eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
Ishit Vachhrajani, Global Head of Enterprise Strategy at AWS, details how managers can guard against the pitfalls of generative AI even as they leverage its enormous potential.
Among the topics we covered:
- As you watch the explosive growth in generative AI, what are the key challenges posed to the enterprise?
- How do you recommend companies best address these challenges – all while leveraging generative AI for competitive advantage?
- How is the AWS Enterprise Strategy team addressing the generative AI needs of its clients? What’s the AWS advantage?
- The future of generative AI in the enterprise? How do you see it evolving over the next few years?
See a transcript for this interview below the video.
Listen to the podcast:
<
Also available on Apple Podcasts
Watch the video:
This transcript has been edited for length and clarity:
The Power of Generative AI: Four Key Considerations
So generative AI, we believe, is truly a transformative technology of our time, with tremendous potential.
AWS and Amazon broadly has been using machine learning and artificial intelligence in our operation for over two decades now. From pretty much everything – how we sort and route packages to optimizing routes within our fulfillment center, to billions of transactions and interaction with Alexa devices.
As we talk to enterprises especially, there are a few considerations when it comes to generative AI.
The first one is to think about: what problem and opportunity are you going to go after?
So it’s really working backward from the problem you’re trying to solve, rather than actually starting with the solution and then applying back to a problem that you’re looking for.
The second piece is how far down you are with your foundational journey, whether it is cloud or data. Because that’s really going to differentiate those who succeed with this versus those who are maybe successful with one or two pilot, but are not able to scale because they haven’t actually invested in the foundational capabilities when it comes to cloud and data.
The third piece is around responsible AI and governance. What are the considerations in terms of security and privacy of your own data?
What guardrails are you going to put in place within your own organization to validate the outcomes and outputs of some of these models? And what practices you’re going to build inside?
And then really the fourth one, which I think underpins pretty much all three of them: the skill and talent. This is going to require leaders, but also employees and staff at all levels to re-skill and learn new things.
And so what’s the strategy that you have from an organization standpoint to prepare for and skill your staff and build a talent pipeline around that?
I think those are the four key considerations that we see when we talk to leaders around the world around generative AI.
Also see: Top Generative AI Apps and Tools
Generative AI Inflection Point
So if you think about the inflection point with generative AI, it is largely enabled by the proliferation and availability of massive amount of data. But it’s also enabled by cloud, our ability to actually process and make sense of and use that data to train foundational model.
And so as we say, Well, don’t try this at home. This is something that you don’t want to try without cloud because you want to make sure that you are integrating this in the rest of your application and workflow so that this is not just some side project, but actually truly solving customer experience, end productivity, all the opportunities that you’re going to go after within your organization.
And so [it’s important to be] integrating that with rest of your infrastructure, the rest of your application, the foundation of data that is unique to your business. Because that’s going to be your differentiation as well, because you want to use data that you have, and then tune and customize some of these foundational model that power generative AI to serve your customers better.
And all of these will mean that companies that have invested in that foundational cloud strategy: they’ve been on the migration and modernization path. They know what their data strategy is or are actually working on their data strategy and are going to actually stand out when it comes to the usage of generative AI.
Many Opportunities, Increased Productivity
There are just so many opportunities for increasing productivity, from your developer productivity with, let’s say, coding companions like Code Whisperer, Amazon Code Whisperer, which is increasing developer productivity by 57%.
Now, think about pretty much any organization today, and someone is writing a code. And you can actually start to use that today by enabling those developers and engineers to spend more time on solving business problems and challenges rather than the undifferentiated work of writing the basics of the code.
This is where the code companions can actually increase the productivity. What that also means is you can now roll out features and functionality faster for your customers. Think about the opportunity to reimagine a customer interaction.
We have customers who are looking to build a better conversational agent when they have interaction with their customers using generative AI. Where generative AI is assisting the agent on the call with the insight to make sure that that experience every time a customer has an interaction with your business and the company is as fruitful and productive for the customer as possible.
And then you get all the insights back from those interaction, which you can then use to fine tune the model further to then build that flywheel of improvement. Supply chain is another area, content generation, marketing, media and entertainment.
So there are a number of areas where customers are looking to reinvent their business using Generative AI to stand out with the competition.
Also see: The Benefits of Generative AI
Large Language Models
So I think if you think about large language models, they are part of what we broadly describe as foundation models. They’re one type of foundation models that are trained based on massive amounts of data.
What we believe is that there isn’t going to be one model that is going to rule all. That means based on the use case, based on the problem that you’re trying to solve, you want flexibility and choice of the models.
And this is why the approach that AWS has taken, especially with Amazon Bedrock, which is a service that we have announced in preview. It makes a number of these models from third parties like Entropic, Stability AI, AI21 Labs, as well as our own foundational model from Amazon called Titan, available for customers.
So based on the use case that you’re trying to solve for, you may want to choose a different model for that particular use case and the challenge.
Now there will be some companies that actually want to build the foundation model, right?
They may have large amount of data already or they are trying to build a business where they provide others with these large language models like some of our other partners are doing.
And in that case, what is really important is to get a better price performance out of it because it is very expensive and very compute intensive to train and build these large language models.
