New Delhi: The rise of artificial intelligence has triggered large scale fears about job losses. But according to Vishal Mishra of Columbia University, current AI systems like Large Language Models (LLMs) are not as big a threat to jobs as many believe—at least not yet.
LLMs Are Powerful, But Constrained
LLMs such as ChatGPT and other generative AI tools are designed to process and generate text based on existing data. The LLMs can write emails, summarize documents, generate code, and help with research. Yet, their intelligence is largely based on patterns, not true understanding.
Add Zee News as a Preferred Source
Prof. Mishra explains that these systems do not “think” or “create” in the human sense. Instead, they predict the most likely next word or idea based on massive datasets. This means they are excellent assistants—but not replacements for human workers in most roles.
For example, an LLM can help draft a business proposal, but it cannot independently identify a completely new market opportunity or design a breakthrough business model from scratch.
Why Most of Jobs Are Still Safe—for Now
According to Mishra, most jobs involve a mix of routine and non-routine tasks. While LLMs can automate repetitive parts—like documentation, data entry, or basic analysis—they struggle with tasks that require judgment, creativity, or real-world context.
Look at professions like doctors, engineers, or journalists. AI can help them by speeding up research or drafting reports, but final decisions still depend on human expertise and accountability.
Even in creative fields, LLMs rely primally on existing content. They modify ideas rather than invent truly inventing original ones. This restrict their ability to disrupt jobs that rely on innovation.
Columbia CS Prof Vishal Misra explains why LLMs can’t generate new science ideas.
Bcz LLMs learn a structured map, Bayesian manifold of known data & work well within it, but fail outside it.
True discovery requires creating new maps, which LLMs can’t dopic.twitter.com/PzI0YrTTeT
— Rohan Paul (@rohanpaul_ai) April 21, 2026
AGI: The Real Deal
The larger concern, Mishra notes, is Artificial General Intelligence (AGI)—a future form of AI that could match or exceed human intelligence across a wide range of tasks.
Unlike LLMs, AGI would not just follow patterns. It would be capable of reasoning, learning independently, and generating genuinely new ideas. This is where the real disruption could begin.
For example, while an LLM can suggest improvements to an existing product, an AGI system could invent an entirely new product category—something humans haven’t even imagined yet.
Likewise, in scientific research, today’s AI can help analyze data. But AGI could potentially form new theories, design experiments, and make discoveries without human guidance.
Creativity: The Key Difference
One of the stark distinctions lies in creativity. LLMs are constrained by their training data. They can combine ideas in interesting ways but cannot truly “break the mold.”
AGI, on the other hand, is expected to go beyond this limitation. It could think conceptually, challenge assumptions, and generate original insights—something that defines human intelligence today.
Mishra highlights that this ability to create new knowledge, rather than just process existing information, is what could eventually make AGI a serious competitor to human workers.
The Way Ahead
For now, the impact of AI is more about augmentation than replacement. LLMs are tools that enhance productivity rather than eliminate jobs entirely. Workers who learn to use these tools effectively may even gain an advantage.
However, the long-term picture could change dramatically if AGI becomes a reality. While that milestone is still uncertain, experts agree it would mark a fundamental shift in how work is done.
Until then, the message is clear: today’s AI is a assistant, not a replacement. But tomorrow’s AI could be something entirely different.