In case there was any doubt about the scarcity of collective creative resources, humanity continues to turn to yesterday to try to solve its concerns for tomorrow.
With the retro aroma of past essences, we star in a reissue of the space race and the Cold War to see which of the three players reaches the top of the LLM first: Open AI, Google or Anthropic.
While this is happening—and with the anxiety and novelty that was experienced in the eighties with the dotcom boom—dozens of platform innovations pass before our eyes, ranging from generating podcasts from a document to uploading your identity online to have conversations with yourself, in a more intelligent version. Will the same thing happen with AI, as with websites and apps, over the years?
And the agents’ agenda?
It is not difficult to sympathize with the human spirit in times when robots appear around every corner. The next function in this room is carried out by the so-called “agents”, systems that are conceived as autonomous and that have the capacity to execute specific tasks with a degree of precision and speed that sends the imagination to probable film scripts.
The big technology companies not only invest, but also name their creations with religious enthusiasm: Agentforce, Agentstudios. The semantics are not coincidental; These tools promise to revolutionize—once again—business management and its entire cascade of effects.
Only these multi-agent systems are not new. Like several developments that were ahead of their time, they have waited until the market was desperate or excited enough to be adopted.
Paradox looks like a joke in this sideboard. In a context that seeks to delegate the maximum number of tasks to machines, intelligent agents require, according to Microsoft’s Ece Kamar, greater human supervision. The illusion of an autonomous office still requires the insomnia of managers who wake up in the middle of the night with existential doubts: can we trust what is not understood about ethics, but what is understood about efficiency?
Once again: generalization vs. specialization
This dilemma leads to the second great earthquake of AI: specialization. General-purpose models—those titans that seemed invincible like OpenAI—now face a new challenge: being relevant thanks to usefulness and proximity.
Those who have bet on the towers of reports and their patterns to anticipate what is coming in 2025, agree that the industry will necessarily turn towards specialized models. If an algorithm can’t speak like a prosecutor or diagnose like an allergist, what good is it in the long term? In the realm of infinite data, specialization is the only possible form of government.
Thus we arrive at 2025, a year that is announced as a watershed. Interconnected networks of agents, models trained in knowledge niches, but also the need to justify each peso invested and try to make it exponential in the shortest possible time.
In reality, the challenge will not be technological, but human: beyond managing expectations, moderating illusions and understanding that even in the age of artificial intelligence, human intelligence remains irreplaceable, knowing direction and purpose above any anxiety and rush will be crucial.
Natalie Glance, Chief Engineering Officer at Duolingo, offers advice that may seem obvious, but in the context of inflated expectations seems revolutionary: “the key is not to reduce costs, but to build value.” Shouldn’t that have been the initial premise?
Contact:
* Eduardo Navarrete is a specialist in Futures Studies, journalist, photographer and Head of Content in UX Marketing.
Linkedin: https://www.linkedin.com/in/eduardo-navarrete
Mail: (email protected)
Instagram: @elnavarrete
Follow information about business and current events in Forbes Mexico
