TheSequence Scope: Architectures for Building AI at Scale

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

Jul 05, 2020

From the Editor: Architectures for Building AI at Scale

Building artificial intelligence (AI) systems at scale seem like a problem that only a few companies have and, as a result, there is not much available guidance about what works and what doesn’t. If you are a machine learning architect looking to implement an infrastructure to run ML programs at scale, where do you start? As a new industry, we are all learning and trying to figure out the best practices as we go along.

In our experience, the best reference architectures for implementing machine learning at scale are coming from big technology players like Microsoft, Google, Facebook, Uber, Amazon, etc. After all, they are the companies dealing with these challenges. The great thing is that these companies have been open-sourcing many of the frameworks and tools of their infrastructure. Just this week, Uber open-sourced its Fiber framework for running parallel, highly scale computations.

To follow the new releases of those ML tools and frameworks is a great way to spot new best practices and ideas that could be adapted to your machine learning architecture.

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

AI Research:

Adversarial Neural Networks at Amazon

The Amazon Research team published some details about how the retail giant uses adversarial neural networks to improve product discovery.

Training Robots to Navigate Using Simulations

Microsoft Research published a paper and a dataset about how to use photorealistic simulated environments to teach robots how to see and navigate

The image credit: Microsoft

Translating and Summarizing Documents Without Fine-Tuning

Facebook AI Research published a paper introducing MARGE, a pre-trained language model that learns to paraphrase, translate, and summarize text without any fine-tuning.

Cool AI Tech Releases:

AWS CodeGuru

Amazon Web Services (AWS) announced CodeGuru, a machine learning engine that can automate code reviews.

>Read more in the AWS press release

Uber Fiber

Uber open sourced Fiber, a framework for scaling computations, like the ones used in machine learning models, to hundreds of thousands of machines

Connected Sheets

Google announced the general availability of Connected Sheets, which enables the analysis of petabytes of data using the familiar Google Sheets interface.

AI in the Real World:

Amazon, Google Back Nation Research Cloud Bill

Tech giants Amazon and Google have joined over 20 organizations supporting a US Congress bill to create a National AI Research Cloud that will give AI researchers access to compute resources and datasets.

YouTube Gets a Smart Reply

Google has adapted its famous SmartReply feature to YouTube.

MIT Takes Down Popular Dataset

MIT has taken offline its high cited Tiny Images dataset over allegations that it contains racist and biased labels that can influence the outcome of machine learning models.

‘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