Using Artificial Intelligence in Finance
More and more people are predicting that AI will soon automate all jobs. How much truth is there to this? If we look at...
Read MoreBlog
Types of Artificial Intelligence
The term artificial intelligence (AI) is often used regardless of the type of AI in question or to refer to a subcategory like Machine Learning (ML). What differences exist and why...
Read More
Big Data
30% of the data we have access to is being generated in real time, are we over-informed? Is there more data available than we can consume? I don't think so, I explain it below...
Read MoreStrategy
I observe with some concern the hype that is being given to artificial intelligence within companies, especially from a business strategy perspective...
Read More
×
Using Artificial Intelligence in Finance
More and more people are predicting that AI will soon automate all jobs.
How much truth is there to this?
If we look at what has happened so far, algorithms based on artificial intelligence have indeed been able to be better in closed and static systems applied to a specific task, with frankly impressive results in many cases. However, in dynamic systems like those we find in financial markets, the track record of artificial intelligence has been more modest so far. The fact is that beating the famous "wisdom of crowds" is one of the most difficult problems.
Does this mean that artificial intelligence is not useful?
Clearly not. In fact, quite the opposite. Artificial intelligence is a key tool and will continue to be. Among other things, it allows us to expand the range of quantitative models, process data much faster, and mitigate behavioral biases that we humans suffer from.
What will AI be like in the future?
The combination of artificial intelligence and human intelligence (better at inferring, making analogies, relating, and interpreting cause-and-effect relationships) will be key for developing algorithms that allow us to reduce uncertainty.
In this sense, the combined experience of artificial and human intelligence will be necessary and key in the future.
What do you think?
×
Types of Artificial Intelligence
The term artificial intelligence (AI) is often used regardless of the type of AI in question or to refer to a subcategory like Machine Learning (ML).
What differences exist and why is it important to keep them in mind?
Artificial intelligence focuses on developing machines capable of performing tasks or solving problems as a human would, and here we can distinguish between two types: narrow artificial intelligence and general artificial intelligence. The first focuses on performing a specific task while the second is more ambitious, where the machine would be capable of performing any task like a human. In this sense, it would have the ability to learn, reason, understand, adapt to different environments, and even feel like a human. On the other hand, Machine Learning is a form of artificial intelligence that encompasses algorithms that learn through examples and experiences (they learn by themselves).
Why is this distinction important?
So far, everything that has been developed, even tools as sophisticated and impactful as ChatGPT, DALL-E, Sora, or Gemini would fall within the category of narrow artificial intelligence, and in my opinion are still far from resembling something similar to general artificial intelligence.
And it is undeniable that, just as the Industrial Revolution did, the adoption of artificial intelligence will transform work.
Should we then be afraid of the irruption of artificial intelligence?
My opinion is no. Machines are far from being able to encompass the tasks a human professional does, since we are still talking about narrow artificial intelligence. In this sense, I very much doubt that artificial intelligence can replace professionals in the short/medium term, or at least in the financial field. But what I do think could happen is that professionals who leverage artificial intelligence end up being more productive and could have a comparative advantage over professionals who don't use it.
×
Big Data
30% of the data we have access to is being generated in real time, are we over-informed? Is there more data available than we can consume? I don't think so, I explain it below.
➡️ Today we live in the Big Data era. Without data, artificial intelligence could not have experienced the acceleration we have witnessed. In fact, much of the algorithms and artificial intelligence techniques were developed decades ago, but only now are they bearing fruit because they are leveraging Big Data.
📈 The growth of data has been exponential, with 90% of data created in the last two years. For next year it is estimated that we will reach 175 zettabytes; to put it in perspective, imagine we represent the Digital Universe with 128GB tablets like the ones we have at home. Well, 175 zettabytes are equivalent to 25 columns of those tablets stacked, each column with a height equivalent to the distance from Earth to the Moon, which is 384,400 km.
📖 So, how can we consume all this information? What works best for me is following some guidelines:
1️⃣ Trying to extract data that has value: keep in mind that it is estimated that only 0.5% of data is currently being analyzed. We have a long way to go, but data quality is also important, that the data is truthful. In the financial world it is also relevant that it is point-in-time; that is, that we use the data that was available at each moment in time when backtesting our models (for example, inflation or GDP series are published with a delay and in many cases are revised; that revised series I didn't have in the past, so I couldn't have made a decision based on it)
2️⃣ Expanding the spectrum of data: we tend to use only structured data, but only about 20% of it is (Excels to understand), with the rest being either semi-structured (HTML) or unstructured (social networks, satellite images, etc.). It's true that in many cases structured data is what works, but resorting to, for example, the sentiment that exists in social networks is usually very useful and allows you to anticipate trends
3️⃣ Using new tools: there is a tendency to use linear models as prediction tools (linear regression, for example), but the reality is that relationships in financial data are usually not linear. Linear models in this sense are not useful in many cases because they can be interpreted better, but when the relationships are not linear it is important to use models that don't force them.
✅ By way of conclusion, the Big Data era has been a great milestone and has allowed many of the artificial intelligence techniques that were envisioned decades ago to become reality today. Knowing how to make use of all this amount of data is important, but you must follow a method to avoid falling into the "garbage in garbage out" trap (no matter how good our model is, if the input data doesn't have enough quality our output won't be useful).
×
Strategy
I observe with some concern the hype that is being given to artificial intelligence within companies, especially from a business strategy perspective.
Is artificial intelligence a strategy? The answer is NO.
➡️ Strategy has to do with being innovative, different, while deploying artificial intelligence without a clear objective, a deep understanding of the environment, and an objective assessment of available resources will only serve in the best of cases to improve operational efficiency (being faster/more productive). In this sense, artificial intelligence should not be the end, but the means. If we really want to get the most out of it, it's no good adapting the use case to artificial intelligence, but rather the other way around, adapting artificial intelligence to the use case; it's no good deploying just for the sake of it. We need more compass and less stopwatch.
📖 So, how can we develop a successful strategy? Below is what has worked for me and following business strategy guidelines:
1️⃣ Simple long-term objectives: I want to leverage artificial intelligence as a predictive tool in financial markets and at the company level to improve my decision-making or create semi-automatic decisions and know what worries investors most
2️⃣ Deep knowledge of the competitive environment: there is no prediction tool for exchange rates and interest rates that I need, so the previous objective makes sense. At the company level there are some tools that I can adapt
3️⃣ Objective assessment of resources: I have access to data and my team has great knowledge of algorithms, but we need greater computational capacity for the level of model complexity. In this sense, we resorted to cloud computing to have access to greater capacity
✅ By way of conclusion, be careful with deploying artificial intelligence without a clear strategy. If all we want is to improve productivity (which is fine) perhaps it makes sense to give teams access to Copilots and Bots that allow them to be more agile. But if what we want is to design a strategy, we have to be clear about the objective and whether artificial intelligence is useful for the specific use case we have in mind.