And this is why we are also investing in custom silicon at the chip level to our processors, which is actually improving the price performance so companies have the ability to do this at a price point that they are comfortable with.
So we’re trying to address this at each stack of this whole challenge because we believe that customers are going to come from different angles, they’re going to have different requirements, and we want to help them all.
Also see: Generative AI Companies: Top 12 Leaders
AWS Enterprise Strategy Team
Let me start with the Enterprise Strategy team, which is the team that I’m part of. We have a team of former customer CXOs from large enterprises. In our previous lives, each one of us led major transformation using data, AI and cloud in large enterprises. So we’ve been in our customer shoes before.
And so what we do is we work with customers to help them with mental models, strategies, lessons learned, in many cases, mistakes that we made – so that they don’t have to make them.
And generative AI, like a lot of other past technological revolution, relies on some of the mental models around operating model changes, culture, reskilling, getting buy-in from your peer executives.
How do you actually bring the technology, but also make it really solve your business problems?
And so those are the kinds of conversations that enterprise strategy team has with executives around the world.
Also see: 100+ Top AI Companies 2023
AWS Generative AI Innovation Center
If we think about broadly across AWS, there are a number of things that AWS is doing to help our customers. We recently announced a Generative AI Innovation Center because a lot of our customers are coming to us and saying, “Hey, Amazon, you’ve been at this for over two decades. You use machine learning in your operation. You’ve been building machine learning and AI services and helping over a hundred thousand customers with machine learning at AWS. How can you help me?”
Because I already have applications running on AWS. I’ve been using AWS for a lot of my infrastructure and my workloads are already there.
Come and help me be successful with the generative AI. So what the Generative AI Innovation Center does is: we bring our expertise of solution architects, strategists, data scientists and engineers to work with customer in a very tangible and practical way to apply generative AI to their business.
And we’re making a hundred-million-dollar investment in the Generative AI Innovation Center. Skill and training and talent development was the other part that I mentioned earlier.
We launched a course in partnership with Andrew Yang on Coursera on generative AI with large language model. It has gotten tremendous response because we believe that not only large companies but pretty much across the sector – companies of all size, even individual builders and developers – would want the skill to take advantage of this.
So there are a number of ways beyond our choice of product and services that I talked about that we are helping our customers be successful.
Innovation Center is basically a function or a team. Think about that as a cross-functional team of experts from AWS that then work with the customers to deeply understand their problem and then apply generative AI toward it.
Looking Ahead: Generative AI Evolving in the Enterprise
We like to joke in our team that “today was the longest year in generative AI.” (Laughing.) The pace of change is remarkable and truly mind boggling and it’s amazing, it’s fascinating.
But when you think about the pace of change with any technology and what’s going to change, I like to actually flip it and think about what is not going to change.
And if you think about generative AI, what is not going to change is that businesses are still going to look for flexibility and choice. They’re still going to look for convenience and ease of use. And they’re going to still look for better price performance overall.
And so if you think about those three aspects from a business standpoint, from an enterprise standpoint, as a company leader within an enterprise, you want to make sure that when you are building your generative AI strategy, you’re partnering and thinking about choice of models.
Because if you look at just, in the last six to 12 months, the size of the models, the number of parameters that they are now trained on, has been exponentially growing. And that pace will continue to change.
There’ll be better models in the future. There’ll be newer models. There’ll be models that will be very specific to industries and use cases that you may be trying to solve.
And so that flexibility and choice is going to be really, really important. And this is why the approach that we were taken with Bedrock is to offer the choice and flexibility.
The second is the ease of use. Again, I’ll go back to integrating this in your workflow and application. So that generative AI does not become something that people go to take advantage of a generative AI powered application somewhere else, and then come back to their workflow, where they’re still using traditional infrastructure in the tool.
You want to integrate that into your core systems and workflow. This is why end-to-end stack and connectivity throughout different services and products that you use is going to be critical.
And then the third piece is price performance. Because even within a company, you may have some use cases. For example, I talked about a coding companion where you don’t need to build your own foundational model. You just want to call an API or provide a service to your developers where they just use generative AI in their job.
But there may be other use cases where you may want to take some foundational model, tune it with your data and build your own application on top of that.
So those are the three things that will remain unchanged. And I advise customers to build your generative AI strategy around that. What is going to change is obviously the investment in training and reskilling. And so that is something that leaders have to stay on top of for themselves, as well as enabling that people.
How Generative AI Moves Life Faster
Think about, this is why I like to apply the lesson of history to the future, right?
Because history has taught us many things. Think about data, right? Several years back when companies started to build data strategies, the data team was independent, data tools were independent. But really when the power of data was unleashed, when it was actually made available in the context of a workflow or a transaction.
So if you’re a salesperson talking to a customer about a deal, you want the data in the context of that transaction. If you are a customer support agent on a call with the customer, you want data in the context of that transaction.
And I think same mental model applies to generative AI, whether you’re trying to solve for supply chain, or if you’re in marketing, building, and using generative AI for your next branded campaign. Or you are a storyteller in media and trying to build the script ideas for your next show.
You want generative AI to be part of what you do everyday, not just yet another thing in a side show that you need to go to.
