S7E9: Using AI Agents and Models Effectively feat. Derek Egan

We continue our series of conversations about Artificial
Intelligence and how it's changing the world around us. In
today’s episode, Frances talks to an AI engineer at Google about
the ins and outs of the newest crop of AI models and agents.





Joining the show to give an insider’s perspective is Derek Egan.
Derek is on the Vertex AI team at Google Cloud where he works on
helping developers build generative AI-powered agents for their
own applications. Derek is also a graduate of the David Eccles
School of Business with a degree in Operations Management. 





David and Frances discuss the difference between models and
agents, and why humans are still very much integral to the
process. They address the strengths and weaknesses of these types
of systems and the ways in which they might be used by
organizations most effectively to help them work more
efficiently. Derek speaks to the realistic worries of some AI
critics and gives some of his own real-life examples of how AI
agents help streamline his work life and personal life. 





Eccles Business Buzz is a production of the David Eccles School
of Business and is produced by University fm.



Show Links:

Derek Egan | LinkedIn

Vertex AI | Google Cloud

Master of Science in Business Analytics - The David Eccles
School of Business

David Eccles School of Business (@ubusiness) • Instagram

Eccles Alumni Network (@ecclesalumni) • Instagram 

Eccles Experience Magazine




Episode Quotes:




Google Cloud invests in explainable AI and safety filters
to build AI that benefits everyone





[09:06] How can we build AI that's beneficial for all is a
critical question. At Google Cloud, we take that really seriously
for responsible AI. It all starts with a set of AI principles
that translates into multiple things in practice as we build
these products and up front and not as an afterthought. A few
examples of that are ethical reviews. So,  internally, there
are dedicated teams that essentially conduct  ethical
analysis and risk assessments for any new product that's going to
launch and their job is to identify potential biases. [09:57] We
haven't been perfect here and no one has been perfect here
because it's a process and we're all learning. And that's also
why, as we gain learnings and learn how to interact with these
models and how to control the types of responses, we're also
building explainability into the models themselves. That's really
important. 





AI will not take away your job but someone who uses it
better might. 





18:35 If I'm a student at the University of Utah right now in the
David Eccles School of Business or an alumni, I would be pretty
anxious about this, right? Like, hey, how does AI impact my job?
What I would encourage people though, is to think about reframing
that question.  And the way I reframe it is I don't think AI
is going to replace your job, and I don't think you're competing
directly with AI for your job. I do think you're going to compete
with other people who can competently use AI to do your job and
that will lead to disruption. That will lead to a change in the
way that works done, day in and day out. So, my guidance, my
invitation would be to try it out, embrace AI and start using it.
See how it can make you more productive. See how it can take the
nitty gritty tasks off of your plate and help you to automate
them on your behalf.





Derek shares a developer's guide for building AI
applications.





32:39 If you are a developer and you want to build an
application, my guidance would be to identify something that's
fairly low hanging fruit, fairly simple, where you're going to
get started. And where it's going to be an interaction between
the person and the agent and not something that's going to go run
autonomously and not have the right feedback loop. And then you
can measure and get a sense for how helpful is that agent, and
then iterate on it. And I'd encourage you to check out some of
the use cases. We have like 185 customer use cases on a blog post
published from Google Cloud. But there's a bunch of examples out
there and there's a bunch of documentation that we offer as well
to help people get started.