Every week, Invector Labs publishes a newsletter that covers the most recent developments in AI research and technology. You can find this week’s issue below. You can sign up for it below. Please do so, our guys worked really hard on this:
From the Editor: The Emergence Self-Learning Models
Machine learning literature typically divide the world in supervised and unsupervised models but reality is much more complex. While supervised models rule the current artificial intelligence(AI) solution space, they often result unpractical given their dependency on training data. As a result, machine learning practitioners have been exploring new disciplines such as semi-supervised or omni-supervised learning that can effectively operate in scenarios that lack large labeled datasets. One of those disciplines that have been receiving some attention is called self-learning.
The idea of self-learning dates back to the 1980s but has recently experienced a renaissance in technologies such as the Amazon Alexa assistant. Conceptually, self-learning systems perfect their knowledge by creating feedback loops with the environment. When Alexa tries to improve an answer after you appear frustrated, that’s self-learning in action. Among the tech giants, Amazon has been pushing the boundaries of self-learning models incorporating all sorts of new techniques into Alexa’s brain. Soon we should see more mainstream machine learning applications that can benefit from self-learning models.
Now let’s take a look at the core developments in AI research and technology this week:
Privacy in Textual Data
AI researchers from Amazon published a paper proposing a technique to re-phrase customer specific data in text datasets in order to achieve higher levels of privacy.
Alexa Learns to Correct her Own Defects
Amazon also published a fascinating paper about how they used a technique called self-learning to train the Alexa digital assistant to infer and correct her mistakes.
Is Bias in AI Models Self-Reinforcing?
Microsoft Research published a paper exploring the negative feedback loop of bias in AI models.
Cool AI Tech Releases
Open Source Video Analytics
Microsoft open sourced Project Rocket Platform, to streamline AI-based video analytics.
Google BERT in ONNX
Staying with Microsoft, the tech giant open sourced an ONNX-based version of Google’s famous BERT model optimized for high performance on NVIDIA GPUs.
GitHub Intelligent Repositories
The Github engineering team published a great blog post detailing how they used AI to build its “good first issues” feature, which matches contributors with issues that are likely to fit their interests.
AI in the Real World
AI Funding Record
Venture funding for AI startups hit a record $26 billion in 2019.
Facebook has trained an AI model to learn navigate an unfamiliar environment without needing a map.
Google CEO Wants AI to be Regulated
Sundar Pichai, CEO of Google and Alphabet, published an op-ed in the Financial Times making a case for AI regulation.