S25 Ep2: Eric Siegel - The AI Playbook: Leveraging machine learning to grow your business
ISF Podcast
00:00
Exploring the Evolution and Limits of Generative AI Towards Artificial General Intelligence
Discussion on the potential of AI to simulate human conversation and provide accurate data within specific domains, while skepticism remains around achieving true artificial general intelligence.
Transcript
Play full episode
Transcript
Episode notes
Today, Steve is in conversation with AI expert Eric Siegel. A former professor at Columbia University, Eric is the founder of the long-running Machine Learning Week conference series and a bestselling author. His latest book, The AI Playbook, looks at how businesses outside Big Tech can leverage machine learning to grow. He and Steve discuss the differences between generative and predictive AI, the most effective ways to implement AI into an organization’s operations, and how we might expect this technology to be useful in the future.
Key Takeaways:
1. No matter how controlled or well thought out a project is, any project relying on AI is only as good as its data inputs.
2. The more we learn to differentiate types of AI and apply their functions skillfully, the more we will learn about what is possible.
3. As predictive AI systems emerge, applying quality data analysis to a well chosen project could make a measurable difference for a company’s bottom line.
Tune in to hear more about:
1. Designing a project involving predictive analytics does require quality data and specific domain areas. (3:00)
2. Generative analytics is still in early stages, and popular notions around its use currently differ from what can reasonably be expected or achieved (4:42)
3. Using AI to work with errors and improve a system requires quality data and carefully applied labels (11:59)
Standout Quotes:
1. “It's absolutely critical to have a fine scope, a reasonable scope, well defined for the first project. But the most well defined, sort of, well, scoped project is, in another way, the biggest because really what we're talking about, if you're looking at what should your first opportunity be with predictive AI that you want to pursue, it should be your largest scale operation that stands to improve the most, and that even an incremental improvement provides a tremendous bottom line. -Eric Siegel
2. “ … It's such a funny time, because predictive and generative are really apples and oranges. They're both built on machine learning, which learns from data to predict. But generative isn't a reference to really something specific in terms of the technology; it's just how you're using it, which is to generate new content items. So, writing a first draft in human language, like English, or of code, or creating a first image or video — these endeavors typically need a human in the loop to review everything that it's generated. They're not autonomous. And the question is, how autonomous could they be?” -Eric Siegel
3. “You can only predict better than guessing, which turns out to be more than sufficient to drive an improvement to the bottom line. So who's going to click, buy, lie or die, or commit an act of fraud, or turn out to cancel or be a bad debtor? These are human behaviors for those examples, or it could be a corporate client, or it could be a mechanism like a satellite, or the wheel of a train that might fail. But whatever it is, we don't have clairvoyance or a magic crystal ball. We can't expect your computers to, either. So it's about tipping the odds in these numbers games and predicting better than guessing … no matter how good the data is and how devoid of wrong values and those types of errors, you're still going to have that limitation. There’s still a ceiling. No matter how advanced the method is, it's not going to become supernatural. There's a thing called chaos theory, which basically says that even if you knew all the neurons of every cell of the person's brain, you wouldn't necessarily be able to predict very far into the future. And of course, we don’t. So it's always limited data anyway.” -Eric Siegel
4. “I wrote this new book, The AI Playbook, because we need an organizational practice to make sure that we're sort of planning the project not just technically but organizationally and operationally, so that it actually gets deployed and makes a difference and actually improves operations. And in general, the awareness and understanding of it and how it can be integrated into organizations is still only improving.” -Eric Siegel
Mentioned in this episode:
Key Takeaways:
1. No matter how controlled or well thought out a project is, any project relying on AI is only as good as its data inputs.
2. The more we learn to differentiate types of AI and apply their functions skillfully, the more we will learn about what is possible.
3. As predictive AI systems emerge, applying quality data analysis to a well chosen project could make a measurable difference for a company’s bottom line.
Tune in to hear more about:
1. Designing a project involving predictive analytics does require quality data and specific domain areas. (3:00)
2. Generative analytics is still in early stages, and popular notions around its use currently differ from what can reasonably be expected or achieved (4:42)
3. Using AI to work with errors and improve a system requires quality data and carefully applied labels (11:59)
Standout Quotes:
1. “It's absolutely critical to have a fine scope, a reasonable scope, well defined for the first project. But the most well defined, sort of, well, scoped project is, in another way, the biggest because really what we're talking about, if you're looking at what should your first opportunity be with predictive AI that you want to pursue, it should be your largest scale operation that stands to improve the most, and that even an incremental improvement provides a tremendous bottom line. -Eric Siegel
2. “ … It's such a funny time, because predictive and generative are really apples and oranges. They're both built on machine learning, which learns from data to predict. But generative isn't a reference to really something specific in terms of the technology; it's just how you're using it, which is to generate new content items. So, writing a first draft in human language, like English, or of code, or creating a first image or video — these endeavors typically need a human in the loop to review everything that it's generated. They're not autonomous. And the question is, how autonomous could they be?” -Eric Siegel
3. “You can only predict better than guessing, which turns out to be more than sufficient to drive an improvement to the bottom line. So who's going to click, buy, lie or die, or commit an act of fraud, or turn out to cancel or be a bad debtor? These are human behaviors for those examples, or it could be a corporate client, or it could be a mechanism like a satellite, or the wheel of a train that might fail. But whatever it is, we don't have clairvoyance or a magic crystal ball. We can't expect your computers to, either. So it's about tipping the odds in these numbers games and predicting better than guessing … no matter how good the data is and how devoid of wrong values and those types of errors, you're still going to have that limitation. There’s still a ceiling. No matter how advanced the method is, it's not going to become supernatural. There's a thing called chaos theory, which basically says that even if you knew all the neurons of every cell of the person's brain, you wouldn't necessarily be able to predict very far into the future. And of course, we don’t. So it's always limited data anyway.” -Eric Siegel
4. “I wrote this new book, The AI Playbook, because we need an organizational practice to make sure that we're sort of planning the project not just technically but organizationally and operationally, so that it actually gets deployed and makes a difference and actually improves operations. And in general, the awareness and understanding of it and how it can be integrated into organizations is still only improving.” -Eric Siegel
Mentioned in this episode:
Read the transcript of this episode
Subscribe to the ISF Podcast wherever you listen to podcasts
Connect with us on LinkedIn and Twitter
From the Information Security Forum, the leading authority on cyber, information security, and risk management.
Subscribe to the ISF Podcast wherever you listen to podcasts
Connect with us on LinkedIn and Twitter
From the Information Security Forum, the leading authority on cyber, information security, and risk management.
The AI-powered Podcast Player
Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
