Table of Contents
The brain is the most complex organ in the human body. And no matter how much we convince ourselves otherwise, we still do not understand exactly how it works. Even holders of a degree in neuroscience admit that our knowledge in this area is limited. Fortunately, this does not prevent scientists from working on the creation of a neural network – a tool capable of self-learning and designed in the image and likeness of the human brain.
The fact is that both the brain and the neural network develop according to similar physical principles (like the systems of galaxies in the observable Universe), and the neural network not only recognizes faces, diagnoses diseases and draws pictures, but can, by analyzing information, predict future events. Recently, a TikTok user asked the neural network to show what they think the last photos of people before the end of the world will be like. The resulting images paint a chilling picture – natural disasters, burning cities and … zombies.
“And I remembered the Fourteenth volume of the works of Bokonon – last night I read it all in its entirety. The fourteenth volume is titled: “Can a reasonable person, considering the experience of past centuries, harbor even the slightest hope for a bright future for mankind?” Read the Fourteenth volume for a short time. It consists of just one word and a period: “No.”
Kurt Vonnegut “Cat’s Cradle”
Brain and neural network
Despite significant advances in the study of the brain, much about how it works remains a mystery. Yes, the functioning of individual cells has been explained quite well, but understanding of the process by which the interaction of thousands and millions of neurons allows the brain to work as a whole is largely lacking and requires further research.
Recall that the neurons of the brain are interconnected by synapses and transmit electrochemical impulses that ensure the functioning of the whole organism. The neural network, in turn, consists of artificial neurons, and it is based on the results of scientific research on how the brain strengthens neural connections. So, when we learn new information or acquire skills, neurons connect with each other, forming new connections.
Today, this computer program (neural network) is able to learn from the information provided and solve some types of problems. Researchers refer models of artificial neural networks to machine learning, and those, in turn, to artificial intelligence (AI).
How a neural network works
Since machine learning allows the program to work on its own, expanding its own knowledge of the world, there are many algorithms capable of deep learning these days. While the basic unit of the brain is the neuron, the basic building block of an artificial neural network is the perceptron , which performs simple signal processing and then connects them into a larger, unified network.
A common example of such tasks is the recognition of objects of a certain type, presented in large numbers. By analyzing the input data, neural networks learn to create and classify images, extracting them from all corners of the World Wide Web (including based on user text messages).
And in 2020, neural networks helped scientists answer the question of why we dream.
DALL-E and the apocalypse
In January 2021, the DALL-E neural network, named after the Spanish surrealist artist Salvador Dali, learned how to turn text into images and understands strange queries well. It works like this: a person describes a situation or an object, and the algorithm creates an image that is often difficult to distinguish from human hands.
Over the course of its work, the DALL-E neural network has managed to create an image of a radish, a manikin, and even an emoji with a baby penguin. Through a process called diffusion, a set of random dots gradually changes to recognize certain aspects of the image. The researchers also intend to train the neural network to recognize speech, textual data, and visual images. However, the data entry process is limited and extremely time consuming.
Since the neural network accepts abstract and unusual requests, a TikTok social network user asked DALL-E to show what the “last selfie on Earth” would look like. The result exceeded expectations: the neural network considered that the end of the world would be terrible.
Each image shows a man with a phone in his hands, behind which you can see burning cities, raging tornadoes and bombings. And there are a lot of zombies hanging around in the middle of the destruction.
According to NBC News, DALL-E over time, the neural network will be able to improve image generation. But there is a small problem – the apocalyptic paintings performed by AI are filled with racial and gender bias. So, if you ask the algorithm to depict a flight attendant, DALL-E gives out images of women, and by typing CEO (general director), only white men will appear in front of your eyes. All in all, it was awkward.
In the future, new methods will help DALL-E create a variety of images, OpenAI announced. Well, today machine learning algorithms respond to such requests as “unknown species of animals”, “human mutants” and “edge of the universe”.
It is impossible not to note the artistic skills of AI – the resulting paintings look depressing. Disfigured people with huge eyes and elongated fingers are in every drawing. Ultimately, over 14,000 users found the images truly scary. The images were shared by “Robot Overloads” on TikTok, which regularly posts AI-generated images based on user text prompts.
Moreover, the DALL-E neural network regularly hits the top news. To date, this AI has mastered the skills of image editing, working with Google’s online translator, and even made a cover for a glossy magazine. Earlier in 2022, the developers announced a new version of DALL-E. At the time of writing, access to it is limited, but in the near future everyone will be able to ask their questions. Will wait!