AI – Everything you want to know about
Artificial intelligence (AI) is a term that has been around almost as long as electronic computers themselves, having been coined back in 1955 by a team including legendary Harvard computer scientist Marvin Minsky. While AI is already present in some forms in our daily lives, such as in the special effects used in some films or in voice assistants like Amazon’s Alexa, in the current debate it has come to mean something much more complex.
At its core, AI involves training computers to learn in a way that imitates human behavior. While traditional computers simply follow instructions given to them in the form of code, to tackle more complex tasks computers need to be taught how to analyze information and draw inferences from patterns within datasets. The most successful versions of machine learning in recent years have used a system known as a neural network, which is modeled on how we think a brain works.
However, with no strict definition of the phrase, and the promise of billions of dollars of funding for anyone who incorporates AI into their pitch documents, almost anything more complex than a calculator has been called artificial intelligence by someone. As such, there is no easy categorization of AI and the field is growing so quickly that new approaches are being uncovered every month.
Some of the main types of AI that you may come across include reinforcement learning, large-language models, generative adversarial networks (GANs), and symbolic AI.
- Reinforcement learning is a type of artificial intelligence training that involves providing feedback to a system each time it performs a task, allowing it to learn from its successes and failures. This type of learning can be slow and resource-intensive, but it can be particularly effective for systems that interact with the real world. In such cases, reinforcement learning may be the best way to train the system to behave appropriately in complex, unpredictable environments. For example, autonomous vehicles may use reinforcement learning to learn how to navigate busy streets and avoid accidents.
- Large-language models, also known as LLMs, are a type of neural network used in natural language processing. These models are trained by inputting vast amounts of text data, such as books, articles, social media posts, and other sources of written or spoken language. The LLMs analyze this text data and use it to predict and generate new words and sentences in a certain sequence. The more text data that is fed into the LLMs, the more accurate they become in generating human-like language. LLMs are widely used in a variety of applications, including chatbots, language translation, and text summarization. They are also used in cutting-edge research areas such as generative art and writing.
- Generative adversarial networks (GANs) are a type of neural network that consists of two networks: a generator network and a discriminator network. The generator network is responsible for creating new data, such as images, music, or text, while the discriminator network evaluates the generated data and provides feedback to the generator network. The two networks are trained together in a process called adversarial training, in which the generator network learns to produce data that is increasingly difficult for the discriminator network to distinguish from real data. This process continues until the generator network can produce data that is indistinguishable from real data. GANs are often used in creative work, such as music composition, visual art, or film-making, where the generator network is given the role of creator while the discriminator network plays the role of a critic. The generator network learns to create things that the discriminator network will approve of, resulting in the production of new and original creative works. GANs have also been used in other applications such as image and speech recognition.
- Symbolic AI is an AI technique that takes inspiration from past approaches and rejects the idea that a simple neural network is always the best option. Instead, it seeks to combine machine learning with more structured information about the world. This approach involves creating a set of rules or symbols that represent knowledge, which can then be used to reason and make decisions. Symbolic AI has been used in a variety of applications, including natural language processing, robotics, and expert systems. While it has been overshadowed in recent years by more data-driven approaches like neural networks, it continues to be an important area of research in the AI field.
