DSS Miami 2023: Generative AI's Transformative Journey

By Data Science Salon

From data visualization to LLMs in cybersecurity, and everything else in between - Data Science Salon Miami 2023 was a great opportunity to meet like-minded professionals and exchange experiences about the ever-changing realm of AI and Machine Learning.

Last year marked a transformative moment in generative AI. On July 12, Midjourney made its debut as a notable text-to-image service. A couple of months later in September, OpenAI launched Dall-e, another influential text-to-image tool. The online space was soon flowded with captivating images generated through these platforms, including memorable creations like Pope Francis wearing a puffer jacket.

Fast forward to November 2022, ChatGPT made its entry, reshaping perceptions about the capabilities of gen AI. This seismic shift hasn't gone unnoticed; businesses globally are recognizing its potential. Early adopters, armed with the insights and strategies to harness this technology, stand at the forefront of tapping into this vast economic potential. Insights from these pioneers were recently spotlighted at Data Science Salon in Miami.

Digital transformation of the cruise industry with AI

“We don’t only want to know what their experiences were, but also to change that,” said Matthew Denesuk, SVP, Data Analytics & AI at Royal Caribbean Group. During his speech, he shared remarks and thoughts on the transformational role of AI in the cruise industry. Yet the challenge is the fact that cruising is not traditionally a highly digitized industry.

Matthew-Denesuk-DSS-MIA-2023Matthew Denesuk at DSS MIA 2023

“First of all, we need to talk about digital transformation in business. We have seen it in telecoms, media, and banking,” says the expert. “Industries that have become all about data and analytics have seen disruption and enormous gains in productivity and performance,” he adds. The same can be applied in less “datafied” industries.

While the concept may sound straightforward, putting it into practice has its complexities. Matthew Denesuk talked about the nuances and best practices adopted at Royal Caribbean. He also  shared insights into challenges faced, from catering to a diverse customer base to navigating intricate business processes in an increasingly complex and dynamic environment.

“And in the end, you have all that using bad data, that was gathered for other reasons,” he adds.

Data literacy is essential 

Laura Gabrysiak, Senior Manager - Data Products & Solutions at Visa, shared her experience in enhancing data experiences through generative AI.

"Better customer experience, better customer experience, we have these topics, there is usability," enumerates Laura. "In 2005, we saw data, and we started to talk about the benefits data brings for business," she adds. Later during her presentation, she shared a more modern approach to the matter, showcasing the types of endpoints data-related solutions usually have, such as a dashboard.

Laura-Gabrysiak-DSS-MIA-2023Laura Gabrysiak at DSS MIA 2023

"Building a good dashboard, a good data visualization, requires a lot of work," says Laura. Additionally, she adds that the overall data experience encompasses many fields and disciplines, including user experience, design thinking, and data science. "And basically, it is the framework you need to use when developing your data-based solution," she adds.

Data visualization in the age of AI

Natalie Porter, Data & Solutions Strategist from Southern Glazer's Wine and Spirits,covered the topic of data visualization. She began by talking about Florence Nightingale's use of data visualization to convince decision-makers to implement hygienic measures that reduced unnecessary death rates to 4%. Then, Porter showed modern data visualization tools that are more sophisticated than those used by Nightingale.

Porter also discussed several use cases for generative AI in data visualization. One such use case is data cleaning, which can be time-consuming and tedious. Generative AI tools like ChatGPT can help by automatically cleaning and formatting data.

Natalie-Porter-DSS-MIA-2023Natalie Porter at DSS MIA 2023

Another use case for generative AI in data visualization is interpreting data series. Porter shared an example of asking GPT-4 to interpret a data series and suggest ways to tune the outcome. This can be helpful for data scientists who want to better understand their data and identify trends and patterns. 
Overall, Porter's talk highlighted the potential of generative AI to revolutionize the field of data visualization. By automating tasks and providing new insights, generative AI can help data scientists create more informative and effective visualizations.

Deliver value to your customers

Sergey Ferguson, Vice President and Head of Data Science at TelevisaUnivision, one of the largest Spanish-language broadcasting companies in the world, shared his insights on using data to transform the media business.

Ferguson explained that TelevisaUnivision combines first-party and third-party data to create a household graph. This graph allows the company to interact with its consumers and other Spanish-speakers in the digital ecosystem.

