Engineers and specialists Acquiring deep knowledgeRodolphe publishes Gelin Flammarion Edition The latest news from artificial intelligence. As the title does not indicate, the author devotes himself to a brief historical, but technical, invention of AIs, which he transcends by presenting their limits.
Gary Kasparov and Lee Saddle felt it: In the wake of decades of astonishing technological advances, we’ve learned artificial intelligence … to learn. In the mid-1970s, Dr. Gordon Earle Moore argued that the number of transistors on a microprocessor chip would now double every two years. What is now called “Moore’s Law” follows a series of theoretical, conceptual, scientific, and computer inventions that drove machines to defeat the world’s best chess players (including Deep Blue) and Go players. Turing testing and manufacturing deepfakeSuch as malicious, continuous-stream multimedia strategies.
In her The latest news from artificial intelligence, Rodolphe Gelin goes back to the first few dozen pages in the long history of AI. Aristotle (syllogism, coherence, non-contradiction, exclusion), English mathematician George Bull’s binary logic (prediction of bits and computer systems based on 0’s and 1’s), Claude Shannon’s conversion of analog signals to digital signals. Automated processing, through Boolean algebra), the invention of the Alan Turing software policy or even the communal and theoretical contributions of John McCarthy and Marvin Minsky have all contributed to the formation of artificial intelligence that we know today – and that cinema has never stopped on stage. Terminator, Blade Runner, 2001, A Space Odyssey Or, more recently, HisWhich is also discussed in the book.
Once this origin is brief, the author digs deeper into the typology of AIs and their component principles. With scholarship and a passion for pedagogy, he reminds us that classical computer science is based on mathematics, while artificial intelligence is based on the experience and concept of interacting with each other. And to quote, instead, expert software (which helps in decision making according to pre-encoded diagrams), neural networks (more complex, by level, and according to the principle of self-learning) Machine learning) Or even Generative Artificial Intelligence (The deepfakes For example, based on an opposing generative neural network, two opposite machines learn from each other). This innovation is not only used to identify different types of AI, as it also makes it possible to purposefully address their limitations and their underlying questions.
This is especially true of AI. If an expert software works according to pre-encoded data, a self-learning machine expert may make confusing decisions without being able to interpret them. Rudolf Jillian recalls some memorable failures, including the horrific accident of an autonomous vehicle (because he never noticed the presence of a truck lying on the road) or the classification of a dog as a wolf (because of the snowy background) later learning AI reinforcement. . The author hammers it home: generalizing their knowledge from our artificial intelligence experience. But it can create all kinds of biases (specification problems, learning bias, discriminatory bias, etc.). It also addresses the inferiority of electronic circuits compared to human neurons in matters of morality, proletariat (famous mechanical turkey) under artificial intelligence, or energy consumption.
Short and accessible, but not impure, The latest news from artificial intelligence The current strategy of AI makes it possible to take stock of usage and brakes.
The latest news from artificial intelligenceRudolf Jelin
Flamarion, February 2022, 176 pages
See side by side