published on 21 February
AI is a technology, often based on algorithms, that enables machines to learn and make decisions, in order to perform tasks by imitating human intelligence. The tasks to be performed are often predefined, with the primary goal being to train the AI to perform a specific task as effectively as possible.
Indeed, as is the case with autonomous vehicles, AI must be “trained” by absorbing a large amount of data to learn and be capable of executing the requested task; for example, distinguishing between a pedestrian and a cyclist.
The great strength of AI is its ability to process a monumental amount of data in a short time, which gives it remarkable predictive capability. In construction projects, for example, it can enable time savings, resource and budgetary savings, or even enhance the safety of site workers.
Although Alan Turing is often considered the “father” of modern computing and one of the precursors of Artificial Intelligence, the origins of its appearance are a bit more complex, and the idea of a “thinking machine” has been present in the human mind since antiquity. It is the research and publications of numerous scientists over decades that have led to the emergence of AI as we know it today.
Warren McCullough and Walter Pitts wrote the article “A Logical Calculus of Ideas Immanent in Nervous Activity,” in which they presented a mathematical model of neurons for machines, aiming to replicate the functioning of the human brain.
Alan Turing, already known for the success of his machine that decrypted Nazi-coded messages, published “Computing Machinery and Intelligence,” in which he attempted to answer the question “Can machines think?” He conceptualized a test bearing his name, supposed to determine if a computer can imitate or achieve the same results as human intelligence.
John McCarthy first used the term “Artificial Intelligence” at the “Dartmouth Summer Research Project on Artificial Intelligence” conference at Dartmouth College. Later that year, Herbert Simon and Allen Newell created the first computer program with AI, called the “Logic theorist.” A few years later, along with Marvin Minsky, they founded the Massachusetts Institute of Technology.
Frank Rosenblatt, a psychologist, designed the Perceptron, a mathematical model inspired by the human neural network with a binary output system.
Marvin Minsky and Seymour Papert published a book titled “Perceptrons,” which became a reference on neural networks.
AI experienced a period known as the “AI winter.” Due to high expectations and uncertain results, researchers gradually lost their investments and public support.
The emergence of more sophisticated algorythms and the discovery of backpropagation algorythms marked a revival for machine learning.
Deep Blue, the supercomputer developed by IBM, became the first machine to defeat a human, the reigning world chess champion at the time, Garry Kasparov. This event marked a major turning point in the history of AI, which has continued to develop and diversify to this day, promising a multitude of possibilities.
Artificial Intelligence is often categorized into two main types.
1. Narrow AI (ANI, Artificial Narrow Intelligence):
➡️ This is the AI that currently exists. It relies on algorithms for task automation. The learning capability of this form of AI enables tasks such as decision-making for autonomous vehicles and virtual assistants (e.g., Amazon’s Alexa). This form of AI also includes “generative” applications like ChatGPT (text) or Midjourney and Dall-E (images).
2. Strong AI, composed of General AI (AGI) and Artificial Superintelligence (ASI):
➡️For now, at the theoretical stage, it involves imagining a machine capable of thinking equivalent to humans. It would possess its own consciousness and be capable of emancipating itself from the data initially integrated into it. Strong AI would also be able to fully mimic human reasoning: planning for the future, problem-solving, questioning, or even expressing an opinion.
AI operates through algorithms, which are based on specific techniques such as Machine Learning and Deep Learning, a subcategory of Machine Learning.
In construction, AI can intervene at various stages of a project to optimize planning, costs, resource management, and safety:
AI is present in some solutions, such as Scodify, our solution for automating the transformation of data from CAD plans into 100% compatible GIS data.
Indeed, thanks to Machine Learning, Scodify can interpret and automatically recognize elements of a plan (pipelines, networks, etc.) and standardize them to generate a single plan. The strength of this technology is that it learns and strengthens its knowledge with each new plan integrated into the application, regardless of its graphical charter.
Scodify not only uses Machine Learning but also has AI capable of linking text to the correct element of a plan or transforming the drawing of a network into a topological graph.
Sogelink has recently established an “Innovation Center,” an internal incubator with its own team dedicated to researching various topics, including AI, with the aim of offering solutions in line with new technologies while addressing business challenges.
Some of our solutions stem from our Innovation Center and examples of its success:
