A debate has become very complicated in the technology world right now: Will artificial intelligence (AI) replace humans? Or will it qualitatively enhance human efficiency?

The AI ​​conference of the ‘Web Weekend Kathmandu’ held in Kathmandu on Saturday debated this very question. The conference, which brought together more than 400 software engineers, designers and technology experts from various countries, including Nepal, brainstormed on how to connect AI to real work performance and human value, without limiting it to the realm of ‘chatbots’. Manu Chatterjee, a speaker at the

conference and AI head at Leapfrog Technology, discussed the current state of AI and its technical limitations, emphasizing that it should be understood as a powerful tool rather than magic. “The perception that AI is just a chatbot that asks and answers questions is incomplete,” Chatterjee said. “Its real power lies in how to turn messy and scattered data into meaningful decisions.” Discussing the great leap forward in

technology, he said that while earlier people planned and computers worked, computers have now developed the ability to plan and work. However, he also warned that one should not trust the ‘Large Language Model (LLM)’ completely. According to him, the LLM is not a database search machine, but rather a ‘guessing machine’ that generates possible results based on patterns. Chatterjee urged people to understand the ‘hallucinations’ (the tendency to give false information) seen in AI not only as a weakness but also as an aspect of its ability to create new things.

Sayantika Banuk, founder of Datajourney HQ, called AI a ‘black box’ and urged caution in its commercial use. ‘AI is now as powerful as the advent of the internet.’ But if AI creates more confusion for users, there is a need for improvement,’ she said. Banuk advised startups and entrepreneurs to separate their business logic from the volatile nature of the AI ​​model. She argued that AI should be used only as a supporting layer, while preserving the ‘main brain’ of the business as technology keeps changing.

AI researcher Haihao Liu has challenged the popular Silicon Valley slogan of ‘move fast and break things’. Liu, who is also the founder of Trinity Education, said that human sensitivity and ethics should be given top priority in the development of AI.

“I encourage people to slow down a bit,” he said, recalling the story of the tortoise and the hare. “Running fast doesn’t always mean winning. Long-term success requires time to think and understand.” He suggested that the AI ​​field should abandon the illusion of “bigger is better” and invest in small and efficient models that consume less energy, are safe, and meet real-world needs.

Platform engineer Vlad Dianchenko offered a different perspective on the future of programming languages. According to him, the programming language that is fully suitable for the AI ​​era has not yet been born. “The winner will be the language that is concise, meaningful, and friendly to AI agents,” he said. He analyzed that the current popularity of Python and TypeScript is not because they are superior, but because of their large communities and “general purpose” nature.

AI expert Dobhan Rai has questioned the ‘intellectual capacity’ of AI. ‘Just because an AI system shows clever behavior does not mean it has a strong world model (view of understanding the world),’ Rai said. She pointed out the positive aspect that AI has forced more discussion about humanity, but expressed concern about ‘agentic AI’ (AI that can make decisions) without understanding. ‘If AI has high decision-making capacity but human empathy and real understanding are zero, that can be a very scary scenario,’ she argued.

At the conference, 21-year-old engineer Amit Timalsina discussed the health sector, giving an example of the practical use of AI. Amit, who has 5 years of experience in the AI ​​field, especially talked about the changes that AI can bring in the clinical trial process of drugs. ‘Currently, it takes about 8 years for a drug or vaccine trial to be approved. This time and complexity can be significantly reduced through the use of generative AI, he said. He claimed that the architecture that extracts accurate information from complex medical documents and simplifies clinical processes will make healthcare accessible in the future.