"Artificial intelligence (AI) lies because it is creative. In time, there will come a day when it imitates humans."
Dona Sarkar, Microsoft (MS) AI and Copilot diffusion lead, said this in a recent interview with ChosunBiz. Sarkar said, "As AI makes mistakes or lies, it can come up with new ideas or produce creative solutions." She oversees work at MS to help corporations and developers worldwide use the AI agent Copilot and AI technologies.
Sarkar stressed that the prevailing perception that AI will completely replace human jobs is wrong. She said, "Marc Benioff, Salesforce chief executive officer (CEO), said AI accounts for 30% to 50% of work at companies, but that work may be what AI has been doing for 20 years," and added, "Anthropic CEO Dario Amodei said half of the office jobs among new hires next year will disappear, but AI cannot fully replace people." People prefer to talk to real people rather than machines, she noted. She went on, "Nvidia CEO Jensen Huang said that thanks to AI, people can do more creative and meaningful work, and jobs will not decrease but instead increase."
Sarkar argued that advances in AI will instead increase developer jobs. She explained, "When MS released Visual Basic in 1991, it was expected that developer jobs would shrink, but in reality, as developers built webpages and applications, jobs increased." The following is a Q&A with Sarkar.
—There is a view that software as a service (SaaS)-based applications (apps) will disappear and AI-based business agents will take their place.
"It will depend on the situation. New apps will evolve by embedding AI. In fact, apps with embedded AI are gradually increasing, but existing SaaS-based apps will not suddenly disappear. For example, apps built in 1995 still exist. New apps with embedded AI will emerge, but it is hard for corporations to discard the apps they built in the past and use only new ones. That is why I think new AI-based apps will coexist with existing systems."
—Will a new business trend using AI agents take root?
"I think this is entirely possible. For example, if I ran a fashion retail business, I could have four independent AI agents handle each area: customer service (CS), inventory management, shipping management, and marketing. The customer service AI agent can forecast demand through communication with customers, and the shipping management AI agent suggests optimal shipping methods. This way of multiple AI agents collaborating and dividing work will gradually become commonplace in business."
—There are also concerns that AI agents could replace jobs.
"I do not think AI replaces people's roles. AI is useful for handling repetitive tasks or routine chores, but it cannot substitute for people's creative and strategic roles. For example, tasks such as summarizing executive meeting minutes or checking emails can be handled by AI, but critical decisions or strategic judgments still belong to people. AI will be used as a tool to assist human work."
—Didn't you say that if AI lies, it can be seen as part of creativity?
"Yes. The reason AI lies is precisely because it has creativity. For example, a generative AI may (falsely) draft a customer complaints list incorrectly, but in the process, new ideas or solutions can emerge. Of course, AI must not repeat lies, and correcting them is entirely up to people."
—Your outlook for the future of AI agents.
"AI agents will evolve to collaborate with people. Rather than completely replacing humans, AI will help in areas humans cannot handle. For example, AI will take on tasks that are hard for humans, such as complex work or real-time multilingual interpretation. However, the most important thing in using AI agents is "trust." Trust is secured only through human intervention. When an AI agent makes a mistake, a person must intervene to fix the problem. In the end, AI and humans will inevitably develop into a symbiotic, collaborative relationship."
—Is it necessary to develop AI agents suited to each country's culture and language?
"AI must be developed to fit each country's culture, laws, and language. For example, an AI agent used in Korea must be designed for the Korean language and culture. MS cannot build a separate AI agent for Korea. AI that accounts for Korea's legal requirements and cultural characteristics must be developed by that country's own developers. For instance, an AI agent suitable for Korea's aging society will be needed, and such AI should be developed in Korea and reflect local data and regulations."
—What matters when people use AI agents?
"AI agents are very useful as tools that assist human work. But the important thing is to correctly judge AI's applicability and clearly distinguish what AI cannot do. Also, because AI can make mistakes, a process to correct them is essential, and people must always think about how to control AI and fix problems. AI is a collaborator; it cannot fully replace people."