Expert's Perspective

The head of the Meta AI department believes that artificial intelligence is closer to the intelligence of cats and dogs rather than humans. According to Chief Meta Scientist Yann LeCun, current AI systems are far from being truly intelligent.

Contrasting Views

Yann LeCun's viewpoint contradicts that of Nvidia CEO Jensen Huang, who believes that AI will be competitive with humans in less than five years. LeCun suggests that Huang's perspective may be influenced by his vested interest in the spread of AI.

AI and Processor Providers

LeCun explains that the development of AI is closely tied to the number of processors purchased from providers like Nvidia. As companies like OpenAI continue their pursuit of artificial general intelligence, they will require more computer processors.

Levels of AI

LeCun predicts that humanity will achieve cat-level or dog-level AI before reaching human-level AI. This means that AI systems with the intelligence of cats and dogs may become a reality sooner than systems with human-like intelligence.

Limitations of Language Models

LeCun highlights that the current focus on language models and textual data is insufficient to create advanced AI systems. Text alone is a weak source of information, and training models on vast amounts of text does not guarantee a deep understanding of the world.

Transformers and Hidden Correlations

MetaAI executives have conducted extensive research on adapter models and transformers. Transformers help AI systems discover hidden correlations between different types of data, such as audio, image, and video information.

Multimedia AI Systems

The development of multimedia AI systems represents the next level in the field of artificial intelligence. These systems require a combination of 3D visual data, text, and audio. However, their development is costly and could give Nvidia an advantage if no other competitor emerges.

Source: CNBC