Название: More than a Chatbot: Language Models Demystified
Автор: Mascha Kurpicz-Briki
Издательство: Springer
Год: 2023
Страниц: 133
Язык: английский
Формат: pdf (true), epub
Размер: 11.8 MB

Did you wonder how tools like ChatGPT or Bard work, but do not have a technical background? Then this is the right book for you! To discuss and better understand what such technologies are capable of and how this will impact our lives and our society, a basic background knowledge about text processing and generation technologies is required. In particular, the book discusses the following questions: How did the field of automated text processing and generation evolve over the last years, and what happened to allow the incredible recent advances? Are chatbots such as ChatGPT or Bard truly understanding humans? What pitfalls exist and how are stereotypes of the society reflected in such models? What is the potential of such technology, and how will the digital society of the future look like in terms of human-chatbot-collaboration?

Recent technological advances have resulted in a new generation of powerful text processing and generation tools, that produce text very similar to what a human could do. Such tools have the potential to change the way we work, learn and teach – and this book enables you to understand and be part of this exciting development.

The latest achievements in the field of Artificial Intelligence (AI) pose many challenges to society. For the broad public, including researchers and professionals from domains outside of AI, the border between what is the technically possible and science fiction is getting blurrier. The term artificial intelligence is a rather generic term. It has many inequivalent interpretations, and the conflation of these leads to exaggerated expectations and assumptions about how AI technology works. The main approach behind the technologies referred to as Artificial Intelligence nowadays is Machine Learning (ML). Machine Learning enables a computer to learn from examples for the purpose of solving tasks on similar data that it has not seen before. Machine Learning can be applied in different fields, and the examples from which the computer learns can be of different data types, such as images, texts, videos, or a conventional data table.

The field of natural language processing (NLP) investigates interactions between computers and human language, including the development of methods for the automatic analysis or generation of text. Different methods have been employed in the past to achieve this, and nowadays, this sort of processing benefits leans heavily on advances in Machine Learning. The meeting of machine learning and NLP has given rise to language models. These are statistical models that aim to describe patterns in written language and can be used, among other things, for text generation. Language models are what’s behind the curtain in some of the latest hyped products, including ChatGPT.

Whereas historically different methods were used to build AI systems, nowadays the term Artificial Intelligence has become almost synonymous with machine learning. In the Chapter 2, we will dive deeper into this topic and understand better how and in what sense machines are capable of learning. To do so, we will introduce some analogies and terminology that will follow us throughout the book when looking at technical aspects of the text processing technologies we will eventually discuss. We begin with the key differences between traditional computer programming and machine learning. Next, we will look at the different subdomains of Machine Learning, namely, supervised, unsupervised, and reinforcement learning.

The book is aimed for a general audience, briefly explaining mathematical or technical background when necessary. After having read this book, you will be confident to participate in public discussions about how this new generation of language models will impact society. You will be aware of the risks and pitfalls these technologies can bring along, and how to deal responsibly when making use of tools built from AI technology in general.




ОТСУТСТВУЕТ ССЫЛКА/ НЕ РАБОЧАЯ ССЫЛКА ЕСТЬ РЕШЕНИЕ, ПИШИМ СЮДА!