This is blog 14 in the blog series about Green IT.
Since the launch of Chat GPT in late 2021, use of this chatbot has skyrocketed. No less than 100 million users registered in the first two months after the launch, giving a huge boost to the use of artificial intelligence. Microsoft and Alfabet (Google’s parent company) quickly followed with similar chatbots: Bing Chat and Bard. While the potential of AI seems boundless, it comes with considerable ecological impact.
In the next four blogs we will explore the world of AI, see how this technology is applied in the heritage sector, and how huge its impact is on our energy use and the corresponding CO2 emission. We will also examine how we can reduce this energy use and which important questions we need to ask when using AI.
What is it?
The notion of ‘artificial intelligence’ was first introduced by the American computer scientist John McCarthy in 1956. He defined artificial intelligence as machine behaviour that resembles the ‘intelligent’ behaviour of humans.
AI comprises a wide range of technologies and methods that enable a machine to show ‘intelligent’ behaviour. The main goal is to have machines perform tasks that normally only people can do. This includes things like learning, reasoning, problem solving, understanding human language, and observing the environment.
Narrow and general AI
There are two main types of artificial intelligence:
(1) Narrow AI: these AI systems are designed and trained to perform specific tasks. They can only perform the tasks for which they have been programmed. Consider things like language recognition, image recognition and chatbots.
(2) General AI: these are systems that are still mainly theoretical at present. These systems would, like humans, have self-awareness, be able to reason and solve problems. The example that most approximates this ideal is the robot called Sophia, produced by the Chinese company Hanson Robotics. In the promotional video, you can see how much Sophia and her family already display human capabilities.
In certain areas, AI is far more capable of performing tasks than humans, particularly where it concerns repetitive and detail-oriented tasks such as analysing large numbers of legal documents, or thousands of X-ray images in a hospital. It can do so very rapidly and with a relatively low error margin. Thanks to the vast amounts of data that AI can process, we can understand information in a very short time. For instance, we can discover trends or find results that were previously unknown or that we would otherwise not have been able to discover so quickly.
How does artificial intelligence work?
Generally speaking, AI systems work by processing huge amounts of labelled training data. These data are analysed to discover patterns, which then enables AI to make predictions. It’s all about pattern recognition, and this can be applied in various ways.
For example, a chatbot’s algorithm learns how sentences are composed on the basis of lots of different text samples, and which words tend to logically combine together. With this ‘knowledge’, a chatbot can generate a text within seconds in response to a ‘prompt’ (that’s the name for your instruction to a chatbot).
Another much-used application is image recognition. Here, an algorithm learns to identify and describe images by viewing a huge amount of images. By processing ever more images, the algorithm becomes increasingly better at recognising what an image shows.
The latest generation of AI tools is even able to create texts, images, videos and music. This emerging generation of AI tools will have a major impact on various fields including education, marketing, product design, and also the heritage sector.
In the next blog we will take a further look at how AI is currently used in the heritage sector. In addition, we will discuss the environmental effects of the large-scale use of AI. The final blog of this series examine the possibilities for ‘green AI’ to mitigate the negative effects of AI use.
Sources
Laskowski, Nicole, Definition: artificial intelligence’, published on TechTarget.
NPO Kennis, ‘Wat is artificial intelligence (AI)?, last viewed on 21 November 2023.
Rijksdienst Digitale Infrastructuur, ‘Wat is kunstmatige intelligentie (AI)?, last viewed on 21 November 2023.