Last Week in AI

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 Agent57 Milestone

Games are often seen as a great benchmark to evaluate the ability of artificial intelligence(AI) algorithms to generalize knowledge. From the different data environments that we can create, games come the closest to resemble real world environments. Game environments present challenges for AI models such as exploration-exploitation, operating under uncertainty and many others. However, the fundamental challenge remains for AI models to generalize skills across different games.

Among the game benchmarks in the market, Atari57 has been notorious for its difficulty. The benchmark includes 57 Atari games with tasks that require different skills. Until recently, no single AI agent was able to solve all games. A few days ago, AI powerhouse DeepMind unveiled Agent57, a reinforcement learning agent that surpass human performance across all Atari57 games. Agent57 represents a very important milestone in the deep learning space showing remarkable abilities to generalize knowledge across heterogenous tasks. The techniques behind Agent57 are likely to propel a new generation of reinforcement learning…