To err is AI, to make a mess is human - The competitive edge of AI over humans
Remember Rajnikant Sir… the good old super hero who at the end of the movie (and also throughout the movie) beat villains, punched them, kicked them and still managed to lit his cigarette by throwing the stick in the air, shooting with a gun and lo and behold, the stick falls and fits his slightly open lips in style… Yenna rascala, mind it.
AI can certainly not do it. But it can do many other things that is below dignity for Rajni Sir.
AI has started giving serious competition to humans, not only the dumb humans who do repetitive tasks (Gadha majoori) but even the ones who carry out highly intellectual work (Lambi race ka Ghoda). Some humans adapt to the challenges posed by AI by joining them and some pass into oblivion by fighting them. As we say, the AI will not replace humans but humans with AI will replace humans without AI.
Healthcare has seen many experiments involving AI and the results are great
Artificial intelligence has done wonders in healthcare. Its application in diagnosis, drug discovery, and genome mapping is increasing every day. The sheer capacity of AI in analyzing the vast set of data, finding patters, clustering based on peculiar characteristics, and delivering output beats the most intelligent human. This is certainly not a new thing as IBM’s blue defeated the world champion Gary Kasparov about 25 years back. What is different this time is the widespread application of the same artificial intelligence in real life to solve common problems.
A study was done in Israel and the USA. The diagnosis was divided into 3 groups, doctors without AI, Doctors with AI, Only AI. AI refers to LLMs here [1].
They were supposed to diagnose few cases. The scores of the three groups were the following.
Doctors with their conventional resources without AI – 74%
Doctors with AI – 76%
AI only – 92%
The study concludes that using LLMs alone has a significant edge over the two groups of humans (with and without AI). The difference of 16% is huge. The reason for this level of performance is attributed to prompts provided to the AI application.
Another report was done on the basis of work done by Google deepmind [2]. The report says even when there are two doctors working together to diagnose breast cancer, the AI diagnosis is as accurate as them. This is when the system is not preloaded with patients’ history which is normally available with doctors. Imagine the result if the AI system is also infused with patients history. The experiment was able to reduce false positive cases (diagnosing breast cancer where there was none) cases by 5.7% and false negative cases (missing to diagnose the breast cancer when it was there) by 9.4%. The report further claimed that another experiments involving 6 radiologists versus AI, the AI was able to outperform all of them.
As per this report on eye disorder diagnosis [3], the same experiment was conducted with three groups of doctors, nonspecialized junior doctors, expert eye doctors, and of course the super experts.
AI driven diagnosis outperformed junior doctors which is to be expected. After all, you spend millions on AI system and pay peanuts to junior doctors. You see... money is incentive. The AI diagnosis was as good as expert doctors. However, the super experts outperformed the AI system.
Does this mean that AI will replace doctors? I have no idea. The prediction of future is something where the biggest brains have failed so I will not go to there. But here is what humans would be required for in such cases.
1. Provide prompt. Standardize it with all possible cases and details. Keep updating this as humans get new cases and deviations.
2. The other important task, as of now is, to study the diagnosis by AI in details and suggest treatment. However, we don’t know how long this edge will remain with humans. AI will take over this too.
3. Relationship and soft power are very important in treatment. This is one area where AI is still lagging behind by a big margin. Humans can interact with patients, tell funny jokes, provide right guidance, and discuss the treatment. Telling funny jokes is optional even though sometimes, that’s the only thing needed to cure a patient.
The most obvious application of AI is investment and trading
This is a no brainer. AI can easily beat human fund managers especially when the history has shown that a monkey throwing dart in a set of stocks to form portfolio has beaten 88% of fund managers. Not directly but if you look at the fact that 88% of fund managers cannot beat the index returns [4].
Monkey throwing dart has outdone the index 88% of the time [5], this logically follows that a money throwing dart and selecting its portfolio from an index can beat 88% of the fund managers.
So the real competition is between the monkey and the AI. Monkeys must devise a new strategy else the dart will be taken out from them. Humans have already capitulated to monkeys long ago.
