post-image

Insights to Inspire / AI

Did you know that AI-driven text generators are already accurate enough to do children’s homework?

author

Nicolás Moreira

Some days ago, a teenager posted a message on Reddit that caught the attention of quite a portion of the AI community: she’s been using an AI text generator to write essays for school and to do her homework. Thanks to this idea, not only has she received straight A’s, but also she’s turning a […]

Did you know that AI-driven text generators are already accurate enough to do children’s homework?

Some days ago, a teenager posted a message on Reddit that caught the attention of quite a portion of the AI community: she’s been using an AI text generator to write essays for school and to do her homework. Thanks to this idea, not only has she received straight A’s, but also she’s turning a profit: by the time she posted her thread online, she had already made USD 100 “writing” other classmates’ homework.

It’s unclear what is the model and platform she’s using, but there is no question that the maturity of the algorithms is surprising. After the re-introduction of neural networks, deep learning techniques have been massively adopted for automating tasks in many domains and are consistently improving their skills for text generation.

One of the major discoveries in this field has been the exploitation of attention mechanisms, which are the core of recent Transformer models. Based on this neural network architecture and by mining from enormous text databases such as Wikipedia or BooksCorpus, AI companies have pretrained state-of-the-art models that are capable of synthesizing quite convincing pieces of text. Google’s BERT was one of the first to do so, but after it we have seen many others such as OpenAI’s GPT (in all its different versions) or Meta’s OPT.

Some of them are even publicly available for you to play around with in platforms like HuggingFaces, where you can write pieces of text that can be automatically completed by AI models.

The implications of having such accurate text generators go beyond our imagination. These algorithms are not only being incorporated into chatbots for a more realistic interaction with clients and users, but they can even write computer code for us and help us incrementally improve the text we write by mimicking the way another human would do it.

What other applications do you envision for text generation models? At Arionkoder we have teams of AI experts focused on Natural Language Processing ready to help you bring your ideas to life and create new solutions. Reach out to us here and discover what we can achieve together.