Sergey-Ferguson-DSS-MIA-2023Sergey Ferguson at DSS MIA 2023

Ferguson also showed how TelevisaUnivision uses this data to improve its targeting and advertising effectiveness. He noted that without access to this data, up to 70% of impressions are wasted when targeting Spanish-speakers.

Overall, Ferguson's talk highlighted the importance of data in the media business. By using data to better understand its audience and target its advertising, TelevisaUnivision is able to achieve better results for both its consumers and its advertisers.

Demystifying AI: Deployment, Challenges, and Opportunities

During the panel discussion, industry leaders dove into the intricacies of AI deployment, addressing challenges, opportunities, and legal dimensions across different sectors. Drawing from their firsthand experiences, they offered unique perspectives tailored to their respective domains.

"In our defense projects, AI is integral. We're aspiring to harness generative AI seamlessly throughout our systems. Yet, a challenge we face is the compartmentalization of our data, limiting its reuse across different initiatives," commented Alexandra Levinson of GMV.

Jennetta George, SVP of AI at AlixPartners, shed light on the criticality of AI explainability: "Meeting compliance mandates is paramount. When rolling out generative AI for a client, we confronted numerous legal hurdles. The foundational step is robust data management."


Panel Discussion at DSS MIA 2023

The discussion squashed some common myths about AI, notably the perception of Gen AI being a costly endeavor. Antonio highlighted the affordability and accessibility of AI models via cloud platforms. Panelists also emphasized the need for upfront cost considerations, the essence of transparency, and the notion that AI is meant to enhance rather than supplant human functions.

Dimitri Berdnikov of LinkedIn emphasized, "It's not about crafting novel, out-of-the-box experiences. The journey begins by comprehending the full spectrum of tasks at hand and pinpointing where generative AI can deliver results more efficiently and economically."

Beyond the hype

"Initiating cost conversations at the outset is paramount, not relegating them to the latter stages," emphasized Dimitri Berdnikov while discussing maximizing returns with generative AI. "Determining the right business model and monetization strategy is crucial," he further pointed out.

Dimitri also talked about how businesses can identify potential use cases for generative AI. "Begin by analyzing the tasks your customers perform. Contemplate how generative AI can augment these activities. For instance, tasks with high structure and volume can achieve up to a tenfold speed boost with AI," he explained.

His insights underscored the value of pinpointing routine, high-frequency tasks, such as those in customer support. Additionally, he accentuated the significance of meticulous methodologies and a profound understanding of prompt engineering.

Why most AI projects fail

"Despite the hype around AI, only 15% of AI projects are completed, and even fewer actually bring value," said Shakeel Hye, Director of Data Engineering at TracFone Wireless.

Hye identified several key non-AI reasons why AI projects fail, including:

  • Wrong identification of the business problem or opportunity to leverage AI
  • Lack of alignment between business and data strategy
  • Improper solution designs
  • Lack of the right data sources

Hye also went into more detail about these reasons:

"One of the key reasons is that organizations jump onto the most difficult use cases upfront. Don't do that! Choose the easiest use case, with the lowest costs and the lowest investment. There is no free computing! We always have limited resources, even though now we have scalable, cloud systems," he said. "Sometimes companies run out of fuel halfway through the project!"


Shakeel Hye at DSS MIA 2023

In other words, AI projects are more likely to succeed if organizations start with small, achievable projects that are aligned with their business goals and have the necessary resources to support them.

LLMs: Boosting efficiency and cutting costs in cybersecurity

Jeff Schwartzentruber, Senior Machine Learning Scientist at Esentire, talked about the role of LLMs in cybersecurity, presenting both specific applications and potential future uses.

"One intriguing application is using LLMs for penetration testing, a critical component of cybersecurity," he highlighted.
Jeff also touched on the challenges of tailoring LLMs for cybersecurity tasks. "A primary obstacle is that the datasets typically employed for training or pre-training LLMs often lack substantial cybersecurity-specific data or conventions," he noted. "Elements like extensive log files, scripts, and paths can be intricate, and the way these models tokenize data might hinder their comprehension of such content."


Jeff Schwartzentruber at DSS MIA 2023


At Data Science Salon Miami 2023, hundreds convened to delve into the cutting-edge of machine learning and generative AI. Serving as a pivotal guide to AI's rapidly changing landscape, this event was a standout in a week-long series. It's one of many Data Science Salons held across the US. Up next: San Francisco, New York, with Seattle and Austin slated for Q1 2024.

Get the latest data science news and resources every Friday right to your inbox!