Replay at https://www.youtube.com/YouthAILab
Brief Overview:
Dr. Irfan Essa is a distinguished professor in the school of interactive computing and executive director of the Interdisciplinary Research Center for Machine Learning at the Georgia Institute of Technology. Dr. Essa is also a Senior Staff Research Scientist at Google Inc. and leads the Google Research Office in Atlanta. He is a world-renowned AI expert, who defined what exactly machine learning, artificial intelligence, and computational journalism are in this episode of the Youth AI Speaker Series. He first introduced the conceptual meanings of artificial intelligence-related fields like computer vision, robotics, HCI (human-computer interaction), and computational journalism. He also focused on how a wide applicable range of the term “intelligence” differentiates AI apart from other computer technologies. Artificial intelligence demonstrates natural intelligence displayed by humans through devices that perceive their environment and achieve specific goals. In contrast, machine learning uses a series of algorithms to build mathematical models of data and predict without being programmed to perform a task. Dr. Essa emphasized that there are fields where AI and ML are significantly used, like pattern recognition and data science. After mentioning the value of each subject, Dr. Essa talked about the issues that come along with the increasing usage of data. For example, there are many cases where the auto-correct feature makes a mistake when recognizing words, and face recognition fails to identify the user. He advised not to overestimate AI technologies as some of them have chances of failing as much as they can succeed.
Full Summary:
As a distinguished professor in the school of interactive computing and senior associate dean in the college of computing at Georgia Institute of Technology, Dr. Irfan Essa presented the true definition of machine learning, artificial intelligence, and computational journalism. Because there are a variety of fields associated with these upcoming technologies, he drew a line between each subject and how they interact with one another.
Dr. Essa first started off the talk with his experience as a professor and staff research scientist at Google where he explored fields of computer vision, robotics, computational journalism, and other AI and ML-related technologies. Through his long educational journey, Dr. Essa claimed that it is very important to interact with the upcoming generation of students as he views student-professor interaction as a symbiotic relationship where the educators give the students the basics and foundations of what they want to learn and the students teach the educators about what is new and exciting in today’s world.
He introduced technologies where artificial intelligence and machine learning play a great part. Some of the fields that Dr. Essa mentioned were computer vision, AI, machine learning, robotics, computer graphics, HCI (human-computer interaction), and computational journalism. He stated that while computer vision is about the process of understanding what images are, computer graphics is about the whole process of making those images. Just like how human eyes are able to see and process information through vision ability, computer vision enables computers and machines to see and interact at the same level as humans. In contrast, AI and machine learning use the information to understand the environment around them and react to various types of subjects or make decisions. Dr. Essa went on and explained the similarity between robotics and anatomy, HCI with linguistic ability, and computational journalism with writing ability. He explained the AI-related technologies in terms of how they are modeled upon humans themselves.
Dr. Essa then discussed certain technological fields he and his team are researching. He talked about computational video, which is the analysis of cameras and video, and how it can be used to improve media creation and understanding. He held YouTube as an example, as they consistently dedicate time to research on computational video and update their server through careful evaluation of the website. He also talked about behavior and activity recognition, where they create a model that predicts, analyze and assess movements, and action modeling, where they train the model to move, animate, plan, react and understand dynamic environments. Lastly, Dr, Essa covered computational journalism, which analyzes how information or media is created, shared, and distributed. He stated that AI and ML are extensively used in his research to improve the quality of analysis.
Furthermore, Dr. Essa brought terminology to the table, defining what artificial intelligence and machine learning are from his own perspective. He focused on how a wide applicable range of the term “intelligence” differentiates AI apart from other computer technologies. While artificial intelligence demonstrates natural intelligence displayed by humans and other animals through the usage of devices that perceive their environment and take actions to achieve specific goals, machine learning uses a series of algorithms to build mathematical models of sample data and predict without being explicitly programmed to perform the task. He also mentioned two significant fields that AI and ML created: pattern recognition and data science. Pattern recognition uses machine learning to predict according to the given rules and data science uses scientific methods to extract knowledge from structured and unstructured data. To describe the relationship between the mentioned technologies, he used the following analogy:
“The goal of developing true AI is like going to the moon, machine learning is the rocket fuel, that is needed to get us there, and data science may be the rocket ship”
Following that, Dr. Essa talked in depth about computer vision and computational journalism. As explained briefly in the introduction, he defined computer vision as how computers can gain high-level understanding from digital images or videos, trying to automate tasks that the human visual system can do. Computational journalism, on the other hand, is about how information is gathered, distributed, organized, understood, and shared. He compared the process of computational journalism to computational photography, which is a sub-part of computer graphics and vision; it focuses on the physics of how the light is captured in the sensor of technology. The process of gathering is similar to how light is captured in cameras or sensors while the method of sharing is similar to sharing photos online. However, Dr. Essa wanted to acknowledge that the goal of computational journalism is focused on how information is verified and accurate.
Dr. Essa raised concerns on some issues about AI and ML, such as the accuracy and fairness of the result produced by biased datasets. He emphasized that there are definite limits to ML because of the boundaries of using data. Not only that, he claimed that the term AI is being overused and its related technologies are being overestimated because of the hype around the subject. He advised not to overestimate AI technologies as some of them have chances of failing as much as they can succeed.
With that, Dr. Essa wrapped up his presentation with the conclusion that it is important to acknowledge different usages of AI and ML technologies as they can be used in a wide range of sectors. Be aware of many improvements that will come along with the development of AI technologies but also its consequences. The Q&A closed out with the idea that computational journalism can be effective to be aware of misinformation and that machine learning covers a lot of studies, as Georgia Institute of Technology pursues an interdisciplinary policy where not only the school of computer science works with machine learning but also the school of engineering, sciences, humanities, business, and design. Look for more resources in Udacity’s Georgia Tech-Machine Learning (YouTube).
For previous Youth AI talks, visit https://www.youtube.com/YouthAILab
0 Comments