96
Score Big Insights on AI from Sports!
View this email in your browser
The weekly newsletter of the Business of Tech, giving you new insights into the world of IT service delivery.
Looking for stories from the podcast stories? Check out the pod itself on Apple Podcasts, Spotify, or daily in your inbox. Stories are available to everyone for five days,and Patreon supporters forever.
Was this forwarded to you? Join the list!
Not all AI models are the same
Have you been missing our AI deep dives? You’re in luck! We’re pivoting back to tech media’s favorite topic with another look at one of my ongoing predictions about the future role of AI: for IT services providers, understanding different kinds of AI models is going to be key for remaining relevant to our customers.
You’ve heard plenty about language learning models with the buzz around ChatGPT, but have you heard of data-driven generative AI? I recently welcomed Steve Wasick onto a bonus episode of The Business of Tech to explain what makes this version of AI unique and a fun use case to bring its applications to life.
With a decade of experience in the generative AI space as a CEO, credit as the Technical Editor for the book Enterprise AI for Dummies, and a law degree, Wasick certainly knows a thing or two about making AI work for clients.
Here’s what he shared about the budding field of data-driven generative AI:
Covering Sports with Data-Driven Generative AI
As I’ve been exploring all of the various cases where we might use generative AI, a recent component of Wasick’s work caught my eye: creating sports articles in real-time. I kicked off my conversation with Wasick by asking why this particular application is such a good use case.
His explanation took us back to one of the core principles of generative AI, which is that it works best with scale. While you wouldn’t necessarily want AI covering the Super Bowl, it can help generate coverage for a Tuesday night game between two high school basketball teams.
A smaller sporting event like that wouldn’t be worth sending a reporter to for an organization like CBS (one of Wasick’s clients), but with AI in the picture, suddenly a news company can create a recap, a game preview, a gambling angle, a fantasy perspective, and much, much more from a single game.
On top of scale, generative AI is amenable to this field because of timeliness. In a competitive news environment, AI enables media organizations to be the first to publish key data and reports. If the betting lines change, they can change or generate a new article to reflect it with ease.
In a time when newsrooms don’t have the resources to dedicate individual people to such niche coverage, AI is extremely valuable.
However, you and I both know that one of AI’s current downfalls is that the technology can hallucinate and get things wrong. I asked Wasick how he balances that risk with accuracy and speed, and he explained something very important: that as a data-driven generative AI company, they do not use the same technology that current language learning models like ChatGPT are using.
While LLMs use a probabilistic method to generate content, Wasick is using what’s called a conceptual automotive:
“It’s basically breaking down the stories into their smallest sub-components and then allowing the system to build them back up using all these individual sub-components. Because of that, we can debug it to know exactly how it wrote. Anything that gets written up by our system, we can trace exactly what happened. And if there’s a problem with it, then we can fix it.”
In short, they’re not using AI to summarize an event like ChatGPT would. They’re using AI to build something from scratch. As a result, his clients don’t have to worry about those hallucinations.
Communicating his AI Model to Clients
Because my goal is to understand how IT services organizations can best advise customers on which AI products are relevant, I asked Wasick how he communicates the essence of his technology to clients.
He explained that because there aren’t that many people in his particular field of AI, they focus on how their methodology differs from that of ChatGPT and other similar AIs when talking to customers. They also emphasize the advantages of using his model to analyze and synthesize large data sets – a major win for organizations drowning in data.
Wasick also shared who is and is not a good match for this type of AI with two factors: scale and complexity of output. If someone’s looking for AI’s help to generate a quarterly report, that’s just not enough data for his technology to sink its teeth into. And if the end goal isn’t a complicated analysis, it also isn’t a fit.
For Wasick, the ideal client is someone who has too much information that needs to be distilled down and lacks a way to analyze all of their target situations.
Integrating Data-Driven Generative AI Into the Workflow
As with any AI application, integrating the technology into an organization’s workflow can be super error-prone. Wasick explained how he sets up his AI in three parts:
The first step is ingesting the data, which can be quite complicated when clients have different data silos. While his technology can pull information from a bunch of different places, they still have to be able to know where each piece is coming from, the structure of the data, and where to look if something’s missing.
The second step is creating a model and generating content. They mainly focus on putting data into a conceptual hierarchy and helping clients identify what they want the output to look like with questions like: Do you want it to be a paragraph? Five paragraphs? What are the most important things to the people who are reading this? Do you have any jargon or any sort of tone?
Finally, the third step is deciding where and how these outputs should be distributed – internally, on a website, etc. Luckily, it’s pretty easy to take automated content and set it up in different formats.
“It’s very easy to segment it once you’ve built it to say, ‘oh, here is the RSS and here’s the email version and here’s the website version’. It’s like the system once it’s set up, can very easily scale to a lot of different distribution points,” he said.
Achieving High-Quality Results
The stakes are high for a use case like this, so I was curious how Wasick is ensuring his AI-generated content is readable, makes sense, and seems natural. He pointed back to the idea of traceability – the first step of his 3-part implementation process.
“Everything that we write, we can see exactly how it created that information. And so if something doesn’t read right, we can typically fix it,” he said.”
And to keep their stories interesting – essentially avoiding the ‘Mad Libs’ approach that other data-driven generative AI companies have – they train their models to focus on the more interesting parts of the data.
Looking at the big-picture impact of this type of AI, Wasick is optimistic that the ability to produce millions of unique pieces of content for communities around the country could help reverse the decline of the local newspaper.
The more use cases like these we add to our AI arsenal, the better we’ll be able to help clients implement effective AI strategies of their own. Have an interesting use case you’d like to share? As always, I’m available at [email protected].
More from MSP Radio
Missed Things?
How about our latest videos to catch you up?
The Daily Podcast available as videos
The Key to Success: Engagement, Algorithms, and Helping Others with Beate Chelette
Exploring the Service Leadership Profitability Report with Peter Kujawa
Public Private Partnerships for Technology companies with Debra Lam
What a bootstrapped entrepreneur thinks of AI with James Benham
Want the Daily News?
All the stories from the daily Business of Tech Podcast are available in the daily digest, and stories are available to everyone for the first five days, and Patreon supporters forever. Catch the audio of the show anytime on Apple Podcasts, Spotify, YouTube, or wherever you find podcasts. Links at businessof.tech
Copyright © 2023 MSP Radio, All rights reserved.
Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list.