Can GenAI bots win Nobels? They’ll soon be pushing the borders of knowledge

Time forgot its own shadow.” This five-word profound-sounding statement evokes a world where the past has no presence, the present is untethered, and the future becomes a void, questioning the essence of existence itself. It appears like a statement from a wise sage that one can quote and re-quote in different contexts.

In reality, it was ChatGPT’s reply to a prompt that wanted the shortest possible meaningful statement that had never been recorded before. Google Search verified that this statement does not exist on the internet—or, at least, not until this author started writing about it.

Also read: Errors, high cost among reasons GenAI not moving beyond concept stage

This intriguing result invites us to explore the growing capabilities of Generative AI, or GenAI, and its implications for our understanding of creativity and knowledge.

Going beyond the bounds of traditional AI: Before the rise of GenAI, traditional programs struggled with tasks involving creativity or originality. The idea of creating a meaningful, novel statement was a challenge even for the most sophisticated systems.

Traditional algorithms operated on pre- existing data, often failing to generate truly unique content. These systems could shuffle words and phrases, but lacked the ability to craft something genuinely new.

In contrast, GenAI models like ChatGPT are built to push boundaries. They utilize extensive data-sets and advanced algorithms to generate outputs that are recycled versions of existing knowledge and truly novel creations.

One can now push the latest chatbots to tackle complex tasks like formulating new conjectures akin to Fermat’s Last Theorem or offer fresh explanations in quantum physics. The results are nothing short of astonishing. These advanced models are proving their ability to contrive completely novel ideas seemingly from thin air.

While the quality of these outputs is still under scrutiny, even the staunchest critics must concede that these machines are advancing at an exponential rate, continually improving their creativity and problem-solving prowess.

Mounting evidence suggests that GenAI is not just a mirror reflecting human thought, but is now an active participant in creating new ideas. The most important result is in how these models are reshaping scientific investigation. They can be synthesized to generate hypotheses, identify patterns and suggest novel research directions.

Many renowned theoreticians have believed that machines will not be able to start any scientific investigation by coming up with new hypothesis even if their roles increase in scientific research (shorturl.at/04QIc). Transformer models are already proving these presumptions wrong.

Also read: Who’s making money from GenAI? Big Tech, consultants or data centres?

This leap in capability is due to the underlying architecture of GenAI. These models are trained on vast amounts of data of all type and can identify patterns and structures within this data in a way nothing in our toolkit was able to do. Apart from their abilities to hold vast context, the models unearth inter-relationships across modalities and disciplines.

The Super-Moore’s Law era: As humans, we’re naturally inclined to believe that nothing, not even our clever creations, can surpass our intelligence, at least in critical areas. After all, we’ve already ceded superiority in raw speed, calculation and visual acuity to machines. Admitting inferiority in intelligence is forcing sweeping changes in every realm of life, but the bigger problem is in the pace of change.

When ChatGPT entered the scene about 18 months ago, it had the capacity to digest a mere 2,000 to 4,000 tokens. In layman terms, that is about a 700-word prompt. Recently, a model boasted the ability to process inputs of 10 million tokens, or equivalent to 750 full-length novels. We are in a realm that makes Moore’s Law appear tame.

The pace of change is relentless. This is creating new waves in scientific investigations of every type. As hundreds of articles compiled at geninnov.ai show, GenAI is now an active participant in fields as far away from each other as drug discovery and astrophysics.

If some machine-generated molecules are currently being tested in phase-2 trials, other models are calculating astrophysics variables with a degree of precision unattained by humans so far.

In fact, in recent weeks, a company has announced an AI scientist too!

GenAI era will deliver extreme personalization in every field. The obvious first field is personalized medicine, where new models are identifying novel patterns and correlations in genomic data-sets that are critical for understanding complex biological processes.

Personalization is the theme in fields beyond medicine. Education is about to change with personalized learning experiences. So will financial advisory services. And agriculture. And manufacturing. And possibly every other conceivable domain.

Also read: It’s swallowed billions of dollars, but has AI lived up to the hype?

Of course, this comes with risks and is still in its infancy. But, as these models evolve, their role in advancing scientific knowledge will only grow. The point is that the GenAI era is more than flashy features or incremental productivity gains. It’s about laying the groundwork for a transformative shift.

The author is a Singapore-based innovation investor for LC GenInnov Fund.