Two recent Financial Times and Straits Times articles, “AI’s Double Bubble Trouble” by John Thornhill [1] and “Chatbots Are a Waste of AI’s Real Potential” by Gary Marcus [2], capture a defining paradox of our time. One warns that AI’s market valuations are inflating faster than its real-world impact, while the other laments that AI’s brightest minds are building conversational tools instead of scientific breakthroughs. Both are right, but both miss a larger point.
AI today is not just a technology boom. It is a societal test. Thornhill’s concern about speculative excess is valid: when companies trade at 225 times earnings or promise trillion-dollar transformations before delivering measurable value, the risk of a financial bubble looms large. Yet beneath that excitement lies a quieter revolution where teachers use AI to simplify lesson plans, small businesses optimize energy use, and non-profits analyze data that was once locked behind technical walls. These are not speculative ventures. They are examples of AI democratizing data analytics at the ground level.
Marcus is equally correct that chatbots alone will not cure diseases or engineer new materials. But dismissing them as “a waste” misses their role as gateway technologies. Generative AI interfaces lower the barrier between human intent and machine reasoning. They allow billions of people to converse with data using natural language rather than code. This accessibility forms the foundation for more specialized, domain-specific AI to develop. Chatbots are not the pinnacle of AI. They are the bridge between human imagination and scientific application.
The real danger is not that society focuses too much on chatbots or that investors chase speculative valuations. The danger is that we create an AI divide. If advanced AI systems are accessible only to corporations and research labs while the public remains confined to surface-level tools, we risk reproducing inequality in digital form. The “AI haves” will innovate, while the “AI have-nots” will only consume.
What we need instead is a tiered vision of AI democratization:
1) Entry-level AI such as chatbots to empower everyday users
2) Intermediate AI for professionals in healthcare, education, and engineering
3) Advanced AI for scientific discovery and societal resilience
Each level should strengthen, not isolate, the others.
AI’s ultimate success will not be measured by stock prices or the number of startups it spawns. It will be judged by whether it lifts human capability across all levels of society.
The task ahead is not to choose between speculation and science, or between chatbots and super-intelligence, but to ensure that AI remains a bridge, not a barrier, between progress and people.
References
[1] J. Thornhill, “AI’s double bubble trouble,” Financial Times, Oct. 17, 2025. https://www.ft.com/content/da16e2b1-4fc2-4868-8a37-17030b8c5498
[2] Chatbots are a waste of AI’s real potential
https://www.straitstimes.com/opinion/chatbots-are-a-waste-of-ais-real-potential.
https://www.straitstimes.com/opinion/chatbots-are-a-waste-of-ais-real-potential.
