AI, so we're speaking about, you know, gender and that kind of thing because I guess there's all that bias there.

I mean, so as a woman, like, with your background in AI and now being a venture capitalist, I mean, do you feel that that has been a sort of male, is it still a male dominated field?

I mean, what are the challenges that you, that you're facing?

It is, but it, I, it has changed so much for the better than when I first started.

When I was first starting out, I just, the big challenge for me was getting used to an environment where I was the only woman or one of a few women I remember one of my first jobs, I don't think I saw any women in the bathroom for months.

And so it's just like, one is getting to comfort two.

I think the biggest challenge really, especially early on, was getting sponsors.

There's mentors.

I feel like getting mentors is easy.

You can easily get people to talk to you and to help you.

But getting sponsors is a little bit different.

A lot of times what sponsorship means is that somebody is willing to open up doors for you and give you new opportunities and really vouch for you.

And a lot of times sponsorship is people see earlier versions of themselves and people that they sponsor.

And for me, it was very difficult to approach men who were much older and to figure out how to build relationships.

And so I feel like sponsorship is one that, for me was particularly hard.

I think it's still a challenge when people who are in leadership are predominantly male.

But overall, I, you know, I feel, I feel extremely lucky.

I feel like I was kind of at the right place at the right time in the right field, taken a lot of advantages of this tech wave, so I wouldn't change anything if I could go back.

And I also think it's always important to frame things in a way that's positive and advantageous.

I think there's a lot of attributes and thinking that women can bring to the table that really make things better and improve them.

And so if I was to give advice, I would say, be mindful of the relationships you build over time.

Make sure you cultivate them with high caliber people, people who are super smart.

I think it's always follow the people as opposed to work on a particular problem.

That's really what works.

You need to have technical skills, but you really need to have relationships with people.

You need to cultivate those relationships.

That would be my advice.

That's great, because I think I was reading something earlier this week that was, I think it was on the BBC website.

I have to try and find the clip, but it was just saying that there are women leaving tech because there's just not enough support.

And it's that kind of like that funding for women.

Because you're right, if it's just predominantly sort of, you know, male venture capitalists or men that are kind of investing it, they're just going to invest in what they know, what they recognize.

So it's kind of, we need to, yeah, encourage more women to just, yeah, follow it, I guess, and get into it and stay in to the field and, yeah, support each other.

One other thing I'd say is I feel like tech has been very flexible and forgiving.

And so this has been, especially now that I'm a mother, I really feel that it's helped me and it's worked out well, giving that flexibility where you don't have to work on any kind of a set schedule, you can take care of family responsibilities.

And again, I think I've been lucky that in my roles my managers have been very supportive.

So I guess that's the other component is find people who are supportive and see the magic in you and see that you have superpowers.

And I would always err on the side of following those people as opposed to something which makes, uh, maybe gives you more money or it looks like a better, like, you know, working on a shiny new thing.

Yeah.

Like, and that I can sort of makes me think about not just like, um, bias, but also ethical sort of concerns, because I think you worked in like a responsible AI team or sort of trying to make sure that it's ethical and fair?

Like how do you, how do you ensure that the sort of AI products that you're building or the ones that you're now investing in are ethical and fair and are there any kind of examples you can give about how AI ethics kind of, I don't know, come into play or play a role in development?

Yeah, this is a huge topic.

And part of the reason why it's so difficult to talk about is when you talk about AI, you could, might as well be talking about software, right.

It can literally mean anything.

There's so many different problems that you could solve with AI or software.

And the way I think about it is really, first we had machines, then we had software, now we have AI.

So it's a natural progression of things.

And some of these issues have come up before, but it's really have been exasperated.

Exasperated exerberated.

By aihdem at Meta.

One of the tools we had was something called a fairness flow, which was a diagnostic tool to help machine learning engineers detect potential bias in AI models and labels used at Facebook.

And bias can happen, and it can creep in so many different ways.

But basically what this tool did was it helped to measure bias.

