Generative artificial intelligence (GAI) started making waves a few years ago with the release of systems such as ChatGPT and DALL-E. They are able to produce sophisticated and human-like text, code, or images after the models powering them are trained on large quantities of data. However, it soon became apparent that the specific phrasing of a question or statement input by a user, known as a prompt, had an impact on the quality of the resulting output. "It's a way of unlocking different capabilities from these models," says Andrei Muresanu, an AI researcher at Vector Institute in Toronto, Canada. "If you tell ChatGPT to pretend that it's a professor of mathematics, it will do better on math questions than if you just say, 'answer this question' or'pretend you're a student'." Coming up with prompts that steer a model towards a desired output has emerged as a relatively new profession, called prompt engineering, to help achieve more relevant and accurate results.

ChatGPT exploded into the world in the fall of 2022, sparking a race toward ever more advanced artificial intelligence: GPT-4, Anthropic's Claude, Google Gemini, and so many others. But with every passing month, tech corporations appear more and more stuck, competing over millimeters of progress. The most advanced and attention-grabbing AI models, having consumed most of the text and images available on the internet, are running out of training data, their most precious resource. This, along with the costly and slow process of using human evaluators to develop these systems, has stymied the technology's growth, leading to iterative updates rather than massive paradigm shifts. As researchers are left trying to wring water from stone, they are exploring a new avenue to advance their products: They're using machines to train machines.

In the beginning, the chatbots and their ilk fed on the human-made internet. Various generative-AI models of the sort that power ChatGPT got their start by devouring data from sites including Wikipedia, Getty, and Scribd. They consumed text, images, and other content, learning through algorithmic digestion their flavors and texture, which ingredients go well together and which do not, in order to concoct their own art and writing. Generative AI is utterly reliant on the sustenance it gets from the web: Computers mime intelligence by processing almost unfathomable amounts of data and deriving patterns from them. ChatGPT can write a passable high-school essay because it has read libraries' worth of digitized books and articles, while DALL-E 2 can produce Picasso-esque images because it has analyzed something like the entire trajectory of art history.

Artificial intelligences that are trained using text and images from other AIs, which have themselves been trained on AI outputs, could eventually become functionally useless. AIs such as ChatGPT, known as large language models (LLMs), use vast repositories of human-written text from the internet to create a statistical model of human language, so that they can predict which words are most likely to come next in a sentence. Since they have been available, the internet has become awash with AI-generated text, but the effect this will have on future AIs is unclear. Now, Ilia Shumailov at the University of Oxford and his colleagues have found that AI models trained using the outputs of other AIs become heavily biased, overly simple and disconnected from reality – a problem they call model collapse. This failure happens because of the way that AI models statistically represent text.

Smart speakers like Google Home and Amazon Alexa could be used by criminals to listen to and decipher a password or PIN being typed in on a nearby phone. Researchers from the University of Cambridge built their own version of a smart speaker to closely resemble those which are commercially available. Sound recordings from the gadget were inputted into a computer for analysis and experts investigated if the sound and vibrations caused by typing on a smartphone screen could be used to guess a five-digit passcode. When the phone was placed within 20cm (7.8inches) of the custom-built device, the computer was able to guess the code with 76 per cent accuracy in three attempts. This graphic outlines the general flow of the experiment.