These bleak arguments usually follow three lines of thinking:
- AI systems have already surpassed human intelligence, which is inherently bad. For example, AI systems can beat the world’s best chess and Go players.
- AI and machine learning has crossed a point of no return and will eventually allow us to automate everything, leaving nothing for people to do.
- AI models have become so complex that even the scientists who create them no longer understand their operations.
These arguments may seem intuitive, but they aren’t rooted in fact. Let’s briefly discuss them.
Argument 1
This argument posits that AI has finally surpassed human intelligence – and that that’s an unquestionable source of trouble. The reality is that AI has been around for decades, doing things better than people in tasks where people tend to make errors, exercise poor judgement, or simply can’t meet work demands. For example, the financial industry has been using AI to automate and streamline credit scoring and loan offering decisions since the 1970’s. This has made the credit and loan application process more secure and more accessible to millions of people. That AI systems can make better, faster, and more secure decisions than people in this industry have been a powerful force for good, and the same holds true in other industries and applications, too.
Argument 2
The pace of technological development – and the fear associated with it – is a tale as old as time. The fact is that the pace of AI development should be a cause for celebration and motivation to keep pushing forward. After all, we’ve all benefitted from AI-driven systems that deliver better medical outcomes and recommend treatments and consultations for a host of chronic diseases like high cholesterol or diabetes. In fact, AI systems produce most of today’s lab work, which alerts healthcare professionals to abnormal or noteworthy results without them needing to comb through it manually. This is another example of AI helping unburden already-overworked people, which helps them perform their jobs better, more accurately, and in less time. In general, the medical field is an ideal example of AI and human intuition combining to deliver optimal outcomes.
Argument 3
The last, and most popular argument AI doomsayers often raise is that AI models have grown so complex that data scientists and software developers can no longer understand them. This is hyperbole – and in any case, any data scientist or developer worth their salt wouldn’t let this happen. Today’s AI experts work with explainable AI technology, which gives users exact insight into how AI models gather data and make decisions. Explainable AI is built into many of today’s common software tools and is becoming the industry norm. In fact, many of today’s vendors and tools can explain every decision right down to an individual data point. So far from becoming more convoluted, today’s AI systems are becoming both more powerful, and more explainable.
Conclusion
Even though AI technology isn’t new, it has grown by leaps and bounds because computing power has done the same. When technologies make great strides, there will always be a natural fear of the unknown. But when we implement technology in disciplined, responsible ways – and pair that with laws and best practices that ensure a standardized, ethical landscape – we can make the world a better place. Rather than an opaque dark cloud hanging over the future, we should view AI as a gateway leading to opportunity and progress.
By Mamdouh Refaat, chief data scientist, Altair
