Joseph Sagliocca
IS-251
Blog #1
Training AI to Save Lives
In today’s society, many people are finding new ways to utilize artificial intelligence. These new ways of using AI has some people fascinated, and others frightened. Many people are alarmed by technology with AI because of the various things that can go wrong with it. Society believes that technology with AI is not safe right now due to its inability to understand when a human’s life could be at risk. Many innovations with artificial intelligence do not have ethical reasoning built into them. This means that technological items with AI can make mistakes that lead to dire consequences such as a human getting hurt or even killed. As many people believe this is the reality for all innovations with AI, these innovators are working hard on making their AI smarter. To make them smarter, they are starting to use Reinforcement learning (RL). Reinforcement learning is being used to help companies simulate actions that will help make their AI smarter. They can run tests to collect data on how the AI should perform in situations to reach its desired goals. They then take this data and build reinforcement learning models that in turn could make the AI perform better in these certain situations. This process has been around for a few years, but companies are starting to learn how they can make their inventions with AI even safer for humans.
Kobielus talks about using RL to help AI robots perform better in scenarios such as working collaboratively with humans and other robots. He also says that RL could help safety features on cars with AI. This could include defensive maneuvering and avoiding collisions. This new way of developing AI can change the world. RL has the ability to make AI not only work better at its intended job, but also work better at the jobs it is not good at right now. Some robots with AI have trouble preforming certain parts of their job when they are first built. With this new process, inventions can go through countless changes until it can perform all of its intended tasks with no problems. Kobielus does say however, that it is not easy to change the AI quickly with RL. He instructs companies to continue to rework their AI frequently.
Reinforcement learning is an amazing process. I could’ve never believed that we could continue to teach AI from putting it through situations. As we continue into the future, AI will only continue to get better because of RL. Technology with AI is starting to get smarter by the day and as time goes on, people will begin to feel safer around it and using it.
Source:
Kobielus, James. “Training AI to Save Lives.” InformationWeek, 24 Jan. 2019, www.informationweek.com/big-data/ai-machine-learning/training-ai-to-save-lives/a/d-id/1333719.
1 comment:
This reinforcement learning seems like a great improvement for the world of AI. As stated in the blog AI is not always reliable which is scary for the human race, because we are the ones putting our trust in something that may or may not work. I think that this reinforcement learning could change the game and help gain the trust of people who are skeptical about AI technology that can save lives and prevent injuries. When the day comes that this life-saving technology is in its prime there will still be the question of whether we should trust it with our lives, but as people hop on board society will most likely be influenced and it will become an extremely common thing whether it's used in car technology or other technologies.
Post a Comment