In 2023, organizations across industries raved about the groundbreaking potential of generative Artificial Intelligence (gen AI). But discussions about the promise gen AI offers can only go so far. Fast forward a year later, companies are now actively testing this technology’s advantages by integrating it into their operations.
Cloud service provider Amazon Web Services (AWS) is facilitating gen AI adoption among their organizational clients. Its three-layered gen AI stack has allowed businesses to choose the solution that best first their requirements, whether it is a pre-built, integration-ready AI assistant, access to foundation models (FMs) and Large Language Models (LLMs) to build custom applications, or the infrastructure to set up and train their own FMs.
In an exclusive interview, AWS Asia Pacific chief technologist Olivier Klein gave a preview into developing trends among organizations adopting AWS gen AI solutions and a sneak peek into the predicted use cases for this technology in the Philippines.
From potential to practical, integrating AI today
One of the first trends that Klein spotlighted was the growing need for gen AI that could quickly, easily, and safely integrate into existing operations.
Klein explained that most companies realized they are either looking for FMs and LLMs or assistants that can seamlessly fit into their current processes.
“So instead of just building a gen AI model on the side, which is a nice to have…[we’re] really thinking about how we integrate them into systems. How do we integrate them into existing ERP or CRM systems? How do we integrate them into existing work flows or business flows?” Klein elaborated.
In order for gen AI to smoothly merge into existing systems, Klein stated that models have to be able to securely integrate into a company’s databases, data repositories, and applications.
This way, not only can gen AI generate answers based on a business’ data, they are constantly updated, constantly learning through the apps and data repositories they are linked to.
Klein emphasized: “That is what we see this year being vastly different from last year where it was like, ‘okay we have a model that is intelligent but it only knows what it knows now, so it’s static. Now we have dynamic models’.”
Klein pointed out that since many organizations are already running on AWS Cloud, they can more easily incorporate gen AI into their operations “AWS’ Integration points are a true differentiator… everyone gets access to the same pre-trained foundation models. The real value comes in when I can actually have these models understand my data and do so securely.” he asserted.
Klein added that the AWS AI assistant, Amazon Q, typifies this effortless integration with the capability to connect to both Amazon products and over 40 business tools, including Gmail, Microsoft Exchange, and Salesforce.
Through these integrations, Amazon Q can filter through vast amounts of company data, conduct analyses and summaries, then engage in dialogue with end-users about the data as it updates.
Similarly, Amazon Bedrock, the AWS service that offers a selection of FMs and LLMs for businesses to build their own gen AI apps, also has capabilities that enable it to integrate into existing systems, such as Agents for Amazon Bedrock. These agents give the pre-trained models on Bedrock access to apps that acquire current information.
Boosting productivity with assistants and multimodality
Other than trends concerning the nuts and bolts of how companies are adopting AI, Klein revealed that some of their clients utilizing gen AI are already experiencing productivity gains.
One of the top productivity gains was in the software engineering field. Apart from helping businesses more efficiently access their internal data or speed up daily productivity tasks, AWS bills Amazon Q as an accelerator for software development.
In fact, as early as now, Amazon Q is helping software engineers code faster with real-time code suggestions, streamlining legacy system modernization, and assisted documentation.
For example, an internal AWS study recorded that when upgrading from Java 8 to Java 17 with Amazon Q’s support, over 1000 applications were upgraded in two days with an average of 10 minutes spent on each application.
Klein additionally reported during the interview that a number of their clients reduced the time they spent on code documentation by up to 75% with Amazon Q.
He also drew attention to how businesses are exploring possibilities created by the trend towards multimodality or models’ capability to process information in multiple forms.
In other words, models are not only accepting and outputting one data type like text, but are starting to receive and generate multiple types encompassing photos, videos, audio, and other abstract data forms.
“Last year, we started seeing LLMs and Large Visual Models evolve separately. We can see the combination of both now happening. There’s multimodal models that can both understand visuals, images, graphs, and tables combined with the question you contextually throw against these models that opens up complete new scenarios and use cases that we see across the board already with many customers,” Klein said.
He further illustrated: “For example, in the case of an insurance company, you can bring the model pictures and have it describe the damage to a vehicle. Or in the case of software engineers, you can prompt it to write a python code based on the excel sheet you send it.”
Visualizing Gen AI in the Philippines
When the discussion turned to gen AI in the Philippines, Klein eagerly laid out how enhancements in the LLMs and FMs’ language capabilities are improving the models’ usability in South East Asia, including the Philippines.
At the same time, Klein held that there are ways current gen AI models can be useful for several of the country’s industries, even without full competence in Filipino or other dialects.
For instance, he jumped off the topic on multimodality to point out how this AI trend can open opportunities for the Philippines’ BPO, marketing and advertising, VFX, as well as other creative industries.
He detailed how BPO agencies and call centers can employ today’s Gen AI models to record live voice transcriptions and suggest potential answers to agents while they are on call with customers.
Klein also described how any marketing, digital, or creative agency can enhance their creative process with AI. This technology can support these companies in creating hyper personalized ads or simply speed up the ideating process by producing numerous drafts swiftly.
“If I talk more broadly in terms of what we offer and what we see out there, [gen ai] dramatically boosts the creative process because I go, give me one example, give me a hundred examples of this,” Klein observed.
“I might not use it as is, but it allows me to be faster at iterating my creativity…it’s really just about iteration, the faster I can experiment, the easier it will be for me to come up with new ideas. I think that will be transformative for the entire creative space.”
Klein stressed, however, that gen AI should not replace the artist, but act as a support. “I see it as an augmentation. You’re not going to replace the artist as such because you still want the complete final product… but you iterate a lot quicker,” he said.
In short, gen AI is rapidly becoming a requisite productivity tool for the workplace. Moreover, the its adoption speed will only increase as innovations like multimodality open more applications for it. It is up to individual organizations whether they will be the among the first-movers or the laggards in the gen ai race.
