OpenAI launches GPT-4o
•
2 min read
This article has been machine-translated into English. Therefore, there may be nuances or errors in the content. The Danish version is always up-to-date and accurate. Read the original Danish article here.
OpenAI has just announced their new model, GPT-4o (o stands for omni), which can generate text, audio and images in real time.
A major piece of news is that the model will be available to everyone – including free users (albeit with limits on the number of prompts), which may help minimise the inequality that can arise when some students pay for access to a better model while others make do with the free one. At the same time, Custom-GPTs are also being made available to free users, and the free version is also getting internet access. Within a few weeks a desktop app for Mac will be released. The Windows app will not be launched until the autumn.
OpenAI is starting by giving Plus subscribers access to the new model, but within a few weeks everyone will have access.
The new model is much faster than GPT-3.5 and GPT-4, which we have tried to illustrate below by running a small test (we have access to the model because we pay for GPT-4).
The new model generates the text a minute faster than the previous model.
Built-in limitations in the model
In GPT-4o, OpenAI has built safety into the model by filtering training data and fine-tuning the model’s behaviour after training. More than 70 experts have reviewed the model to assess areas such as social psychology, bias and misinformation, and to identify risks associated with the model’s new capabilities.
Audio modalities in particular present new risks, which is why OpenAI is starting by releasing text and image input.
GPT-4o and education
We have selected three videos that clearly show how GPT-4o can be used in education.
Below, Sal and Imran Khan from Khan Academy test GPT-4o and the model’s potential to act as a maths tutor. In education there are many opportunities for students not just to receive an answer, but to interact with the model to get support, for example in mathematics. Another aspect is that the help is provided through natural language, which appears to support students’ spoken communication.
In this video, OpenAI demonstrates live translation of a conversation between two people.
It is also possible to have ChatGPT tell you what objects are called in different languages simply by filming the object and asking for a translation.
Watch the full presentation from OpenAI here:
Written by
MSc in IT | Co-founder of Viden.AI | Head of IT and Communications at TietgenSkolen
