AI year in review

Jerome Pesenti

At Facebook, we think that artificial intelligence that learns in new, more efficient ways – much like humans do – can play an important role in bringing people together. That core belief helps dri…

Overview

The article reviews Facebook's advancements in artificial intelligence (AI) throughout 2018, highlighting the importance of semi-supervised and unsupervised learning, the transition of AI research to production, and the societal benefits of AI applications. Key developments include the release of PyTorch 1.0, improvements in language translation, and AI tools aimed at enhancing lives.

What You'll Learn

1

How to implement semi-supervised learning techniques to enhance AI model performance

2

Why transitioning AI research to production is critical for real-world applications

3

How to leverage PyTorch 1.0 for developing production-ready AI applications

4

When to apply AI tools for social good, such as suicide prevention and disaster relief

Key Questions Answered

What advancements did Facebook make in AI during 2018?

In 2018, Facebook made significant strides in AI by focusing on semi-supervised and unsupervised learning, releasing PyTorch 1.0, and developing tools for applications like suicide prevention and disaster relief. These advancements aim to enhance AI's versatility and societal impact.

How does Facebook's PyTorch 1.0 improve AI development?

PyTorch 1.0 integrates modular, production-oriented capabilities, streamlining the AI development pipeline from prototyping to deployment. It includes features like a hybrid front end for execution modes and revamped distributed training, making it easier for developers to transition AI research into production.

What is the significance of semi-supervised learning in AI?

Semi-supervised learning reduces the reliance on large labeled datasets, allowing AI systems to learn from less curated data. This approach has shown substantial improvements in tasks like language translation, enabling the translation of low-resource languages.

What AI applications did Facebook develop for social good?

Facebook developed AI applications aimed at social good, such as tools for suicide prevention that analyze posts for suicidal language and AI systems that assist in disaster relief by quickly identifying affected areas. These initiatives demonstrate AI's potential to improve lives.

Key Statistics & Figures

Languages added to automatic translation services

24

This increase was achieved through multilingual modeling techniques that leverage similarities between dialects.

Public Instagram images used for training

3.5 billion

These images were utilized to generate self-labeled data for image recognition tasks.

Speed increase for MRI scans

up to 10x

This improvement is part of the fastMRI collaboration with NYU School of Medicine, aimed at making MRI technology more accessible.

Technologies & Tools

Some links below are affiliate links. We may earn a commission if you make a purchase.

Framework

Pytorch

Used for developing and deploying AI applications, with a focus on production readiness.

Platform

Horizon

An open source platform for applying reinforcement learning in production environments.

Framework

Glow

Enables hardware acceleration for machine learning across various platforms.

Key Actionable Insights

1
Implementing semi-supervised learning techniques can significantly enhance the performance of AI models, especially in scenarios where labeled data is scarce.
By utilizing semi-supervised methods, developers can expand the capabilities of their AI systems to understand and translate more languages, including those with limited resources.

2
Transitioning AI research into production is essential for maximizing the impact of AI technologies in real-world applications.
This transition allows developers to leverage advancements in frameworks like PyTorch 1.0, which facilitate the deployment of AI solutions across various platforms.

3
Utilizing AI tools for social good can create significant societal benefits, such as improving mental health support and enhancing disaster response efforts.
By integrating AI into applications aimed at helping vulnerable populations, organizations can make a meaningful difference in communities.

Common Pitfalls

1

One common pitfall in AI development is the over-reliance on supervised learning, which can limit the versatility of AI systems.

This occurs when developers focus solely on labeled datasets, neglecting the potential of semi-supervised and unsupervised learning techniques that can enhance model performance.

Related Concepts

Semi-supervised Learning Techniques

AI Applications For Social Good

Transitioning AI From Research To Production