Common Misconceptions about Artificial Intelligence
It is common to over-imagine the intelligence of the machine as compared to a living person. Because the machine appears capable of identifying relevant news article to a topic, it is easy for a person to imagine that the machine actually reads and understand a news article like a human person.
However, the truth is closer to one where the machine is looking for patterns (e.g. appearance or absence of certain words in the article, the frequency of certain words, how words/strings of words appear next or near to each other, etc) without assigning any real-world meaning to it. The amazing thing is that there is useful information embedded in such patterns and the machine is learning them without hardcoded rules. It is how we harness such useful information by using it to train machine learning models to produce decisions or actions (e.g. classifying articles by relevant topics) that makes it appear so intelligent.
While good AI systems are able to perform near to human-level of performance, it may not be doing it in a way we think it is. Sometimes, this is exhibited when the AI makes a decision that is illogical and immediately obvious to a human person. This is usually where the expectations gap with users tends to appear.
In this context, what are some of the benefits that can be obtained from adopting AI?
Firstly, there are conventional benefits of automation that are applicable in the use case of using AI to automate human tasks. AI systems can help automate low-level tasks, especially where errors made by AI will not be fatal to the operation. For example in ScoutAsia, Standard Scout AIs help take away the tedium of reading and monitoring the deluge of news coming in daily. This can be easily scaled to large volumes of news articles where it would have taken an army of humans to match up to.
Another related benefit is the consistency that a machine would apply to its decision across all articles that it processes. With a team of people, a usual challenge is getting different persons to make the same decision as each other. Of course, this depends very much on the robustness of the AI model as it may also inherit bias that is present in its training. But in a sense, this is similar to how human could also inherit bias based on the training he or she had received.
One interesting benefit that is often highlighted in the market is how AI systems outperform existing systems or even humans (AlphaGo beating the best human players at the game of Go). This is an example of the AI’s ability to self-learn at scale. In ScoutAsia, the AIs were collectively trained on a proprietary dataset of more than 500,000 training examples and growing each day. In human terms, this would roughly be equivalent to a person with 10 years of training experience in reading and classifying news by its relevant topics!
While ScoutAsia helps to make life better for you, from time to time, you can also expect ScoutAsia to present an interesting but unexpected article to your attention. These are just some of the benefits that AI can bring.