Unlocking the true power of AI hinges on the ability to connect and structure data effectively. This podcast dives deep into how organizations can harness AI by transforming their data into an interconnected web of information, akin to a well-organized kitchen with all the ingredients at hand. The discussion emphasizes the importance of knowledge graphs and ontologies, which serve as essential frameworks for understanding the relationships within data, thereby enabling AI to make more informed and accurate decisions. By exploring innovative concepts like working memory graphs and the role of semantic metadata, the hosts illustrate how these tools can enhance AI's ability to reason and derive insights. Ultimately, the conversation highlights the necessity for a collective effort in creating a data-driven culture that empowers both humans and machines to work together effectively.

Links referenced in this episode:

• schema.org

Companies mentioned in this episode:

• UBS

Takeaways:

• The true potential of AI lies in how well we connect and structure our data.
• Organizations must focus on making their data AI-ready to harness its full capabilities.
• Ontologies serve as crucial frameworks that provide context for AI understanding and reasoning.
• Building a network of interconnected knowledge can enhance resilience and adaptability in AI systems.
• Investing in a strong data foundation is essential for creating intelligent organizations.
• The working memory graph combines LLM capabilities with knowledge graphs for new insights.

The episode articulates a bold vision for the future of AI through the lens of connected data. It unpacks the critical role that data organization plays in the effectiveness of AI systems, drawing parallels between a cluttered kitchen and fragmented data silos within organizations. The hosts emphasize that, much like a chef requires well-organized ingredients to create culinary masterpieces, AI requires structured and interconnected data to unleash its full potential. They discuss the importance of knowledge graphs and ontologies as frameworks that define relationships and concepts within data, which provide the necessary context for AI to function accurately and intelligently.

As the discussion flows, the hosts introduce the concept of the working memory graph, a hybrid model that combines the intuitive capabilities of LLMs with the structured knowledge from graphs. This innovative approach allows AI to recognize patterns and make connections in a way that mimics human cognition, enhancing its ability to reason and understand complex information. The conversation delves into practical strategies organizations can adopt to begin this transition, such as linking existing datasets to public ontologies and adopting a mindset of continual improvement in data management practices.

The episode culminates in a profound reflection on the ethical dimensions of AI deployment. It stresses the necessity of keeping humans in the loop to ensure that AI systems reflect our values and ethical considerations. The hosts advocate for a collaborative approach to AI, where human intuition and expertise guide the development and implementation of these technologies. This episode serves as a comprehensive guide for organizations looking to navigate the complexities of AI, emphasizing that the journey towards a data-driven future requires both strategic planning and a commitment to ethical practices.

Hey, there ya.

Today we're diving into AI, you know, and how to unlock its true power through connected data.

Right.

We've gotten some really intriguing articles and research and even a few philosophical quotes about this whole world.

It's interesting stuff.

So let's unpack all this and figure out how organizations can truly harness the power of AI, because it's looking like the key to just might be how we connect and structure our data.

Yeah.

What's fascinating is this isn't just about the latest AI models, like GPT four, which can write letters, analyze data, even generate code.

Oh, wow.

But it's about understanding that the true power of AI lies in making your data AI ready.

Okay, think of it this way.

You can have the fanciest oven in the world, but if your ingredients are scattered and disorganized, you won't be baking any masterpieces.

Right, exactly.

Yeah.

And these sources really dive deep into that idea.

They do.

They highlight how most organizations organizational data is currently locked away in these separate databases.

It's like having ingredients in different corners of the kitchen.

Yeah, yeah.

Not exactly useful when you're trying to feed a powerful AI system that thrives on connections and relationships.

Exactly.

And it raises this important question, what's the best way to organize data for AI?

And these sources point to knowledge.

Graphs, they're like a recipe book for AI, showing how all the different data points relate to each other, creating this structured, interconnected web of information.

I see.

So it's all about making those connections, like drawing lines between different pieces of information.

Exactly.

It's about providing that context that AI needs to make sense of it all.

And that's where ontologies come in, right?

Yes, exactly.

Ontologies.

Think of them as the language of knowledge.

Graphs.

They define the concepts and relationships within your data.

Okay.

They're like the glossary in our recipe book.

Right.