You can't address bias and fairness unless you think about define what it means and you measure it.

Right.

The second thing is it would assess label bias.

So this all goes back to the data that goes into AI systems.

Labels of data labeling is a crucial part of that makes the AI run.

So it's important to examine the labels used to train models.

And if these labels contain biases, the resulting model might be biased as well.

And finally, the tool had ways to visualize how things were working and metrics that helped engineers understand the nature and extent of the bias that was detected.

There's a lot of benefits to do this if you can detect bias earlier on, before you're built.

You know, as you're building the product, it's much easier to fix and fixing them after the fact, after everything has been deployed, you take greater risks, but it's also much more difficult to fix.

And by understanding and addressing bias, engineers can really create models that are more equitable and inclusive from the get go.

So, a specific example where Facebook used fairness flow was in the ad targeting system.

This was huge.

It impacted many people.

But to simplify it, basically, there's Facebook.

The business is primarily predicated on the ads engine.

They sell a ton of ads, and you have to think about what kind of ads people are seeing.

So an example might be that ads for job opportunities, for good job opportunities, might be shown disproportionately to mendez, while ads for beauty products might be shown more for women.

Right.

So this is clearly an example of where there is bias.

And so Facebook went back, implemented the fairness flow to analyze their ad targeting algorithms and figure out, this is one example.

But there's so many more ways that things can go wrong.

And they were able to detect potential biases across the board, measure the impact, and then third, ensure transparency.

Transparency is something that's coming up more and more in AI products.

It's important for people to understand why AI is doing what it's doing and to have visibility into how their data is used.

So providing transparency to users about what's influencing ad targeting in this particular example really helps reassure people.

And it's something.

Again, I'm seeing more and more adoption of transparency in AI.

So that's one example.

But again, this is such a broad question and there are so many examples of things that can go wrong, especially when you're talking about 3 billion people.

Yeah.

Because it's like otherwise it feels like AI is going to just start telling people how to think.

Because if you're only ever seeing one thing, how do you know that, that there's more out there?

You know, if you're only seeing beauty projects or if you're only seeing this kind of job, or if you're only seeing that kind of holiday or these kind of products or.

Yeah, it kind of is taking the choice away, isn't it?

In a way, it thinks it's helping us or, you know, whoever's designed it, because again, it's a kind of, you know, it's not doing it.

It's, I say it's the data that we're feeding into it and letting it predict from just, I mean, I never accept the cookies anymore.

We had some kind of like it training at work and it was saying it's tracking everything you do.

Just never accept, you know, always reject everything.

So I'm always doing that.

And the other day a friend of mine had her phone and I just went to like, reject talk.

She said, oh, she's like, I always accept everything.

Yeah.

I spent, I spent a lot of time thinking about AI privacy, and so I like you.

Also, I'm very mindful about who is asking for my data, what kind of data it is.

And I think for, you know, for certain things, not a big deal.

It's okay.

Like, I wanted to understand what I want.

Like if I'm shopping for dresses or something, I actually like that AI knows what I'm looking for and what I want, so it can give me better suggestions and it can help me shop.

I don't, I don't know about you, but I do most of my shopping online and I never thought I would be that person because I genuinely enjoyed shopping in person.

But nowadays it's just like, ah, too many things do.

So that's an example.

I actually don't mind, like, what is the worst thing that can happen?

It shows you the wrong kind of dress that you don't like.

Big deal.

Right?

Yeah.

You might find something that you do like.

You might.

Yeah.

And I want it to give me what I like.

I want it to learn my taste.

What if we're talking about, I think the other example you had mentioned earlier is something like period tracking apps.

Yeah.

Because, yeah, it's on all those kind of things.

Yeah.

Is that helpful?

Is that controversial?

Is it?

Yeah, I don't want to share that with it with anyone.

That's my own personal data.

It goes into how I function.

And, yeah, that's one where I'm not a particular fan of that application because.

I think dating apps use a lot of AI as well, don't they?

