By Prof. Vidhu Shekhar is Assistant Professor, Finance & Economics, SPJIMR
Not too far in the past, having robots and machines do our work used to be the stuff of dreams and science fiction. Things of which heavens were made. But as the integration of artificial intelligence (AI) into our daily lives has rapidly shifted from a futuristic vision to a present-day reality, we find ourselves in a lot of trepidation about what the new future has in store for us.
AI technologies, ranging from chatbots and virtual assistants to autonomous vehicles and advanced medical diagnostic tools, have become increasingly sophisticated and ubiquitous. This transformation has led to both optimism and concern regarding the implications of AI, particularly its impact on jobs and various industries. While AI offers remarkable opportunities for enhancing efficiency, productivity, and innovation, it also poses significant challenges. While much debate has focused on what AI can do, it is crucial to understand what AI cannot do, at least yet. This knowledge can help us navigate and thrive in the evolving post-AI world.
AGI vs ASI
Before we go deeper into what AI can or cannot do, it’s essential to understand the difference between Artificial General Intelligence (AGI) and Artificial Specific Intelligence (ASI). Simply put, AGI resembles human intelligence, capable of performing multiple tasks without explicit training. For instance, humans can cook, drive, sweep, and do math, all with varying proficiency. Translating this versatility to a machine would result in AGI.
On the other hand, ASI is specific to task-trained AI. That is, say, it can just do driving. Ask it to cook a bit; it will fail drastically. But driving can do extremely well, maybe even better than the best human driver. Similarly, the one trained to cook can only and only cook. That is why specific intelligence.
All the AI advancement we have seen to date is in the area of ASI. AGI hasn’t been created yet, and there are debates about whether AGI can at all be created within our current understanding. Some argue that AGI needs consciousness, that is, it needs to be alive like humans. And since our understanding of consciousness is very poor, we don’t seem near to creating AGI – a single machine that can do many tasks.
The difference between ASI and AGI also comes from current AI machines working only from their training data. They can only work on what they have been trained to do, which means they cannot generate anything truly ‘original’ or highly creative if it is not in their training data. It takes a human, or an AGI, to make something entirely original or very creative.
So, what really can AI not do today
Based on the above and knowing that today’s AIs are all ASIs, we can safely say some things that they cannot do.
First, anything that requires constant adaptability is not something it cannot do. AIs have low resilience to big changes. In fact, it is a big drawback of automation itself. As an example, one of the reasons that Indian airports were able to adapt faster and better to COVID protocol during the pandemic, as against European or other countries, was because Indian airports are heavy in human-based processes. They can change quickly to a new process. However, try changing the machines installed to a new process. It is a nightmare.
Therefore, tasks that require on-the-fly adaptability and problem-solving in unpredictable environments are challenging for AI. For instance, a human firefighter can make quick decisions based on the changing dynamics of a fire, assessing risks and altering strategies as needed. Similarly, emergency medical responders often face unpredictable scenarios that require rapid judgment and flexibility.
Then, the tasks that require fine motor skills and intricate hand-eye coordination are beyond current AI capabilities. Delicate surgical procedures, assembling intricate components, and crafting detailed handmade items require human precision and adaptability. Complex social interactions are another area that involves context, body language, tone, and cultural nuances and remain challenging for AI. Negotiating business deals, mediating conflicts, or providing emotional support demand empathy and intuition, which AI cannot fully grasp.
Another area is out-of-the-box thinking and innovative problem-solving. While AI can solve problems within its programming and data confines, it struggles with novel and interdisciplinary solutions. Humans excel at inventing new technologies, developing unique business strategies, and creating innovative marketing campaigns, drawing on a wide range of experiences and interdisciplinary knowledge. This requires a level of ingenuity that AI has yet to achieve.
Lastly, AI cannot create truly original works of art. Whether in painting, music, or writing, creating art involves technical skills and deep emotional and cultural understanding. AI can generate art based on learned patterns, but it lacks the originality and emotional depth of human creators. The nuances of human experience and emotion are beyond current AI capabilities.
Where does that leave us
Of the many lessons above, a key takeaway is that being a human generalist may become more valuable going ahead. ASI excels in narrow areas, but without AGI, it cannot be a generalist. Therefore, a generalist with diverse skills may thrive better in an AI-driven world. While AI has transformed many industries, it cannot yet replicate human creativity, emotional intelligence, moral judgment, common sense reasoning, flexibility, or physical dexterity. These areas highlight where humans excel over AI. In fact, my personal theory is that as AIs become more and more common, a true masterpiece of human creativity may become more and more valuable.
In conclusion, understanding the current limitations of AI helps us navigate the post-AI world we live in today. By focusing on the unique strengths of human intelligence and creativity, we can adapt and thrive in an era where AI is increasingly prevalent. The key is to leverage AI’s capabilities while recognising and nurturing the irreplaceable aspects of human cognition and ingenuity.