Joe Lai, chief technology officer of Creativ Company, initially joined as a co-op student nearly a decade ago. The ease with which he picked up new technology stacks was instrumental in getting the company’s first generation of models off the ground in 2016.

Since then, Joe has been working to build models in Spark, building data pipelines using Kafka, enabling API layers using next-generation openWhisk framework, and many more exciting technologies.

Recently, Joe has been building Creativ Insights’ HTML5/Angular responsive interface. This interface sits on top of openWhisk based API layer, which can dynamically scale to handle traffic of over 100 TPS. Joe is fluent in Scala, Java, Python, HTML5/Angular/JS, and building learning modules running on TensorFlow and Spark frameworks using hybrid backends like graph, document, key/value or relational.

A graduate of the University of Waterloo with a degree in computer science and a minor in statistics, Joe lives in Toronto, Canada.

Joe sits down with LBB to discuss integrating AI into the Creativ’s marketing-technology stack, using large language models (LLMs) to cluster data, and the challenges of collaborating with AI…

LBB> ​​What’s the most impactful way that AI is helping you in your current role?

Joe> As the CTO, my entire job revolves around integrating AI into our marketing-technology stack. We have built large language models (LLMs) trained on various datasets, ranging from social media comments to internal customer data, that allows us to identify and provide custom actionable insights for our clients.

These LLMs can identify changes in consumer sentiment and brand perception, generate campaign reports and competitor analysis, and pinpoint topics of interest at a speed and scale that was previously impossible without AI.

This not only allows us to provide our clients with more informed decisions, but it also allows us to focus more on providing high-value strategies where our experience and creativity shines.

LBB> We hear a lot about AI driving efficiencies and saving time. But are there any ways that you see the technology making qualitative improvements to your work, too?

Joe> Absolutely. Our use of AI and LLMs raises our overall quality of insights by being able to uncover patterns that humans may overlook.

There’s a lot of data out there. Using LLMs to cluster data across multiple channels can help us identify emerging topics in conversations happening all over the world – ones that might not be obvious through traditional methods like keyword tracking. This richer level of understanding means that our reports and insights are not only faster, but also more precise, context-driven, and strategically valuable.

LBB> What are the biggest challenges in collaborating with AI as a creative professional, and how have you overcome them?

Joe> One of our biggest challenges is ensuring that our AI outputs are not only factually correct, but also contextually and strategically correct. This is where our years of experience in the tech and creative space kicks in.

We use this experience to curate our LLMs to fit the use case and context of the client, and we involve human validation to validate that our data and results are high quality. This process ensures that our AI models understand the specific nuances of the client, and that our resulting insights align with the client’s views and objectives.

LBB> How do you balance the use of AI with your own creative instincts and intuition?

Joe> I view AI as a powerful tool rather than a replacement for creative instincts. The AI models can surface patterns and ideas that I may not have known or considered, and I can then apply my own knowledge, creativity, and experience to either accept, refine, or reject these suggestions.

This allows me to utilise the powerful capabilities of AI without losing touch with my creative intuition.

LBB> And how do you ensure that the work produced with AI maintains a sense of authenticity or human touch?

Joe> The key is grounding AI in real-life experiences and perspectives. We tailor our LLMs to fit our clients’ individual needs, which means that our models understand the specific nuances behind the industry and company.

For example, gaming communities typically have their own slang and phrases that can be game-specific or industry used terms. We ensure that we capture these contextual differences in our models so that our output properly reflects the real world. We also make sure that a human reviews these outputs to validate that the results are contextually relevant and credible.

LBB> Do you think there are any misconceptions or misunderstandings in the way we currently talk about AI in the industry?

Joe> Yes. I think a common misconception is that AI is an autonomous decision maker. At its core, LLMs are language models. They are able to consume and process information like no human can, but its outputs are ultimately statistical and they do not have the same understanding that humans do.

In its current state, AI still requires high-quality inputs and human validation in order to produce the right results. This is why I treat AI as a very powerful tool in my arsenal that can influence my decision making, but does not make decisions for me.

LBB> What ethical considerations come to mind when using AI to generate or assist with creative content?

Joe> Ethical use of AI is core to the beliefs at Creativ. We are strict with data protection and data privacy, and we only use data that is either publicly available or is provided to us.

When it comes to training our models, we ensure that no sensitive data is exposed or misused. We are also transparent with how we use AI in our work, and we are cognisant of potential bias in our models.

LBB> Have you seen attitudes towards AI change in recent times? If so, how?

Joe> For sure. When ChatGPT first released and gained global popularity, it felt like people were either dismissive of its accuracy and sceptical of its usefulness, or they were fearing for their job safety, feeling that AI would overtake human thinking and creativity.

I think nowadays, as these LLM models have continued to mature, people have become more accepting towards AI. They generally view AI as a legitimate tool that can be very useful in certain situations, but they are still far away from being able to replace people’s thinking and creativity.

LBB> Broadly speaking, does the industry’s current conversation around AI leave you feeling generally positive, or generally concerned, about creativity’s future?

Joe> I feel generally optimistic; there is an enormous potential for AI to augment creativity across all industries. However, I think it has less to do with the tech itself, and more with how it is being used. If we treat AI as a shortcut to creativity, then we risk diluting real creative work with ‘AI slop’. But if we use AI to enhance human creativity, I believe we can unlock innovation that we never thought was possible.

LBB> Do you think AI has the potential to create entirely new forms of art or media that weren’t possible before? If so, how?

Joe> I believe so. AI has the ability to digest and consume information that we’ve never seen before. This includes all forms of media, whether it is text, audio, visual, you name it. I think it won’t be long before someone nudges AI in a direction where it can produce art and media in a format that we’re just beginning to imagine.

LBB> Thinking about your own role/discipline, what kind of impact do you think AI will have in the medium-term future? To what extent will it change the way people in your role work?

Joe> In the medium term, AI will shift my role to creating AI-driven pipelines that increase both the speed and efficiency of our systems. Instead of spending countless hours in cleaning and digesting millions of data points, I can spend time in generating and curating more insightful reports for my clients.

AI handles the heavy analytical lifting while I can focus more on client engagement, technical innovation, and integrating more complex and efficient pipelines for my company.