But again, it's that if it's just showing you that if you're making the wrong decision as a data, then you're going to keep getting the wrong people.

Yeah, that one is fascinating.

I actually had some ideas for how dating, how this could be flipped on its head.

Just randomly share this with you right now.

If you think about the dating apps, all of them are predicated on the way that people look.

But what if we just took away all the pictures?

Oh, my God, I've just started watching love is blind.

I don't know if you've seen that, but I'm not.

No, but I think there might be something to that focus on.

Not the external kind of indicators of success and wealth, but focus on things that actually make you a better person, where a potential dating partner could help you or be a good match.

And I think there might be something there.

I think so, yeah, definitely.

Because you want to, you have the sort of discussions, well, you know, what do you think of this topic?

Or what would you do if this happened?

Or what are your thoughts on that?

Or kind of, what do you think a relationship should look like and this kind of thing, or how would you support each other?

And I think, yeah, if you can kind of go through all those things and match all those bits, then you're kind of halfway there to liking the person already.

Yeah.

And I think that the trouble with, you know, dating apps is sometimes people don't even know what they want.

And I mean, this is true of building any kind of software or Aihdenk, you ask people what they want, they can't, they don't necess, they can't necessarily articulate it.

And it's, you know, sometimes, you know, maybe we'll pick exactly the wrong person.

So, so there's, there's, there should be a component of randomness and serendipity there.

I think that actually might be helpful.

Yeah, no, I think so.

And, um.

Because.

Yeah, because otherwise it's just, yeah, it's, yeah, it's not giving you the choice and that's what it's giving you too much choice.

Yeah.

I mean, where do you think, do you think that's different in b two.

B as it than it is to consumer facing.

Yeah, it is the story of b two B and B two B.

AI is.

It's an entirely different ballgame.

And I mean, again, I think this one, this is one where we can do a whole other podcast, but essentially consumer Internet companies.

So the metas, the Googles, Microsoft is now considered an enterprise company.

But the product I started working on was a consumer like Web search is a consumer application.

And so when you think about consumer applications, they are really completely, I touched on this earlier, they're data driven.

And in order to build these, you need an experimental culture.

So you need the data, you need to have a culture of we're going to experiment into the best product over time.

You need to have the infrastructure where you can collect data, track it and continuously look at the metrics.

You can't know if your product is improving unless you're measuring it.

And it turns out that this setup is actually ideal for AI.

Right.

The B two B world works in a very different way.

It's very customer driven.

So you talk to your customers and you close deals.

And because of this focus really on getting paying customers, it means that historically the companies didn't have the ability to experiment.

They can't experiment and ship product quickly because, you know, they need to have, they have paying customers, they need to trust them.

So these customers are inherently more sensitive to changes, which gives you a less leeway to experiment.

So as a result, B two B companies don't have, many times the enterprise companies specifically, like the big ones, just don't have that experimental culture.

And they need to make an organizational shift to where they figure out how to experiment into AI products to get better over time.

Two, a lot of times they don't have infrastructure that supports collecting data and having a feedback loop where the data that you collect makes the product better over time.

They don't think about product metrics in terms of, you know, like usage and engagement.

Again, the focus is on paying customers and reducing the churn.

So because of all of this, B two B has to make a cultural shift and then it needs to fulfill a lot of the infrastructure requirements.

But again, this is why I'm investing in B two B.

I just feel like that is an enormous opportunity.

It's lagging now, but it will catch up.

And I think there are so many things that are going to get so much better in B two B over time.

And so again, problems are actually opportunities.

Going back to your earlier question, what kind of advice I would give?

I guess the other one was look at problems and challenges as opportunities.

Like, yeah, being a woman, you know, sometimes looks like a problem, but it's also an advantage.

You might be the only woman in the room, but you're also the most visible person.

So, you know, really step up and make sure you show your, you put your best foot forward.

And so every problem, while it might seem daunting, is also a hidden opportunity.

Yeah, 100%.

I agree with that completely.

