Atlas, the humanoid robot famous for its parkour and dance routines, has recently begun demonstrating something altogether more subtle but also a lot more significant: It has learned to both walk and grab things using a single artificial intelligence model. What is more, the robot's single learning model is showing some tantalizingly "emergent" skills, like the ability to instinctively recover when it drops an item without having been trained to do so. Boston Dynamics, the company that makes Atlas, together with the Toyota Research Institute (TRI), developed a generalist model that learns to control both arms and legs from a range of example actions. This is different from the norm: robots equipped with the ability to learn would usually rely on one model to walk and jump and another to grasp items. "The feet are just like additional hands, in some sense, to the model," says Russ Tedrake, a roboticist at the Toyota Research Institute and the Massachusetts Institute of Technology, who led the current work.

Artificial intelligence (AI) must be inclusive to reach its potential. AI applications that solve problems for a small segment of the population will fail to achieve widespread adoption. So, it's important that AI applications be designed and prepared with data that reflects as many segments of the global population as possible. Many moving parts need to be managed well to do that, and one of them is language. The more languages an AI application can handle, the more inclusive it is.

Incorporating AI and machine learning technologies into everyday workflows isn't as easy as the testimonials would have you believe. That's the top-level finding from a survey of 750 business decision makers conducted by Algorithmia, which found that while machine learning maturity in the enterprise is generally increasing, the majority of companies (50%) spend between 8 and 90 days deploying a single machine learning model (with 18% taking longer than 90 days). Most peg the blame on failure to scale (33%), followed by model reproducibility challenges (32%) and lack of executive buy-in (26%). "The findings of our 2020 [State of Enterprise Machine Learning] study are consistent with what we're hearing from customers," said Algorithmia CEO Diego Oppenheimer. "Companies are growing their investments in machine learning, and machine learning operationalization is maturing across all industries, but significant room for growth and improvement remains. The model deployment lifecycle needs to continue to be more efficient and seamless for ML teams. Nevertheless, companies with established ML deployment lifecycles are benefiting from measurable results, including cost reductions, fraud detection, and customer satisfaction. We expect these trends to continue as ML technologies and processes arrive to market and are adopted."

Incorporating AI and machine learning technologies into everyday workflows isn't as easy as the testimonials would have you believe. That's the top-level finding from a survey of 750 business decision makers conducted by Algorithmia, which found that while machine learning maturity in the enterprise is generally increasing, the majority of companies (50%) spend between 8 and 90 days deploying a single machine learning model (with 18% taking longer than 90 days). Most peg the blame on failure to scale (33%), followed by model reproducibility challenges (32%) and lack of executive buy-in (26%). "The findings of our 2020 [State of Enterprise Machine Learning] study are consistent with what we're hearing from customers," said Algorithmia CEO Diego Oppenheimer. "Companies are growing their investments in machine learning, and machine learning operationalization is maturing across all industries, but significant room for growth and improvement remains. The model deployment lifecycle needs to continue to be more efficient and seamless for ML teams. Nevertheless, companies with established ML deployment lifecycles are benefiting from measurable results, including cost reductions, fraud detection, and customer satisfaction. We expect these trends to continue as ML technologies and processes arrive to market and are adopted."

They found that the computational and environmental costs of training grew proportionally to model size and then exploded when additional tuning steps were used to increase the model's final accuracy. In particular, they found that a tuning process known as neural architecture search, which tries to optimize a model by incrementally tweaking a neural network's design through exhaustive trial and error, had extraordinarily high associated costs for little performance benefit. Without it, the most costly model, BERT, had a carbon footprint of roughly 1,400 pounds of carbon dioxide equivalent, close to a round-trip trans-American flight.