Discover Geeks Academy’s articles on: Blockchain, Coding, Cybersecurity, Cloud, Big Data, Artificial Intelligence, Gaming, Digital Innovation
The world of new technologies is constantly developing. A few calendar months within industries such as Web3, blockchain and cybersecurity seem like an eternity: improvements, updates and upgrades are on the agenda and progress and development are limitless. However, the most talked about field lately will be the 2023 protagonist. We are talking about artificial intelligence, specifically a part of AI which is attracting experts, students and enthusiasts.
Generative AI
What is the difference between artificial intelligence (AI) and generative artificial intelligence (Gen-AI)? Artificial Intelligence is a broad definition that refers to all those technologies capable of intelligent behavior. By this we mean simple algorithms that can sort data up to more advanced systems that can mimic human-like computational and thought processes.
On the other hand, generative AI is a specific part of artificial intelligence focused on generating new content, such as text, images or music. Such intelligent systems are trained on large datasets and use machine learning algorithms to generate new content similar to the training datasets. The use cases are varied, such as creating art, music or even generating text for chatbots. These achievements in AI are generally reached through the use of machine learning (ML), particularly thanks to a specific method.
GAN
A generative adversarial network – more commonly known as GAN, generative adversarial network – is a class of machine learning methods, in which two (or more) neural networks are trained with large datasets and, over time, learn to generate similar content but original based on the training datasets. GANs are just one way to create generative models, but most of today's most popular applications are based on or inspired by GANs.
Let's take a look at the most promising applications.
Text-to-Image
It is certainly the most discussed use case at the moment. Thanks to image generators, we can create masterpieces with simple text instructions (usually called prompts). How is it possible? The algorithms have been trained with millions of images, paintings, drawings, starting from contemporary art up to styles dating back centuries. For example, we might ask the AI to generate “a castle on a hill” and within seconds we will have a graphical representation of a castle on a hill:
This image has been generated by Dall-E 2, an AI algorithm introduced in early 2021 by OpenAI, a non-profit artificial intelligence research organization founded by Sam Altman and Elon Musk. In this case, we have not asked the algorithm for specific details or characteristics, such as a specific style. If our prompt is more detailed as “a castle on a hill in the style of Hieronymus Bosch”, then Dall-E 2 creates an image that might be closer to the style of the famous Flemish painter.
Let’s notice how the colors of the new representation are darker, in line with Bosch's style.
Dall-E is just one of many image generators on the market. For some months, other services such as Nightcafé or Midjourney have established themselves among the most trending text-to-image generators. We have been infected as well by the trend indeed (today's cover has been generated by an AI).
Despite recent debates about the ethics of Gen-AI, many artists have compared the innovation of generative art to the birth of photography. Every great innovative discovery is greeted with doubts and uncertainties; however, beyond any criticism, we know that this has been only the beginning and we will hear even more about generative art in 2023.
Music Generator
While they haven't yet had the same impact as text-to-image applications, music generators are on the rise. With a quick Google search, you might discover dozens of mobile or browser-based applications. Music generators basically work as the image generators. The algorithms are trained with large datasets, containing hours and hours of music from every genre and music era. By giving instructions to the AI, the application will generate an original music track. Like the image generators, these tools can be useful to professional musicians themselves: for instance, I have a funky rhythm in mind but I can't find the main idea that will drive my creative process. I can pick one of the available services, such as Amper Music or Amadeus Code, ask the algorithm for a funky rhythm and in a bunch of minutes I have got enough ideas to start my new music track.
Chat GPT-3
Chronologically, this is one of the latest models to attract media attention. What is Chat GPT-3? It is a prototype chatbot based on artificial intelligence. Through machine learning and deep learning, OpenAI, the same organization behind Dall-E, has created an AI capable of speaking in a very similar way to a human being. Chat GPT-3 is based on GPT-3, the third generation of the Chat Generative Pre-trained Transformer, a linguistic model that is based on 175 billion parameters. Thanks to its vast knowledge, the chatbot is already widely used to get specific information on some topics, generate specialized articles and write code in various programming languages. Its usefulness and potential is yet to be discovered and even a large giant like Google is developing its linguistic models thanks to the LaMDA project, Language Model for Dialogue Applications, a powerful algorithm for understanding natural language.
In short, scientific research never stops and 2023 could be the year of artificial intelligence.
The future is AI
During the past years, the total amount of data created has soared, mostly due to the expansion of the Internet of Things (IoT). The data market is therefore growing dramatically and, according to the latest estimates, is going to reach a value of $103 billion by 2027. Here are some figures about the world of data:
- Businesses generate approximately 2,000,000,000,000,000,000 bytes of data per day.
- 97.2% of companies invest in AI and Big Data.
While the numbers speak for themselves, companies struggle to keep up with the ceaseless creation of new data:
- About 95% of companies report an inability to understand and manage unstructured data.
- Only about 26% of companies say they have achieved a data-driven culture.
Don't live the future as a sidekick... be a superhero! Discover Geeks Academy’s training offer in AI & Big Data:
Sources:
https://blog.google/technology/ai/lamda/
https://openai.com/dall-e-2/
https://creator.nightcafe.studio/
https://chat.openai.com/chat