Making sure everyone understands the ingredients and how they combine.

And how they combine.

Yeah.

So it's about having that shared vocabulary, that common understanding.

Exactly.

Ontologies are vital because they provide context for AI.

They're like guardrails.

I like that.

Preventing those hallucinations, you know, where AI models generate inaccurate or misleading information.

Right, right.

You've probably seen this with LLMs like chat, GPT.

Oh, yeah, for sure.

They can be incredibly creative.

Yeah, but sometimes go off on tangents that aren't grounded in reality.

Right, right.

Get a little carried away.

Yeah.

Ontologies help to keep them on track.

Interesting.

So it's like they provide the boundaries.

Yes, exactly.

The structure framework so that AI can be creative, but within a controlled environment.

Precisely.

Okay, so we've got our ingredients, we've got our recipe book, and we've got our glossary.

Right.

Now, where does the working memory graph fit into all of this?

Because the sources bring that up as well.

Yeah.

The working memory graph is a brilliant design pattern that combines LLMSD ontologies and knowledge graphs.

Okay, so it's like the whole package.

It is.

It's like having an AI chef with a photographic memory and a deep understanding of culinary principles.

Okay.

Working seamlessly with a well organized pantry and recipe book.

I see.

So it's like bringing together the best of both worlds.

Exactly.

The working memory graph allows AI to enrich data, make complex connections, and ultimately unlock new levels of capability within organizations.

Okay.

It's about moving beyond simply processing data to actually understanding and reasoning about it.

And one source uses the example of the butcher on the bus to explain this.

Oh, that's a good one.

Have you ever had that experience where you recognize someone but you can't quite place them?

Oh, yeah, all the time.

Like, did I go to school with them or are they the barista from my favorite coffee shop?

Right, right.

And then it hits you.

They're the butcher I see at the market every week.

Yeah, that aha moment.

Yes.

That's the kind of cognitive leap that a working memory graph can help AI achieve.

Exactly.

The working memory graph allows AI to combine that fuzzy intuitive recognition which LLMs are great at, with the specific precise knowledge stored in knowledge graphs.

So it's like having the best of both worlds.

Yes.

Creating a powerful and adaptable system.

So what does this all mean for organizations listening to this deep dive?

Like, what's the takeaway here?

The takeaway is this.

While purchasing external AI models is important, the real game changer is making your own data AI ready.

And that means connecting and structuring your data in a way that AI can truly understand and leverage.

So it's about investing in a strong foundation.

Yes.

That will allow you to build a truly intelligent organization.

Exactly.

And the sources give some practical advice on how to get started.

Oh, good.

They suggest starting small, maybe by linking a few existing datasets to widely used industry standards or public ontologies like schema.org.

That'S a great place to start.

It's like dipping your toes into the water before diving head first.

Exactly.

Get a feel for it.

Yeah.

And they also encourage us to think big.

They suggest using URL's to identify data points, making them globally unique and linkable in distributed graphs.

That's thinking long term.

Yeah.

This is about laying the groundwork for a future where data can flow seamlessly between organizations.

It is.

It's about creating a network of interconnected knowledge.

And one source introduces the DProd specification as a practical tool to get started with connecting data semantically.

Oh, yeah, DPRD is great.

It's a simple yet powerful way to define data products using plain JSON.

Link them to shared schemas.

Right.

Connect them in a distributed graph.

It's like a starter kit for building a connected data ecosystem.

A starter kit, I love that.

Yeah, it's accessible.

So it's about lowering the barrier to entry, making it easier for organizations to take those first steps.

Exactly.

And then from there they can start to build out those connections, create those knowledge graphs, and really start to unlock the power of AI.

Precisely.

It's an exciting time to be working with data.

It really is, yeah.

There's so much potential and so much to learn.

Absolutely.

But I think the key takeaway here is that connected data is the key.

It is.

It's the foundation for everything.

It's a foundation for the future of AI.

Absolutely.

It's not just a technical task, you know, connecting data, it's a deeply human one, I think.

Oh, interesting.

How so?

Well, it requires us to think about the relationships between things, to see the bigger picture, you know, and to make choices about how we want to structure our world.

I see.

So it's about imposing order on chaos.

In a way.

It is.

And one of the sources used this phrase, network of networks.

