Transformers Comprise the Fourth Pillar of Deep Learning

 By ARK Invest - 2021-03-05

Description

Transformers increases our confidence that AI will contribute $30T to global equity market capitalization over the next 20 years.

Summary

  • by James Wang, AnalystMarch 05, 2021 Modern artificial intelligence (AI) is powered predominantly by three categories of neural networks, each with distinct capabilities focused on different markets.
  • For informational purposes only and should not be considered investment advice, or a recommendation to buy, sell or hold any particular security.
  • While AI researchers needed five years to match human performance in the ImageNet Large Scale Visual Recognition Challenge, they took little more than a year to develop deep learning models that achieved human performance in the General Language Understanding Evaluation (GLUE) benchmark, as shown above.
  • A new kind of workload, Transformer-based language models are roughly 10x more demanding computationally than image models, as shown above.

 

Topics

  1. NLP (0.32)
  2. Machine_Learning (0.31)
  3. Stock (0.15)

Similar Articles

New deep learning models: Fewer neurons, more intelligence

By techxplore - 2020-10-14

Artificial intelligence has arrived in our everyday lives—from search engines to self-driving cars. This has to do with the enormous computing power that has become available in recent years. But new ...

Top 10 AI and machine learning stories of 2020

By Healthcare IT News - 2020-12-14

Whether assessing vaccine safety and efficacy, assisting with X-ray readings or tracking communities' vulnerability to COVID-19, artificial intelligence has been put to work in new and innovative ways ...

louisfb01/Best_AI_paper_2020

By GitHub - 2020-12-26

A curated list of the latest breakthroughs in AI by release date with a clear video explanation, link to a more in-depth article, and code - louisfb01/Best_AI_paper_2020