The Double-Edged Sword of AI

Sam Bocetta Feb 10, 2020

AI has become a major aspect of our daily lives and as an industry, it is experiencing rapid growth and innovation.

While it still has a long way to go before it resembles what we see in sci-fi movies, AI has already provided us with plenty of benefits, from how it can be effectively used as a communication tool to paving the way for next-generation marketing and advertising. But at the same time, AI has also undeniably shown us its dark side with several possible hazards.

This double-sided nature has called for many to suggest a total halt in AI research and development, despite its advantages. As with all versatile tools, AI has the potential for great good or great harm. But in which way does it currently lean?

The Endless Advantages of AI

Machine learning is undoubtedly going to play a huge role in how the Internet continues to develop complexity and sophistication. As we enjoy the extra advantages of a more connective and adaptive internet, AI and machine learning technology are greatly benefiting the health industry.

AI bots and analysis systems can properly (and arguably more accurately) diagnose patients at medical centres across the world. This can allow patients to get the care they need without the human bias or margin for error that we currently work with. Even further, AI will most likely be able to consistently detect cancer and other diseases with higher degrees of accuracy than humans can manage.

We’re also seeing how artificial intelligence and machine learning are being used to revolutionize business marketing as we know it. For instance, AI has become the primary marketable asset for cloud and SaaS-based businesses. This is huge because 86% of organizations are expected to turn to have their software needs met by SaaS by 2022.

AI is also going to be a critical part of the development and improvement of transportation and mobility. Naturally, systems that handle millions of humans moving around the same time need to be exceptionally complex. Regular human minds have difficulty handling all the moving parts.

We’ll likely see AI impacting transportation most dramatically with the advent of the self-driving car. If such a vehicle becomes popular enough, AI-driven cars may become more common than human-driven vehicles, causing traffic to become one smooth, thoroughly networked system. Public transportation, of course, can also be improved from AI automation and the enhanced speed at which machines can operate.

In general, AI is effective and valuable because it can handle greater amounts of information and distil large data sets into meaningful actions or suggestions. There are possibilities that machine learning algorithms can help us with communication or PR, such as when public servants need to know how best to address their constituents.

There may even be a silver lining to the problem of AI being used for cyber threats, however. AI can make a hacker’s attempts at theft or digital vandalization more successful, can’t AI also be leveraged to increase network security?

Many businesses and industries are looking into this solution more earnestly, particularly in response to an increase in AI’s ability to mask its synthetic nature. A good example of this can be seen, once again, with Twitter; it’s a known fact that many profiles are actually “bots” – false profiles created by an algorithm to post for the purpose of misinformation or follower-influencing.

AI Presents Lurking Dangers

Conversely, there are already many instances of AI and machine learning causing more trouble than many would say is worth it.

AI-controlled bots can also be used to influence social media threads or to tank the reputation of companies through methods such as review bombing, which is devastating because 90% of consumers always check online reviews before buying from a business. Another easy example of AI’s sophistication and proliferation across social media is the self-learning chatbot that, after being exposed to Twitter for just a few days, developed disturbingly sexist and racist attributes.

AI and machine-learning bots are most infamously being used in recent years as catalysts for the increased proliferation of malware in cyber attacks. Even those who take steps to improve their site’s security may find AI cyber threats to be an issue regardless.

While many look to AI to solve social issues, machine learning also presents several challenges that may truncate its actual usefulness for a given problem. For instance, many machine learning algorithms can accurately take in large amounts of data, but their preferred solutions may fail to address the root cause of such a deep, historical issue as poverty, for example. At this point, AI doesn’t have the depth of thought many human minds can bring to the issue.

The trend of user data being taken and used without owners’ consent by major tech companies is worrying for those concerned with AI, too. AI becomes more sophisticated and can take advantage of this data or provide it to someone with ill intent relatively easily if you aren’t careful.

In fact, the biggest problem with AI and predicting its effects or possible threats is that AI and machine learning developments necessarily make it difficult to create frameworks or safe practices that “control” the technology. Other historically notable technological developments are often accompanied by safety measures to prevent misuse of the tech or accidents.

Machine learning requires extreme adaptability on behalf of the AI in question, making it very difficult to create safety frameworks without harming the bots’ ability to learn. It’s a circular issue that’s tricky for even the best minds to solve.

In addition, humans as a whole are nowhere near consensus when it comes to what is acceptable behaviour among themselves, let alone AI designed to emulate us. Without a moral or ethical guide through which all AI or machine learning developments can benefit, who can really say what is and is not acceptable for machine learning technology?

Finally, all nations and industries thinking about utilizing machine learning for greater productivity will need to consider the jobs that will be lost from that development. This may yet have unseen impacts on the economy and specific businesses. Already bots can handle simple jobs such as customer service assistants for websites, and customer interaction complexity is increasing.

As cybersecurity becomes more tenuous with each year, increased AI development may be seen by some as a bad thing purely because machine learning allows hackers to become more adept at cracking networks and stealing personal information.

Fighting Fire with Fire: When AI Defenses Meet AI Threats

While these machine learning algorithms are becoming sophisticated enough that it’s difficult for us to tell whether we are talking to a human or bot, AI algorithms may be able to screen them out more effectively. Such defences will likely become necessary because of scale as well; millions of bots can be created every month, and human employees can only screen so fast. AI doesn’t necessarily have this limitation.

All in all, time will tell whether AI ends up being a net benefit or hazard for human development. In all likelihood, it'll remain a bit of both, like with most other major historical innovations.