The AI of the future probably won't need large data centers, according to experts

Source: CritpoTendencia Original Title: The AI of the future probably won’t need large data centers, experts say Original Link: Currently, the advancement of artificial intelligence (AI) is shaping a promising future, but also one filled with challenges. Its progress is largely driven by a strategic race among major powers seeking dominance in this technological competition. However, gaining a decisive advantage is becoming increasingly complex due to the scarcity of critical resources.

Launching an advanced AI program is neither an easy nor inexpensive task. It requires colossal investments to build the necessary infrastructure, such as data centers and power generation systems capable of supporting them. These are today the foundations for training advanced language models, which depend on enormous volumes of data and computational capacity.

However, this approach could represent just the prehistory of artificial intelligence development. According to a recent study from Johns Hopkins University, there are much more efficient alternatives to advance this technology. The research suggests that emulating the functioning of the human brain would allow AI to learn without the need for a constant avalanche of data.

In simple terms, humans do not need to be connected to a data center via wiring to acquire knowledge. They learn from direct experience and interaction with their environment. An AI system that autonomously explores the network could be a key to the future.

In this way, an artificial intelligence capable of learning without depending on an increasingly powerful computing system could emerge.

Will AI become a superbrain in the future?

For now, there is no complete understanding of how the human brain works. Neuroscience is advancing rapidly, but it is still far from recreating a system capable of generating knowledge from stimuli received through electromagnetic pulses autonomously and consciously.

Neurons receive information in their dendrites from the axons of other neurons via synapses. This process occurs billions of times in the brain and is precisely what gives rise to memories, knowledge, skills, and other cognitive processes. Replicating this mechanism has not yet been possible, although it is likely that, with scientific progress, it will become a field conquered by future generations.

Once the complex network that allows matter to transform into perception and then into abstract thought is understood, science will be able to replicate the brain organ, or at least something functionally similar. Thus, a non-biological organism with characteristics analogous to the brain could access information without relying on gigantic data centers.

This is precisely what the mentioned study points to. “The way the AI field is moving right now involves throwing enormous amounts of data at models and building computational resources the size of small cities. This requires hundreds of billions of dollars,” says the study’s lead author, Mick Bonner.

He then notes that humans, in contrast, are capable of generating knowledge from a relatively small amount of information and data.

###Learning from evolution

According to the study, the way the human brain generates knowledge demonstrates that it is a highly efficient process. “Evolution probably converged on this point for a good reason,” the author states.

“Our work suggests that architectural designs most similar to the human brain place AI systems at a much more advantageous starting position,” he emphasizes.

The study also raises a huge ethical dilemma. The human brain is not a supernatural organ, nor does knowledge come from divine sources. It is the result of an extremely complex organization of matter that, through evolution, reached a level of development capable of generating self-awareness.

If humanity manages to fully unravel how this organ works and succeeds in replicating it, the result would be clear: the creation of artificial consciousness. This would imply the existence of an entity capable of experiencing emotions such as fear, anger, or distress.

Although imagining this scenario inevitably recalls episodes of dystopian science fiction, the Johns Hopkins University study suggests that science is moving in that direction. Will the development of AI open a new Pandora’s box of ethics?