Director at Facebook AI Research, Yann LeCun presented a paper detailing the latest trends and the future of Deep Learning hardware, at the IEEE’s 2019 International Solid-State Circuits Conference (ISSCC) in San Francisco. LeCun’s presentation illustrated that several Artificial Intelligence-based businesses should consider this technology in the near future.
The presented document highlighted some subject matters; included machines should be given some common sense- Through the Deep Learning advancements, computer understanding of images, texts, and audio has enhanced, which has enabled developers to develop new applications like information exploration and filtering, self-driving, real-time language translation, and virtual assistants. Though, advancements in these applications heavily rely on supervised learning that necessitates human-annotated data or reinforcement learning. LeCun deemed that researchers, in the next decades, should put their efforts into creating machines that learn like a human, by simple observations and occasional actions or in short, by self-supervised manner. In his paper “Deep Learning Hardware: Past, Present, and Future”, LeCun highlighted that of self-supervised learning enables the machine to learn huge amounts of background knowledge regarding the world’s works through observation, it may hypothesize that some form of machine common sense could emerge.
The document further highlighted that Machine learning chips that can fit everyday devices, where LeCun is optimistic that will see the computer, near future, chips that can fit in everyday devices like vacuum cleaners and lawnmowers. Through the integration of Machine Learning chip, any device can be able to make smart decisions, like facial recognition that recognizing a user’s face to unlock the device. In the near future, more work will be put in to make mobile computing chips more sophisticated. The ISSCC 2019 Conference is the foremost global forum, where researchers present the current advances in solid-state circuits and systems-on-a-chip.