As AI applications grow rapidly, Andrew Ng, founder of DeepLearning.AI and managing general partner at AI Fund, emphasizes that Generative AI (GenAI) encompasses much more than just chatbots, highlighting a range of practical scenarios in which the new technology can be applied.
Ng acknowledges that while users may be experiencing chatbot application fatigue, there remains significant potential for enhancing and leveraging the product across a range of sectors.
Taiwan connection
He advises companies to identify industry-specific AI use cases before deciding whether to develop solutions in-house or outsource them. Ng has deep ties to Taiwan, collaborating with Foxconn and the Taiwan-based engineering team at DeepLearning.AI. In May, Taiwan's National Science and Technology Council announced it would work with Ng's AI fund on venture capital projects, potentially with joint ventures and capital investments.
Mr. Ng offered his views at AI Fund's recent online forum, where he spoke with Bloomberg Beta head Roy Bahat. Mr. Ng joined Amazon's board in April but did not represent the company at the forum.
Big League Competition
Discussing the current state of AI development, Ng noted that AI is capable of performing an increasingly wide range of tasks, bringing opportunities for growth but also associated risks.
Responding to Sequoia Capital's article, “AI's $600 Billion Question,” Ng explained that the article focused on factors such as model training, GPU procurement, and capital investments. He reassured that companies need not worry too much as long as they are not directly competing with model training providers such as OpenAI, Anthropic, or Google.
He further explained that the significant investments by these leading companies will benefit the entire ecosystem by reducing costs for application developers.
Build or buy?
Ng emphasized that it is important for companies to define industry-specific use cases when deciding whether to build or buy an AI solution. He suggested that knowledge workers should leverage GenAI tools to increase productivity and that companies should provide employees with the necessary training.
“Those who are skilled in using AI will outperform those who are not,” Ng said. But he noted that the transition won't happen overnight and that some of the risks may be overblown. He also said that certain industries, such as finance and healthcare, typically have inherent risk management mechanisms for AI applications.
Chatbot fatigue: The applications are multiplying
Ng also noted that tools provided by companies such as OpenAI have lowered the barrier to developing chatbots, resulting in a proliferation of chatbot applications and user fatigue. But there are many other areas of consumer application to consider. For example, payment service providers can integrate AI capabilities into their back-end systems to classify spending into entertainment, equipment and other categories, improving the usability of their products.
A few years ago, developing an application could take up to six months, including market research and product development, and there was concern about recovering development costs. The barriers to development have been significantly lowered, allowing developers to release multiple applications simultaneously to test market reactions, allowing even a few successes to generate significant value.
Ng also highlighted the emerging trend of agent workflows or AI agents that automate lengthy processes from a single command. For example, in document writing, an AI assistant can draft, research, revise, identify issues in a document, and suggest improvements to make the corrections. This holistic approach can significantly increase user productivity compared to issuing a single command to a language model.
