NVIDIA’s GTC event this year not only featured announcements from NVIDIA themselves but also showcased a plethora of leading researchers in various subfields. One notable attendee this time around was Yann LeCun, Chief AI Scientist at Meta, a pioneer in convolutional neural networks (CNN) who painted a picture of the future of artificial intelligence.
LeCun envisions AI in the future to expand beyond the current standard of LLM, as the output format is limited (according to token dictionaries) and training solely on text data using model expansion and continuous training data may not lead to the creation of human-level intelligence or AGI. Humans learn from a diverse range of data, including vast amounts of images and sounds from childhood.
He also expressed his dislike for the term AGI, as even humans are considered to possess unique intelligence. He prefers to refer to highly intelligent AI as AMI, or advanced machine intelligence. Furthermore, the process of developing AMI/AGI is not a sudden breakthrough by any single company but rather a continuous evolution. He concluded by emphasizing that creating AMI/AGI requires more affordable hardware in the long run.
Source: NVIDIA GTC event
TLDR: Yann LeCun, Chief AI Scientist at Meta, discusses the future of AI and believes that creating advanced machine intelligence (AMI) or artificial general intelligence (AGI) requires a more diverse training approach and continuous development process rather than a sudden breakthrough.
Leave a Comment