Network of networks, to describe this vision.

It's like each organization creating its own miniature web of connected data.

Okay.

And then those webs linking together to form a larger, more complex ecosystem.

So it's like the Internet, but for data, in a way.

Exactly.

And that network of networks is the foundation for a future where AI can be truly decentralized and distributed, where intelligence emerges not from a single monolithic system, but from the interactions of many different agents.

It sounds almost like a biological organism.

Yes.

With different cells and organs working together to create a whole that's greater than the sum of its parts.

That's a great way to think about it.

And just like in a biological organism, diversity and interconnectivity are essential for resilience and adaptability.

So a network of networks can be more robust.

Exactly.

More responsive to change than a single centralized system.

Wow.

It's amazing to think about the potential of this.

It is.

Imagine a world where AI can help us to solve complex global challenges like climate change or poverty by connecting these different sources of data and insights from around the world.

Exactly.

That's the kind of future we can create if we start thinking about data differently.

Okay, so it's not just about collecting the data.

It's not just about collecting it.

No, it.

It's about connecting it, structuring it, and making it meaningful.

Creating those relationships.

Exactly.

It's about creating a shared understanding that can empower us to make better decisions, you know?

Right.

To actually use that data to make.

A difference and to build a better world.

So what can organizations do to kind of move towards this vision?

Like, what would be a good first step?

Well, one of the sources suggested starting by just looking at the data we already have.

They encourage us to think about how we can give each data point a unique identifier, like a URL, so that it can be linked to other data points in a distributed graph.

So it's like giving each piece of data its own address.

Exactly.

Its own address in this global network.

Okay, I like that.

And that's where things like schema.org come in, right?

Yes, exactly.

Those shared ontologies that we talked about earlier, they provide a common language for describing data.

They do.

They're like a universal dictionary for data, helping us to establish a shared understanding, not just within our organization, but across, you, different organizations and domains.

So it's about finding that common ground, that shared language that everyone can agree on.

Exactly.

And the beauty of schema.org is it's open and accessible.

Right.

Anyone can use it.

Anyone can contribute to its development.

And that's important.

It is.

It's about collaboration.

Yeah.

Because this isn't something that one organization can do alone.

No.

It's a collective effort.

It's about everyone working together to create this connected future.

One of the sources makes this interesting point about the transformer architecture used in LLMs, and they suggest that we could actually think of transformers themselves as a type of graph neural network.

Oh, that's a pretty mind blowing idea.

It is, it is.

And it highlights this deep connection between language and graphs.

Okay.

Transformers excel at understanding the relationships between words and a sentence.

Right.

That's what they're known for.

Exactly.

And those relationships can be represented as edges in a graph.

Yeah.

So, in a sense, transformers are already working with graph like structures.

They are.

Even if we don't always think of them that way.

And if we kind of embrace this idea.

Yeah.

It opens up some really exciting possibilities.

It does, because it means that we can start to bridge the gap between structured data, like what we have in our databases.

And unstructured data, like text and images.

Exactly.

And that's the key to unlocking the full potential of AI, I think.

Yeah.

It's about bringing together all of those different forms of information.

It is, yeah.

We need to move beyond thinking of data as either structured or unstructured and start to see these underlying connections and relationships that exist across all forms of information.

It's about seeing the connections, not the silos.

Exactly.

And that reminds me of one of the sources that talks about the importance of semantic metadata.

Oh, yeah.

Adding meaning to data.

Right.

Making it more understandable for both humans and machines.

Semantic metadata is crucial because it allows us to move beyond simply storing data.

Right.

To actually understanding what it means.

It's like adding labels to the ingredients in our pantry.

Exactly.

So we know what they are and.

How to use them.

And how to use them.

Yeah.

And that's where ontologies come in again.

Right.

Yes.

They provide the vocabulary for describing those semantic relationships.

Exactly.

Ontologies are like the language of meaning.

Right.

Helping us to connect data points in a way that makes sense and that.

Allows us to extract insights that would otherwise be hidden.

Precisely.

One of the sources talks about this idea of compressibility in data.

Oh, interesting.

And they argue that the better we connect and structure our data.

Okay.

The more compressible it becomes.

Interesting.

