Last month, I was invited by the #Mex-IA community to talk about the evolution of Artificial Intelligence from its early stages to its foreseeable future.

In short, it was a blast! 

It had been a while since I had so much joy creating a presentation. I talked about Good Old-Fashioned AI (GOFAI) to the Agentic and Causal AI passing through Machine Learning, Deep Learning, and Generative AI, explaining along the way the different inference approaches on each case: logical, heuristic, mathematical and neural-network based.

The topic that drew the most attention and discussion was what I called the convergence of the Artificial Narrow Intelligence (ANI) into Artificial General Intelligence (AGI). 

For this topic I created a couple of slides inspired by the now-famous AI maturity framework introduced by Google’s DeepMind team a year ago.

I used this chart to highlight in yellow a very significant change since the summer of 2024, when this document was first published, which is AGI’s achievement of “Competent Level” thanks to the introduction of the “reasoning” (time-to-compute) feature on the most advanced LLM models.

I then used this second chart to explain how communication protocols, such as MCP, will enable the interaction of multiple ANI agents to make them behave as a distributed AGI tool that would inevitably evolve into an Artificial Super Intelligence (ASI).

Besides revealing my view on achieving ASI through the orchestration of multiple ANI agents, I also shared an example of the kind of prompts we will be able to do within the next five years. 

Part of this means that as a CEO, I would be able to ask for a five-year strategic plan to achieve business goals set by a Board of Directors. This would include the likely market scenarios during this timeframe and the current financial and operational circumstances of the business which these tools will be able to determine by analyzing the current and historical data available on transactional systems such as ERPs and legacy systems. These well-choreographed ANI agents would be able to formulate a “competent” if not “expert” plan by effectively combining all sorts of intelligence types: recommendation engines, causal AI, deep and machine learning, and of course, GOFAI =).

In my view, AI would inevitably evolve into Artificial Super Intelligence (ASI) but not by the means of a single technology platform but rather as the result of a careful orchestration of multiple ANI agents that will in turn utilize all sorts of inference types: logical, heuristic, mathematical and neural-network based.

If you’re interested in this presentation or want to know more, email me directly at heriberto.perez@wizeline.com