My team at Invector Labs started a new newsletter to track the most recent developments in AI research and technology. I published the first issue below. You can sign up for it below. Please do so, our guys worked really hard on this:
From the Editor:
Building trust into AI systems is one of those pivotal challenges for the mainstream adoption of AI technologies. But how do we quantify trust? Recently, researchers from IBM published a paper in which they identify four key pillars to trusted AI:
· Fairness: AI systems should use training data and models that are free of bias, to avoid unfair treatment of certain groups.
· Robustness: AI systems should be safe and secure, not vulnerable to tampering or compromising the data they are trained on.
· Explainability: AI systems should provide decisions or suggestions that can be understood by their users and developers.
· Lineage: AI systems should include details of their development, deployment, and maintenance so they can be audited throughout their lifecycle.
As AI systems evolve, IBM proposes that they should include a factsheet that details some of these aspects. If we trust factsheet in food items or appliances why not in AI agents? You can read more about IBM’s ideas in this article.
Now let’s take a look at the core developments in AI research and technology last week:
Facebook announced the winners of their Testing and Verification Awards which looks for proposals that advance the implementation of machine learning systems in the real world
AI agents that know how to ask the right questions? Google released a research paper and open source implementation of a new reinforcement learning method that reformulates a target question in multiple ways in order to obtain the right answer.
Eliminating bias in training data? IBM recently published a research paper that shows how to use adversarial neural networks to generate training datasets with minimum bias
AI Technology Releases:
LinkedIn open sources Tony, a framework for running TensorFlow on Hadoop YARN. We have been able to run TensorFlow on Spark for years but the support for Hadoop has been lacking
Microsoft finally released an open source version of its probabilistic programming framework Infer.net. Probabilistic programming is becoming a very relevant trend in machine learning applications and frameworks like Infer.net help to push that value proposition.
Staying with Microsoft, the Redmond giant just announced the new version of its ML.NET framework that brings machine learning to .NET developers.
AI in the Real World:
Facebook shows how it uses artificial intelligence in its marketplace. The highlights use cases for many of the open source machine learning technologies we have come to love from Facebook.
LinkedIn explains the best practices for scaling machine learning within their organization. Most of these practices are incredibly helpful to organizations starting their machine learning journey.
Cutting edge AI research is expensive. Google’s AI subsidiary DeepMind has been growing its losses year after year.