There's so many opportunities and AI is really transforming the whole landscape, isn't it?

I mean, how do you see AI transforming everyday life in the next ten years?

Well, it could be the next two years.

Ten years seems quite long.

Big, big.

I mean, again, I think it's just a natural shift.

First it was the industrial revolution, then the technology and software, and now we're moving into the AI age.

And so it's going to move into more and more aspects of our lives and it's going to be, and it already is.

It's getting into all of our devices, it's getting embedded everywhere.

I.

And there's a tremendous amount of hype around AI, but I also think that there's real value there and there is real promise.

And so I'll give a couple examples.

So drug discovery, actually, we're now able to find new drugs and personalize them much easier with AI.

And there's been a lot of examples of this recently.

There was something called de novo antibody design, regenerative AI.

And so this really is something that traditionally drug discovery takes a lot of time and resources, and usually people have very little control over the outputs.

So in this case, the antibodies that need to be generated are really, always suboptimal.

Now, regenerative AI models, you can create antibodies with one round of model generation and reduce the time you need fewer optimizations.

And so that one, I think, is really promising and really interesting.

There's, you know, chip design is another one.

I think Nvidia has had a ball with this particular one.

Overall, the trend is that AI is making us a lot more productive and I think can be used to leave to give us higher quality work.

So just in, just last year, there were several studies that assessed AI's impact on labor.

And really what was found is AI enables people to complete tasks more quickly and improve the quality of the output.

And so there's really potential to bridge the skills gap between low and high skilled workers.

It decreases costs and increases revenues.

So from a business perspective, like, incredible things are happening here, so the potential is there, but, you know, it's like anything.

Double edged sword.

There's also some danger if it's not done properly.

And so I think the future is bright.

It depends.

It's up to us to make the best of it.

And, you know, I'm very excited about what's going to happen next, although I'm at times a little bit scared, too.

Me too.

I could talk to you all day.

There are many different podcasts.

We need to have a spin off, so many topics, but where can our audience learn more about everything that you've been telling us?

I think the best way to learn is to build something.

And there's tons of easy to use AI applications that I suggest people should play with to get a sense of what are the things that you can do with the technology.

So if you're at the very kind of beginning, let's say you're in the spectrum or you're an AI beginner, you can play around with the tools easily and figure it out.

It just requires a time investment and a will to learn.

And the best thing to do is find something you care about and then use AI to support it.

So let's say you like to travel.

One of the things I love to do when I travel is I love to take videos.

There's tons of AI tools that can make your videos better.

So focus on what you want to do, what are the kind of outcomes, what are the things you're passionate and interested in.

And then think about how AI can help you if you want to get more in depth with the AI.

There are so many resources out there that you can learn things easily.

I mean, MIT offers free online courses.

This is incredible.

You know, we talked about how technology can be used for good and how it can democrat.

Literally anyone in the world can go online.

All they need is an Internet connection, and they can get MIT courses and learn.

So I really feel like technology has commoditized education.

It's made things easier.

My favorite resources.

I really like Coursera.

Andrew Ng, who is an AI pioneer, has his own set of courses called, I think he actually founded Coursera, but he has his own set of courses called deeplearning AI.

You could take those courses and you could go in depth as much as you want to, to learn how to build, to learn how the guts of AI work.

But again, I'd say don't learn for the sake of learning.

Figure out how you want to use the AI, what are the problems you want to solve, and then go back and figure out what are the tools and what are the courses you need to take to get there.

That's great advice.

Brilliant.

So finally, where can people find you?

Is LinkedIn the best place?

Yeah, I'm pretty active on LinkedIn.

I'm on Twitter.

I have my podcast.

I'll be restarting that.

I took a little bit of a break, and it's also on substack and YouTube.

And I love hearing from people, so please feel free to reach out.

Yeah, Britt.

We'll put all the links in the show notes so people can find you.

Natalia Barina, thank you so much for coming on.

Women with AI.

Thank you so much, Jo.

It's been a pleasure.