A hedge fund in Japan replaced analysts with AI and beat the market in performance in the first 6 months as this report points out [6]. There are many robo-advisory startups that are doing pretty well in investment recommendation. The advices are based on data, devoid of individuals’ biases and thoughts. The best part about AI driven investment advisory is that it ignores the story and narratives that we humans are really taken for granted by. Stories, while engaging, distort data. Investing being a number driven game has to give way to hard hitting numbers and not get swayed by stories. The story of second tier cities booming in India has killed more startups than expenses of running a startup in south Mumbai.
However, there are areas where AI will take some time to learn. This is where human fund managers, traders, investors still have advantage over monkeys and AI.
1. Incorporating new information immediately. Special situations like Covid, War (what is known as systematic risk) are still better handled by humans. We are yet to develop an AI which would immediately factor such events.
2. Psychology in investing is extremely important. It is not a science as depicted in valuation. What would an investor do if he or she has taken a buy position based on the charts and the stock falls further? Wait or come out immediately. What if you come out and the stock starts rising. If wait is a better strategy, wait till what. Would you define wait in terms of days or loss percentage. These are few things that will come only as a result of experience and keen observations. AI may be able to reflect that too in future but as of today, this is still in the realm of human.
3. Role of Governments keeps changing as a response to stock market. Random stimulus, interest rate cuts/raises based on its own reading are still beyond the capacity of AI. The nature of Governments are responsible for this. For example, dems and reps in USA or Cong and BJP in India have different views on economy and this reflects on their action. Moreover, their response and views are not consistent which is fine as Governments have to respond to current situation and people and not to their ideologies. Humans can judge these situations better than AI can.
The frequency of such events and even rarest of the rare events has increased in recent times.
There are many areas where AI has been shown to do better than humans. Areas like law, management, supply chain, logistics planning have shown great use of AI in increasing efficiency and building competitive advantage. But these will be for next article if at all. Human’s attention span is not more than 5 minutes. So going beyond that is meaningless.
In short, everything that your mind can do can potentially be replaced by AI or someone with AI. Irrespective of the outcome, being humans in the age of AI is challenging. This is similar to what any technology brings to the table. It makes the future unknown because we cannot see beyond few quarters in future. After a few quarters, we can't even see the present. But human ingenuity has often overcome the challenges and that’s why despite hundreds of wars, disruptions, viruses, pandemics, and recession, we are thriving and progressing. The arrival of AI will not change that.
Rajnikant Sir knew this. That’s why he never used guns or weapons to fight villains, he used good old hands. In the end he would throw away all the weapons and use his raw skills to kill enemies. He knew AI will replace all of them but not the mighty hands. Not only that, the hands will control AI and not the vice versa. He showed the way, now we have to follow it. Like he said in one of his movies “Naan oru dhadavai sonna, nooru dhadavai sonna maadiri”… translated for lesser humans… “If I said it once, I said it 100 times”.
References
[1] JAMA Network Open. "Diagnostic Accuracy of Deep Learning for Detection of Diabetic Retinopathy in Retinal Fundus Photographs." Accessed May 4, 2025. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/28253.
[2] Franck, Dan. "Google's DeepMind AI Beats Doctors in Breast Cancer Screening Trial." CNBC, January 2, 2020. https://www.cnbc.com/2020/01/02/googles-deepmind-ai-beats-doctors-in-breast-cancer-screening-trial.html.
[3] University of Cambridge. "Artificial Intelligence Beats Doctors in Accurately Assessing Eye Problems." Last modified February 14, 2025. https://www.cam.ac.uk/research/news/artificial-intelligence-beats-doctors-in-accurately-assessing-eye-problems.
[4] Nasdaq. "You Can Outperform 88% Professional Fund Managers Using Simple Investment Strategy." Accessed May 4, 2025. https://www.nasdaq.com/articles/you-can-outperform-88-professional-fund-managers-using-simple-investment-strategy.
[5] Ferri, Rick. "Any Monkey Can Beat the Market." Forbes, December 20, 2012. https://www.forbes.com/sites/rickferri/2012/12/20/any-monkey-can-beat-the-market/.
[6] Japan Times. "Hedge Fund Startup AI." February 14, 2025. https://www.japantimes.co.jp/business/2025/02/14/markets/hedge-fund-startup-ai/.