News classification
Contact us
- Add: No. 9, North Fourth Ring Road, Haidian District, Beijing. It mainly includes face recognition, living detection, ID card recognition, bank card recognition, business card recognition, license plate recognition, OCR recognition, and intelligent recognition technology.
- Tel: 13146317170 廖经理
- Fax:
- Email: 398017534@qq.com
Driverless cars need at least human common sense
Driverless cars need at least human common sense
Boston's hot drivers and chaotic traffic conditions are well known all over the world, but this could be a perfect place for different types of automated driving tests.
ISee, a derivative of the Massachusetts Institute of Technology, is developing and testing an autopilot system that employs a novel approach to artificial intelligence. The start-up company does not rely on simple rules or machine learning algorithms to train cars, but is inspired by cognitive science, which allows the machine to have a common sense and is able to quickly cope with new situations. It is developing an algorithm that tries to match the way human beings understand and learn the physical world, including interacting with others. This approach may allow auto drivers to better cope with unfamiliar situations and complex traffic conditions.
"Human thinking is very sensitive to physics and social cues," says YibiaoZhao, co-founder of iSee. "Current AI is relatively limited in these areas, and we think this is actually a missing link in driving."
Zhao company has just begun to take shape, is a new investment company created by Massachusetts Institute of Technology, to provide funding for the innovation of local technology companies, a small team of engineers working in a small space in the laboratory. Just a short distance from the campus of Massachusetts Institute of Technology, it overlooks a street where drivers scramble for parking spaces and actively crowd into the traffic.
The laser radar sensor and all kinds of hardware are installed in the iSee space on the table, the team has assembled the first prototype of its products, this is a belong to one of the company co-founder of Lexus car. Several engineers sat behind large computer monitors, staring at the code.
Compared to driverless cars such as Waymo, Uber, or Ford, iSee may seem trivial, but the technology it is developing may have a major impact on many areas of the current application of artificial intelligence. By allowing machines to learn from less data and build some kind of common sense, their technology can make industrial robots more intelligent, especially in unfamiliar situations. In the field of artificial intelligence, artificial intelligence has made remarkable progress, thanks to the depth learning technology. The technology uses neural networks that are powerful enough to handle huge amounts of data.
When inputting large amounts of data, a sensitive neural network can recognize fine patterns. For example, to provide a neural network with a large number of dog pictures, it will know how to find dogs in all images. But the ability to learn in depth is limited, and we need some more radical new ideas for the next leap. Dogs, for example, recognize depth learning systems that do not understand that dogs usually have four legs, fur and a wet nose. Moreover, without further training, it could not identify other animals, or draw a dog.
Lack of common sense must have brought some problems to the autopilot system. Last year, for example, a Tesla operated in a semi-automatic mode in Florida, where a truck lost its judgment when a truck crossed the motorway. A human pilot may quickly and safely find out exactly what happened.
Zhao and Debbie Yu, one of his co founders, showed in the video a Tesla accident in China, when the car crashed into a street cleaner. "The system was trained in Israel or Europe, and they didn't have this kind of truck," Zhao said. "It's based on testing, and it doesn't really understand what's going on," he says."
ISee was built to understand how human beings understood the world and to design machines that could imitate humans. The other founders of Zhao and iSee are from Josh Tenenbaum laboratories, a professor of brain and cognitive sciences at the Massachusetts Institute of Technology, and now an adviser to the company.
Tenenbaum is good at exploring how human intelligence works and uses this insight to design new AI systems. This includes, for example, an intuitive study of physics, or even a child's performance. Children are able to understand the behavior of the real world so that they can predict how unfamiliar situations will develop. Tenenbaum explained that the physical world is closely related to the understanding and intuitive understanding of psychology, including the prediction of the behavior of others, for example, by observing a person's behavior to predict he picked up the cup action.
The ability to learn in different situations is also a sign of human intelligence, even the most
ISee, a derivative of the Massachusetts Institute of Technology, is developing and testing an autopilot system that employs a novel approach to artificial intelligence. The start-up company does not rely on simple rules or machine learning algorithms to train cars, but is inspired by cognitive science, which allows the machine to have a common sense and is able to quickly cope with new situations. It is developing an algorithm that tries to match the way human beings understand and learn the physical world, including interacting with others. This approach may allow auto drivers to better cope with unfamiliar situations and complex traffic conditions.
"Human thinking is very sensitive to physics and social cues," says YibiaoZhao, co-founder of iSee. "Current AI is relatively limited in these areas, and we think this is actually a missing link in driving."
Zhao company has just begun to take shape, is a new investment company created by Massachusetts Institute of Technology, to provide funding for the innovation of local technology companies, a small team of engineers working in a small space in the laboratory. Just a short distance from the campus of Massachusetts Institute of Technology, it overlooks a street where drivers scramble for parking spaces and actively crowd into the traffic.
The laser radar sensor and all kinds of hardware are installed in the iSee space on the table, the team has assembled the first prototype of its products, this is a belong to one of the company co-founder of Lexus car. Several engineers sat behind large computer monitors, staring at the code.
Compared to driverless cars such as Waymo, Uber, or Ford, iSee may seem trivial, but the technology it is developing may have a major impact on many areas of the current application of artificial intelligence. By allowing machines to learn from less data and build some kind of common sense, their technology can make industrial robots more intelligent, especially in unfamiliar situations. In the field of artificial intelligence, artificial intelligence has made remarkable progress, thanks to the depth learning technology. The technology uses neural networks that are powerful enough to handle huge amounts of data.
When inputting large amounts of data, a sensitive neural network can recognize fine patterns. For example, to provide a neural network with a large number of dog pictures, it will know how to find dogs in all images. But the ability to learn in depth is limited, and we need some more radical new ideas for the next leap. Dogs, for example, recognize depth learning systems that do not understand that dogs usually have four legs, fur and a wet nose. Moreover, without further training, it could not identify other animals, or draw a dog.
Driving is more than pattern recognition. Human drivers always rely on common sense about the world. For example, they know that buses take longer to stop, and a large number of pedestrians are suddenly produced. It's unrealistic to want to spin out every situation that an automatic car is going to run into. But people can use their common sense of the world, understand it, accumulate it through a lifetime of experience, and act wisely in all kinds of new situations.
Lack of common sense must have brought some problems to the autopilot system. Last year, for example, a Tesla operated in a semi-automatic mode in Florida, where a truck lost its judgment when a truck crossed the motorway. A human pilot may quickly and safely find out exactly what happened.
Zhao and Debbie Yu, one of his co founders, showed in the video a Tesla accident in China, when the car crashed into a street cleaner. "The system was trained in Israel or Europe, and they didn't have this kind of truck," Zhao said. "It's based on testing, and it doesn't really understand what's going on," he says."
ISee was built to understand how human beings understood the world and to design machines that could imitate humans. The other founders of Zhao and iSee are from Josh Tenenbaum laboratories, a professor of brain and cognitive sciences at the Massachusetts Institute of Technology, and now an adviser to the company.
Tenenbaum is good at exploring how human intelligence works and uses this insight to design new AI systems. This includes, for example, an intuitive study of physics, or even a child's performance. Children are able to understand the behavior of the real world so that they can predict how unfamiliar situations will develop. Tenenbaum explained that the physical world is closely related to the understanding and intuitive understanding of psychology, including the prediction of the behavior of others, for example, by observing a person's behavior to predict he picked up the cup action.
The ability to learn in different situations is also a sign of human intelligence, even the most