Is AI better than Humans now?
The answer is Yes and NO.
On the picture below you can see a model of the biological neuron. In a simplified form, the work process of a neuron can be represented as follows: neuron receives a signal at the input, process the signal, pass it through, and give modified signal at the output, to other neurons. If you want more info about neurons you can find it here.
What is really important for us now is that inside human brains there are approximately eighty-six billion neurons. Each neuron has on average 7000 synaptic connections to other neurons. Quite similar to a giant network of neurons.
And what if we could create an artificial neuron that serves the same function as humans does? Well, actually, we can create one, just like on the picture below.
And what if we could unite thousands and millions of artificial neurons into the network? Then we could possibly get a network of artificial neurons just like in humans brains. Is such a network will be Artificial Intelligence? Not yet. First, somehow we need to train that network to give us the expected output signal on a certain input. And just at this point machine learning begins. After such an artificial network of neurons has been trained we can use it as an AI but only for the task, it has been trained for.
So now you can see that artificial brains are quite similar to biological ones. They work on the same principles and can have same structure. And I will ask you a question: having artificial brains, can we program them for the task of financial statements analysis? Yes, we can.
New Model generation
There is one problem with Artificial Intelligence (AI for short). It cant program itself yet. And this problem derives from lack of motivation. I mean that AI doesn’t have any motivation, unlike humans. There are many theories about motivations but I do prefer neuroscience theory but you can find broad info here.
So AI has to be “guided” or supervised by humans to learn something. Humans give their motivation to AI and then it works. And here we need a group of human specialists who can design models architecture for financial analysis, stocks price prediction and portfolio investments.
The Cognitive Bias
Here we come to the bias problem. Amos Tversky and Daniel Kahneman did a great job introducing the notion of cognitive biases to the world. Because of cognitive bias, humans tend to deviate from rational choice in the decision-making process. You can read more about cognitive bias problem here.
AI, unlike humans, has an easily manageable bias. AI supervisor can tweak it as he wants and any time he wishes to. That is a very good advantage of AI.
Errors in the Model
Imagine that a human has learned a way to do something in his life, for instance brushing his teeth or shopping, or cooking. He or she will use this way without any changes to the end. It is a great luck if this way leads to success and maximizes the result. But what if this way leads to a mediocre outcome and the particular human is fine with it? Well, he won’t change his behavior to the end unless he experiences very stressful pressure for a long time or life-threatening circumstances. And those two last things are very rare unless made artificially. So what does this means? It means that if a common person has made a mistake he will constantly repeat it to the end.
Another bad thing is that most of the humans don’t even understand that they are making mistakes. Usually, they prefer to deny mistakes instead of correcting them especially when someone from the side points at those mistakes. The reason for such outcome is that once or maybe several times in previous human life a certain behavioral pattern was successful and human’s brains have learned this, so under the similar circumstances, human brains use the same pattern. And a process of changing conditions of behavioral patterns usage inside human brains requires a huge amount of energy, indeed he has to rebuild a lot of neural connections. And human’s brains seek energy economy thanks to the evolution. That is why changes in human behavior are possible under very stressful pressure for a long period of time or under life-threatening circumstances.
As for AI, he doesn’t care for the economy either doesn’t deny own mistakes and can constantly seek better solutions guided by human motivation.
Physical resources of humans are limited,
In tasks such as evaluating huge amounts of companies financial statements data AI is already better. In building new models for companies assessment humans are still better for now. To generalize AI is more accurate, fast and cheap in routine simple and repetitive tasks especially where you can scale up. In our case, it is predictions of equities prices based on companies financial statements, technical analysis of shares prices, and social sentiment analysis. So we can expand the use of AI in portfolio investments, stocks trading.