To understand how we came to be immersed in a world that constantly talks about artificial intelligence, It’s helpful to analyze and reflect on the fact that this is not entirely new, but rather something that the scientific community has been working for many years to theorize and attempt to build machines that resemble us or at least simulate our human capabilities.
It can be inferred with certainty that the first venture into AI was the Turing machine. From there, during the 1950s and 1960s, the theoretical foundations of AI were established. These theories were based on how machines could simulate human intelligence.
By the late 1960s and throughout the 1970s, symbolic approaches and logic- based rules were primarily used, with IBM's Deep Blue chess-playing program at an expert level being an iconic example of the era.
The symbolic connection approach emerged in the 1980s and was extended into the 1990s, a period during which techniques were explored that combined symbolic reasoning with methods based on knowledge and learning from the connections between data.
In the early 2000s, there was a resurgence of research into techniques that had begun in the previous decade but had not proven sufficiently fruitful: neural networks. This period stood out for significant advances in learning algorithms, such as supervised and unsupervised learning, as well as techniques in natural language processing and computer vision.
In the 2010s, with the massive increase in data availability, computational power, and communications, the focus on deep learning began to dominate the AI field. Image recognition, autonomous driving, and natural language processing are examples of techniques that reached levels that Turing and McCarthy could barely have dreamed of decades earlier.
Modern artificial intelligence aims to create intelligent agents that assist individuals and organizations in performing their tasks with outstanding levels of efficiency and effectiveness, generating a high return on technological investment and providing a competitive advantage.
Among the forces driving the rapid evolution of AI, we can highlight some of the most influential ones:
- The development of Big Data
- The boom of E-commerce
- The Internet of Things (IoT)
- Social media
- The growth in computational power
- The drastic reduction in processing costs
These forces provide a significant boost to technology investments and, particularly in the case of AI, carry expectations of building systems capable of accelerating business processes. To make this possible, the goal is to equip systems with capabilities that allow them to:
- Perceive their environment
- Represent it
- Learn
- Make autonomous decisions
The representation of the environment is what we call a MODEL.
Model tuning is the learning process, in which metrics are adjusted by interacting with a simulator or through repeated exposure to new information, and from that learning, making autonomous decisions. This process aims to understand the distribution of input data and, from there, enable learning.
Pero no todo es fantástico, también hay desafíos para mantener a la IA alejada de nuestras propias limitaciones. En este sentido, uno de los mayores desafíos que hoy presenta la IA es resolver la influencia de los sesgos. Al no tener programación específica, los agentes de IA adquieren los sesgos de los datos. Por ejemplo, puede darse que un auto autónomo no reconozca a un peatón que cruza en la mitad de la calle, porque fue entrenado para “entender” que los peatones cruzan en las esquinas. Esto trae los problemas obvios que los modelos deberán solucionar para adaptarse a tareas críticas, y brindar decisiones objetivas tanto operativas como estratégicas.
Currently, the trend in AI adoption indicates that within the next 24 months, most organizations plan to integrate custom solutions based on this technology. Statistically speaking, companies are seeing a return of $3.70 for every dollar invested in AI, making it a highly attractive area for technology investment.
However, the main barrier to AI adoption is the lack of qualified personnel. This challenge is one that organizations and technology companies face—finding individuals who can identify needs, design solutions, and make the necessary adjustments to extract maximum value.