We have been talking about the power of machine learning and AI all over the year. AI is described as the super intelligence that can think, learn, and reason about the world. It seems like we sometimes forget that AI, to its nature, is just code and statistics with various limitations.
Data scientists are building machine learning algorithms to understand the world and extract marketable/profitable insights from the big data humankind is generating. In many cases, AI is painted as a living entity that has its own thoughts about the world while in fact, it is machine algorithms that can find patterns in big data through analytics and programming. AI creators say that their AI can learn while all they do is providing machine algorithms with new sets of data that they have picked and curated.
It is misleading when data scientists claim that algorithms have “created” something while all they do is to use statistics they get from training data to generate some unexpected results. If that is what makes AI the superhuman entity that can end human civilization, we don’t need to worry about that illusional prospect.
As the research and development on AI and machine learning progress, some developers and businesses even bring their success in AI further by exaggerating the ability of their AI algorithms. It is understandable that they want to make people think of AI as the superpower that can change the world for marketability and profits. However, such marketing tactics can be dangerous and harmful if those machine learning algorithms are rushed to production into products that can affect human life like those applications in healthcare, driverless car, or government operations.
Nobody can deny the power of AI and deep learning; nonetheless, they should be seen as they are so that they can be properly developed and used. AI, even when placed into a silicon body, or an artificial brain, is still a mathematical incarnation, not an intelligent entity. Algorithms can be trained to analyze and make sense of the big data from the world, but they are purely machine instruction that humans need to guide and manually tune for each application. They can’t learn to adapt themselves to the new environment as living entities can.
Many people may argue that the above issues are simply the matter of terminology and shouldn’t be taken too seriously. But we should. Because we are deploying AI applications into many areas that can directly affect human life. If we continue to praise AI and machine learning as the almighty, we may overlook their limitations and risks when adopting them. The consequences of a single mistake in AI deployment can be severe, dependent on the scale of the operation.
AI and machine learning will go far beyond their current states, bringing critical changes to the world. We should not deter their advance and development; however, we should stop overhype about them. The key to successfully developing and deploying AI is to be realistic about both its abilities and its limitations. That is the only we can leverage the technology for our sake without risking our own safety.