And the more value it holds for AI.

That's a really interesting point.

Think about it like packing a suitcase.

Okay.

If you just throw everything in randomly.

Right.

It's gonna be a mess.

Yeah.

And you won't be able to fit very much.

You're wasting space.

But if you carefully fold and organize your clothes, you can pack much more efficiently.

Right.

You're using the space effectively, and it's.

The same with data.

It is.

The better we organize it, the more.

Efficiently we can store it, the more.

We can fit into a given space.

Right.

And the more useful it becomes.

Exactly.

Yeah.

So it's about efficiency, but it's also about making that data more meaningful.

It is.

It's about packing the most meaning into the smallest possible space.

And it sounds like knowledge graphs are kind of the key to achieving that efficient packing.

They are.

They help us to connect and structure data in a way that makes it more compact and meaningful.

Precisely.

Knowledge graphs are like the ultimate packing organizers.

Okay.

I like that for data.

Allowing us to fit more information into a smaller space and make it easier.

To access and understand.

Exactly.

One source talks about the semantic web and how it can be used to create a shared semantic layer within an organization.

Yeah.

It's like a common understanding that can.

Be accessed by different departments and applications.

Exactly.

The semantic web is a powerful vision.

It is.

And it's starting to become a reality.

It is?

Yeah.

Thanks to technologies like linked data and schema.org dot.

These open standards allow us to connect data from different sources and describe it.

In a common language, making it easier to share and reuse information across different systems.

Exactly.

One source uses this really cool analogy.

They talk about building an internal version of schema.org dot.

Oh, interesting.

Called like your dot schema.org dot.

Your dot schema.org dot.

Okay.

That reflects the specific concepts and relationships that are important to your business.

That's a great way to think about it.

So it's like creating your own custom dictionary for data tailored to your specific needs.

That's a brilliant way to think about it.

It's about taking ownership of your data and creating a shared understanding that reflects your organization's unique perspective.

And then you can use that shared understanding to build these powerful AI applications that are truly tailored to your business.

Exactly.

It's like having a team of AI experts who speak your language and understand your goals.

And it's not just about building custom applications.

It's about creating a data driven culture, okay.

Where everyone in the organization can access and understand data in a way that is meaningful to them.

It's about democratizing data in a way.

It is, yeah.

Making it accessible to everyone.

One of the sources talks about the concept of data agents.

Data agents, okay.

And they envision this future where each data product in an organization is essentially an intelligent agent.

Interesting.

With its own embedded AI model.

Okay.

These agents can communicate with each other, sharing information and collaborating to solve complex problems.

That's a really fascinating idea.

It's like a network of intelligent beings, each with its own specialized knowledge and skills.

Right.

Working together.

Working together to achieve this common goal.

And that network can be incredibly powerful because it can draw on a wide range of perspectives and insights.

Right.

And it can solve problems that would be impossible for a single agent to tackle alone.

Exactly.

It's about the power of collaboration.

It is, yeah.

And that's something that AI can help us to achieve.

It can, yeah.

It can connect us in ways we never thought possible.

One source talks about how UBS, the swiss bank, is already using a similar approach with their semantic router.

Oh, interesting.

Semantic router.

It's a system that allows them to analyze news articles and understand how they might impact their trades.

So they're connecting news data with financial data.

Exactly.

To gain insights into the market.

It's like having a team of AI analysts who are constantly monitoring the news and providing insights that can help them to make better decisions.

That's a great example of how this kind of technology can be used in the real world.

It is.

And it highlights the importance of connecting different sources of information.

Right.

Because it's often the connections between data points that reveal the most valuable insights.

It is, yeah.

It's about seeing the bigger picture.

It's fascinating to see how all these ideas are starting to come together.

It is.

You know, we're moving towards this future where data is not just something we collect and store.

Right.

But something we actively connect and structure to create meaning.

It's about making data meaningful.

Yeah.

And unlock new levels of intelligence.

And that future is not just about machines.

It's not.

It's about humans and machines working together.

Working together in a way that is.

Both symbiotic and empowering.

And one of the sources really emphasizes that, you know, the importance of keeping humans in the loop.

Oh, absolutely, yeah.

They argue that we need to be careful not to let AI become too autonomous.

