We need to develop simpler models, which can drastically reduce the carbon footprint of AI. Current research estimates that training a large deep-learning model produces 626,000 pounds of carbon dioxide, equal to the lifetime emissions of five cars. But to do all that, AI systems must consume enormous amounts of energy. Additionally, we will understand more clearly the carbon footprint of machine learning and get more serious about Sustainable AI.ĪI innovations can help optimize many of our activities to slow the impacts of warming: optimizing the electricity cost of technology, making transportation more efficient, monitoring and stopping deforestation, preserving biodiversity or ensuring there’s enough food to go around. New technical ideas and funding to back them are needed to ensure progress.
It is important to remember that today’s greatest advances are due to decades-old ideas enhanced by vast amounts of data and computation. We will also see new ideas for AI, get more serious about AI and privacy, and address AI sustainability. While the industrial robots of the past 60 years are made of hard plastics and metal, I believe the next 60 years will bring us machines made of materials available to us naturally, or through engineered processes like wood, plastics, paper, ice or even food. While the industrial robots of the past 60 years have mostly been inspired by the human form, the next stage will be soft robots inspired by the animal kingdom: form and diversity modeled by our own built environment, with broad potential to mimic our natural state. While the past 60 years have defined the field of industrial robots and empowered hard-bodied machines to execute complex assembly tasks in constrained industrial settings, the next sixty years will usher in robots in human-centric environments to assist humans with physical tasks. What will 2022 bring for these categories? As we move forward into our rapidly changing world, I believe that robots and AI can help us unlock our human potential, as individuals and as a collective. The research community is working to address these challenges. There are also robustness problems as the trained models are often unstable, and we need to understand that the systems do not do “deep reasoning”, they mostly perform shallow pattern matching. Furthermore, these systems are black boxes - there is no way for users of the systems to truly “learn” anything based on the AI’s innerworkings. The quality of that data needs to be very high, and if the data is biased or bad, the performance of the systems trained on this data will be equally bad. But it has also become clear that these methods require data availability, meaning massive data sets that have to be manually labeled and are not easily obtained in every field. Industry has adopted many deep neural network applications that are enabling tools to augment work in a variety of fields. In AI, we have seen heightened awareness about the challenges with today’s AI solutions. Research has made great strides on safer and more capable robots, with advancements in soft robot bodies and machine-learning powered robot brains. Industry has seen increased adoption of robots for manufacturing and logistics applications, where autonomy can deliver value, yet autonomy on the roadways in the form of a robotaxi is still a long way away. What was the defining robotics/AI/automation trend of 2021? Regarding robotics and automation, the pandemic and subsequent labor shortages made it abundantly clear that there is a critical role for robots in the workplace. She’s really gone above and beyond in her responses for this last Actuator of 2021, so I’m going to let her kick things off. This week we’ve asked MIT CSAIL Director Daniela Rus to weigh in on the matter. Image Credits: Paul Marotta/Getty Images for TechCrunch
0 kommentar(er)
