In the fourth episode of Inside Alana Podcast, we will discover how some areas of Biology positively influence the development of artificial intelligence technologies.
Despite appearing to be totally different subjects, Biology studies some aspects that help in understanding how the human brain works and evolves. This capacity for evolution is what inspires scholars around the world in the search for a Strong AI, and is the main influence in the creation of genetic algorithms.
Evolution theory and genetic algorithms
Charles Darwin was not a programmer or statistician, but he played an important role in the innovation of artificial intelligence techniques.
Without imagining that his theory of evolution could help technology, he claimed that natural selection was an essential evolutionary principle since better-adapted species would have a better chance of survival. Based on this thinking, AI scholars applied rules in the form of an evolutionary algorithm, whose main positive metric is adaptation.
Evolutionary algorithms are called genetic algorithms, and they are programmed to follow patterns that replicate the theory of the evolution of machines. This process works as follows:
- Algorithms are created and put to the test;
- A small group will survive longer, which can be considered as a natural selection;
- The surviving group will undergo recombination (crossing over);
- The process is repeated so that the algorithms continue to evolve and thus become stronger.
The technique of genetic algorithms was developed mainly by John Henry Holland, in the 1970s and by some of his students, such as David E. Goldberg, who popularized the topic in the 1980s.
Among the main benefits of genetic algorithms are:
- They are easy to implement;
- They are highly adaptable;
- They work with continuous parameters;
- They can be used in conjunction with other techniques
According to Marcellus Amadeus, CTO of Alana AI, besides the theory of evolution, there is another area of Biology of great importance for artificial intelligence: neuroanatomy.
Neuroanatomy and Deep Learning
As neuroanatomy studies the nervous system, which is composed of neural networks, it helps to understand facts for the evolution of artificial neural network algorithms, which imitate the biological structure of the brain and are the basis for the deep learning technique (Deep Learning).
As mentioned in the podcast episode about how machines think, artificial neural networks are like systems of interconnected neurons, and they process input values, simulating behaviors of biological neural networks.
Deep learning and its peculiarities
The most amazing thing about the Deep Learning technique is that, in most cases, it is not possible to know how the machine came to a certain conclusion. This is called the black box and, for sectors that need to understand the behavior of the machine, it can be a problem.
On the other hand, there are Explainable Artificial Intelligence (Explainable AI), also known as XAI.
This area of study covers simpler and older techniques, such as decision trees. The great challenge in this field is to explain how Deep Learning learns, that is, to make it not a black box algorithm.
Deep Learning Algorithm Types
There are several Deep Learning algorithms, and it is possible to use several simultaneously since each is ideal for a specific type of activity.
The predominant techniques are:
- Convolutional Neural Network (CNN);
- Recurrent Neural Network (RNN);
- Generative Adversarial Network (GAN);
- Long Short-Term Memory (LSTM);
LSTM is one of the most common in the area of machine language development. It is related to memory, both short term, and long term.
This algorithm (LSTM) is a recurring neural network architecture that “remembers” values at random intervals. It is suitable for classifying, processing, and forecasting time series with unknown time intervals.
Some of the main, and most famous, applications of LSTMs include:
- Language Modeling;
- Language Translation;
- Captions in Images;
- Text Composition;
- Spelling Recognition;
- Rhythm Learning.
Artificial intelligence is multidisciplinary
The link between AI and other areas is fantastic and, at the same time, intriguing. In addition to biology, other fields of impact are statistics, mathematics, and linguistics.
We know that it is a difficult task to create a machine with full awareness in the coming decades or centuries, but the study of these varied areas, for sure, will continue to contribute to the development of artificial intelligence technologies.
To learn more about the basics of AI, we recommend that you listen to previous episodes of the Inside Alana Podcast.