We need to ensure that human values and ethics are embedded in every stage of its development and deployment.

That's a crucial point.

Yeah.

AI is a powerful tool.

It is, but it's only as good as the intentions of the people who create and use it.

We need to be mindful of the potential risks as well as the benefits.

Absolutely.

We need to make sure that AI is used in a way that is beneficial to humanity as a whole.

It really is.

And this is a great point to emphasize that it's not about taking humans out of the equation entirely, right?

Absolutely not.

No.

It's about keeping humans in the loop, making sure that AI reflects our values, not replacing them.

Right.

So how do we do that?

Well, one source suggested incorporating human expertise into those knowledge graphs and ontologies we talked about.

Humans can curate data, define those relationships, relationships, and ensure that AI aligns with our goals.

It's like training an apprentice.

Oh, I like that.

Passing on the wisdom of generations, but in this case, to a digital student.

And this really stood out to me.

You know how sometimes you just recognize someone, but you can't quite place them?

Oh, yeah.

Like, did I go to school with them, or are they the barista from my favorite coffee shop?

Right.

You see them all the time and.

Then it hits you.

They're the butcher I see at the market every week.

They butcher on the bus.

Yeah, exactly.

One source used that butcher on the bus analogy to talk about how our brains make these leaps.

Yeah, it's a great analogy.

And it made me think about how we can design AI systems that can make those same kinds of connections.

It is.

It gets to the heart of what we can learn from human cognition as we design these AI systems.

You know, we have this incredible capacity for both fuzzy associative thinking and this precise fact based recall.

Right.

We can think both intuitively and logically.

Exactly.

And the butcher on the bus moment is the perfect example.

It's where those two modes of thinking intersect.

Right.

Where the fuzzy becomes clear.

Exactly.

And with AI, we can strive for that same duality by combining the strengths of LLMs and knowledge graphs.

So it's not about picking one over the other, but finding ways to combine them.

Right.

To create something that's a more than the sum of its parts.

Exactly.

And that's where the real magic happens, I think, when we start to bridge those different approaches.

So, we've covered a lot of ground here today, from the raw power of LLMs to the elegant structure of knowledge graphs we have.

It's been a fascinating conversation.

It really has.

But for someone listening who's thinking, okay, this all sounds amazing, but where do I even begin?

What's a good first step?

Start small, you know, don't try to boil the ocean all at once.

Right.

Take it one step at a time.

Exactly.

One of the sources had a fantastic suggestion.

They said to start thinking about your data as if each piece of it needs a unique address, like a URL.

Okay.

Like giving each data point its own digital identity.

Precisely.

Imagine taking your customer database, your sales records, your marketing materials, right.

All those different data silos, and giving.

Each data point its own unique identifier.

This is how you begin building those connections, creating a web of information that AI can actually navigate.

And this connects back to schema.org dot, right, using those shared ontologies.

Exactly.

Schema.org is like providing a common dictionary for your data.

It helps establish that shared understanding, not just within your organization, but potentially across entire industries.

So if I'm hearing you correctly, the real magic happens when we move beyond just having data to actively connecting it in a way that's meaningful.

Right.

You got it.

Think about it like this.

What's more valuable?

A library full of books randomly scattered on the floor, or a library where those books are meticulously organized, categorized, and easily searchable?

I definitely choose the organized library.

Of course, knowledge graphs are like building those organizational structures for your data.

They make it discoverable, they make it usable.

And just like a well organized library, unlocks the potential of those books.

A well structured data ecosystem unlocks the potential of AI.

Precisely.

It allows AI to become more than just a tool for automation.

It becomes a partner in discovery, helping us uncover hidden patterns, make new connections and gain insights that would have been impossible to see before.

This has been an incredible deep dive and I have to say I'm walking away feeling incredibly optimistic about the future of AIh.

Me too.

There's a real sense of opportunity here.

We're on the verge of a new era of data driven innovation and it's.

Up to all of us to shape it responsibly.

Absolutely.

It's a collective effort and we all have a role to play.

To everyone listening, thank you for joining us on this journey into the world of connected data and AI.

We hope you found it as thought provoking as we have.

Yes, thank you for listening.

Keep those questions coming, keep exploring, and most importantly, keep connecting.