Every week, my team at Invector Labs publishes a newsletter to track 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 AI that can Improvise in Jazz
Improvisation has been the hallmark of jazz since the bebop revolution. Improvisation doesn’t only requires an incredibly musical talent but also the ability to create compositions in real time. Even though AI has made a lot of progress in the music space, most of it has been done in non-real time environments in which the models undergo an extensive training. Improvisation has long been seen as one human artistic, cognitive skills that would be impossible for AI to tackle.
If we think about musical improvisation in the context of AI, the challenge results overwhelming. First of all, models will need to process musical information real time generate from different instruments and generate something that doesn’t only fit well with the melodies but that is also unique and that it aligns well with the band chemistry. Recently, a team from Google unveiled ML-Jam, a machine learning model that is able to improvise real time using different instruments. The experiment has been so successful that the new system is being used as part of live performances.
Now let’s take a look at the core developments in AI research and technology this week:
In an impressive display of machine learning and user experience, AI researchers from the Massachusetts Institute of Technology(MIT) unveiled an interactice data science system called Northstar.
Google AI researchers published a paper about new improvements in graph representation learning, a technique that can streamline many deep learning models.
Microsoft Research published a paper introducing a pre-training method that vastly outperformed similar state-of-the-art methods in language generation tasks.
Cool AI Tech Releases
Facebook released a new took for searching code snippets using natural language.
Google open sourced YouTub-8M Segments, a new dataset to train video analysis models.
Microsoft open sourced TensorWatch, a new tool for debugging deep learning models using Jupyter Notebooks.
MLPerf, a consortium of 40 organizations like Alibaba, Google, Facebook or Google released a new dataset for interference models.
AI in the Real World
AI pioneer Stuart Russell, was the guest in a Harvard Business Review podcast to discuss the future of AI and its usage to improve mankind.
Google unveiled ML-Jam, a machine learning system that can improvise melodies on the fly.
A team of AI researchers use simulations to understand how the universe work and the results are enlightening.