TheSequence Scope: The Biggest Roadblock for the Mainstream Adoption of Machine Learning

Initially published as 'This Week in AI', a newsletter about the AI news, and the latest developments in ML research and technology

Apr 12, 2020

From the Editor: The Biggest Roadblock for the Mainstream Adoption of Machine Learning

One word: data. Up to this day, we haven’t seen the complete potential of unsupervised models and supervised methods rule the machine learning space. Supervised training requires high quality labeled datasets and those are incredibly expensive to produce on an ongoing basis. These challenges have prevented even large organizations from adopting machine learning at scale and are a major roadblock for startups entering the space.

The labeled data hurtle has two-man solutions: Either we develop methods for producing labeled datasets more effectively or we develop methods that can learn with smaller datasets. In the first solution, areas such as generative models are showing some promise to generate high-quality training datasets. In terms of the latter, we have seen incredible advances in research in areas such as semi-supervised models, one-shot or zero-shot learning are improving very rapidly. While the adoption of these techniques remains in a very nascent stage, the problem that they are trying to solve remains the biggest challenge for the adoption of modern machine learning solutions.

Now let’s take a look at the core developments in AI research and technology this week.

AI Research:

Understanding Glass Using Neural Networks

The image credit: DeepMind.com

DeepMind published a very intriguing paper proposing a method based on graph neural networks to understand the puzzling phenomenon of the glass transition.

Advancing Generative Networks

Microsoft Research published an insightful summary of three different projects in generative models that can facilitate the implementation of self-supervised learning methods.

Learning Tasks from Single Examples

Researchers from Amazon published a paper proposing a method that improves performance in meta-learning tasks without increasing the training data requirements.

Cool AI Tech Releases:

TensorFlow QAT

The TensorFlow team released the Quantization Aware Training (QAT) API which allows the implementation of faster and smaller machine learning models.

Sound Separation Dataset

Google open-sourced the Free Universal Sound Separation data set, intended to support the development of AI models that can separate distinct sounds from recording mixes.

AI in the Real World:

$30M for AI Research

The US Department of Energy announced $30 million in funding for AI research projects.

Node AI

AI startup Node raised $6 million to help companies launch AI projects without requiring data science expertise.

AI to Study the Oceans

We know very little about our oceans. AI methods are helping researchers to overcome some of the traditional challenges analyzing the super large datasets related to oceanic data.

“This Week in AI” is a newsletter curated by industry insiders and the Invector Labs team, every week it brings you the latest developments in AI research and technology.

From July 14th the newsletter will change its name and format to develop ideas of systematic AI education.

To stay up-to-date and know more about TheSequence, please consider to ➡️

TheSequence