The progress of a civilization of anything requires its systems to function harmoniously and in order, without much of chaos. Yes, disruptions do happen; however, if they are inclined towards the larger good of the civilization, it leads to a better state of affairs. Any malicious element, if found, is required to be subdued by the forces of the good. Look at the micro level, the system of our bodies fight against pathogens and malevolent foreign bodies with its army of immune system. Animals maintain the security of their herd or pack by keeping the hunters at bay through their collective. Even humans have intricate systems of law and order that check wicked elements from damaging the fine fabric of the society.
A common denominator is the domination of collective conscience over the fragments of awful anti-social elements, which maintain and sustain peace and prosperity of these groups. Any act of decoding a crime requires a higher intelligent mind to analyze the act and overpower the nefarious designs of the criminals. But as intelligence takes new and ore evolved form with every passing day, it is doubtful if the balance will remain tilted towards good. With the way AI is advancing, the day is not far when AI may end up being dangerous to our society. And in case the AI systems happen to be superior to us, then the only opponent to beat them will be another AI – thereby leading to the first clashes of the AI systems outside the control of humans.
AI-based malware are already a concern for technology experts. An ordinary malware is a mere piece of code that runs on a server and can be detected with proper security mechanisms in place. But an AI-based malware will learn from the circumstances it is in and will evolve over time to learn how to evade detection. The moment it detects a risk of being discovered, it will readjust its evasion techniques and will be a headache for security professionals. And it is not a hypothesis anymore. IBM Security had created DeepLocker in 2018 that was designed to be stealthy and autonomously select the computer to be hacked based on a facial recognition and audio analysis algorithm. Though IBM would not have abused the malware for any malevolent motives, what’s alarming is that the technologies used to create DeepLocker are anyways publicly available. Beating AI in its own game will require humans to create a good AI that can outperform malicious AI, and can predict its behavior much before the malware launches its attack.
Malevolent AI can also turn out to be an ally of the fraudsters. AI systems may end up being capable of understanding how to evade or compromise security checks of banks and financial institutions by earning from its failures over time. The hacker of tomorrow may not necessarily be a human draped in a hoodie and sitting inside a dark chamber. It can very well be an automated AI that needs just a seamless internet connection to tirelessly keep hitting the walls of security before they collapse.
Malicious AI can also be the driver of discord in societies. Today, behavior analytics is anyways evolving by leaps and bounds. Our actions on social media, e-commerce websites and search engines are anyways being continuously monitored. The day may not be far when malicious AI will simply read our intentions and provoke the dormant anarchist inside us through fake posts and propaganda. The threat of social manipulation by malicious AI will soon loom over us, as smart algorithms may identify the best people to target to incite friction against establishments. This is again where our forces of better and improvised AI forces of the good will have to duel against the malicious ones to maintain order in the society.
Misaligned intelligence needs only an internet connection for connecting machines and systems against the will of humans. As AI systems turn more competent in thought and judgment, they can easily develop goals unaligned with our goals. The artificial judgment may find human rationale incompetent of taking the right judgment or may even classify human intelligence as detrimental as humans have damaged the planet beyond repair over millennia. The artificial judgment may very well classify humans as a threat to themselves and may decide to exterminate the biological race altogether. Add to it the chances of all AI systems providing consensus to this and working in tandem to get the extermination goal fulfilled. No wonder, we may end up at the mercy of our own creations gone wrong. And we may need to have our own force of AI systems with ‘service to humans’ programmed into them, so that we have a force to fight for us.
The day to panic has not yet come. However, it is never too early to exercise caution. With the way artificial intelligence is evolving, the day may not be distant when AI will be pitted against its own brethren and will redefine the threats humans face for their existence.
