AI practitioners can draw lessons from scientists working with animals to train reinforcement learning agents, a recent paper found.
Researchers at MIT’s CSAIL developed an AI and machine learning system that can decipher ‘lost’ languages it hasn’t seen before.
Microsoft’s T-URLv2 model can translate between 94 different langauges, falling just short of Facebook’s recently announced M2M-100 model.
Researchers at DeepMind have open-sourced the FermiNet, a neural network that simulates the positions and arrangements of electrons.
LaND is a navigation system that learns from disengagements to help delivery robots successfully travel longer distances.
Google Assistant now integrates with more mobile apps on Android phones. In addition, it supports macro-like voice shortcuts.
More than 70 companies now count themselves, members of Amazon’s Voice Interoperability Initiative, which recently published a design guide.
Researchers developed a data set called Libra-Adapt to show the challenges in adapting voice recognition models to new devices and settings.
Nvidia’s DLSS neural upscaling technology now works with virtual reality titles. This could enable some VR games to make a visual leap.