TMTPOST--As artificial intelligence (AI) technology rapidly advances, the focus has shifted from merely developing the technology to applying it effectively. For companies providing these large AI models, the crucial task is now to help users implement them efficiently and harness AI's power to drive success.

In light of this, Zhao Hejuan, the founder, chairperson, and CEO of TMTPost had an in-depth discussion with Ishit Vachhrajani, the Global Head of Global Enterprise Strategy at Amazon Web Services (AWS).

Ishit told TMTPost that businesses across various sectors are actively embracing the benefits brought by generative AI (GenAI), which create certain types of images, text, videos, and other content in response to prompts. In this process, companies must choose models suited to their specific business scenarios while ensuring data security and selecting the appropriate model products for different use cases.

If 2022 is the first year of large AI models, 2024 marks the first year for practical applications of the models. Based on the current market trends, it’s clear that large model vendors are focusing not just on increasing parameters but on how to apply these models in industry. The emergence of models based on the Mixture-of-Expert (MoE) architecture and the application of few-shot learning and fine-tuning techniques support this shift.

AI large models are infiltrating businesses and industries at an unstoppable pace, and the advent of GenAI has shifted the demand for cloud computing and AI technology from technical fields to the business community. Market research predicts that by 2025, the global GenAI market will exceed $10 billion, with a significant portion dedicated to enterprise-level GenAI applications.

However, Ishit believes the GenAI race among tech giants has just begun, but there are a wide array of applications going forward. He offers some advice for Chinese companies looking to leverage GenAI for their businesses.

Firstly, companies should focus on solving problems rather than the solutions themselves and select influential, relatable content.

Secondly, companies need to refine their data strategies, ensuring they have the right data culture to unlock more value. Ishit emphasizes the importance of investing in infrastructure and data environments to gain an advantage when using data.

Thirdly, users should consider the deployment intent of GenAI. "Your goal should be to truly unleash the productive value of GenAI. Once the proof of concept (POC) is successful, you can continue to expand," said Ishit.

Fourthly, corporate leaders should pay close attention to GenAI. "Leaders need to take the time to accumulate their knowledge," Ishit noted. AWS has been committed to training over two million people by 2025 under the AI Readiness program to help companies better leverage GenAI.

When asked about the anxiety of Chinese companies in terms of GenAI, Ishit reassured them that they are not alone. The development of GenAI has just begun, and companies worldwide are still exploring its potential. He pointed out that the challenges faced by Chinese companies in using GenAI are similar to those encountered by companies in the U.S. and other regions.

To better assist enterprises in implementing GenAI applications, more tech giants are launching platform-based large model products. Ishit thinks that from an application perspective, larger models are not always better. Companies need to select AI model products that suit their specific scenarios, choosing different tools for different tasks.

For instance, in pharmaceutical research, larger models with more tokens and parameters are crucial. However, in fields requiring quick responses like intelligent customer service, smaller models can offer lower latency and higher cost-efficiency. Companies need to balance accuracy, performance, and cost to achieve the best results.

Giants like AWS, IBM, Microsoft, Alibaba Cloud, and Baidu are focusing on platform-based large model products as a key area of development. Clearly, platform-based, diversified large model products have become a competitive market that many tech giants are vying to dominate.

Despite the differences between China and the U.S. in the GenAI market, Ishit sees China's mobile-first consumer ecosystem as a significant advantage in building innovative applications, even if there are still technological gaps with the U.S. He also highlighted the unique challenges posed by regulations and compliance in different countries.

AWS has a long-term strategy of helping enterprises implement GenAI applications through platform-based products. In 2023, AWS launched Amazon Bedrock and its self-developed large language model Titan. This year, AWS introduced new features for Amazon Bedrock, providing customers with a simpler, faster, and more secure way to develop advanced GenAI applications.

Customers now seek more cost-effective, high-performance, low-latency multi-model hybrid solutions instead of just one large model product. AWS's Amazon Bedrock allows users to access leading foundational models from companies like AI21 Labs, Anthropic, Cohere, Meta, Mistral AI, Stability AI, and Titan. The newly upgraded Amazon Bedrock also features proprietary model import capabilities, enabling enterprises to integrate custom models into the platform with just a few clicks.

Ishit mentioned that regulated industries like financial services, healthcare, and life sciences are ahead in applying GenAI. These industries use GenAI to enhance customer experiences and drive data-driven product value.

For example, global pharmaceutical company Merck used AWS's GenAI technology to reduce the false rejection rate in drug manufacturing, achieving significant cost savings. In the CAD field, companies like Autodesk apply GenAI for predictive maintenance, reducing downtime and optimizing manufacturing processes.

In the finance sector, GenAI is used to enhance customer interactions and provide personalized services, such as customizing insurance policies or streamlining loan documentation.

As enterprises adopt a "cloud-first" approach, data security and privacy protection become crucial challenges in the AI era. Ishit emphasized the importance of privacy, security, and responsible AI for all users. AWS remains cautious in launching AI-related products, focusing on providing secure, resilient, and compliant solutions.

Security is a priority for AWS, which offers solutions like sensitive data protection using machine learning and pattern matching to help customers identify and manage sensitive data. AWS's approach ensures that AI and machine learning are integral to the company’s DNA, improving customer experience and driving innovation across all business areas.

By integrating AI into their corporate culture, enterprises can better leverage AI technologies. AWS aims to use GenAI to innovate at every layer of the IT stack, ensuring comprehensive, secure, and scalable AI solutions for their customers.

Using retail services as an example, Amazon leverages AI technology to optimize mobile tray movement and robotic picking routes, thereby enhancing product supply chain production and inventory management. This is not a common scenario. In fact, AI technology, represented by machine learning, has been embedded into Amazon's DNA, Ishit pointed out.

Meanwhile, Ishit told TMTPost that AWS views GenAI in the same way it does cloud computing: not only by using these advanced technologies themselves but also by scaling and transitioning their practical experiences to benefit their clients' businesses.

AWS's approach to using GenAI is to ensure innovation at every layer of the IT stack, Ishit said.