The announcement was eagerly awaited: OpenAI has just unveiled GPT-4, its new multimodal AI model. Here are 7 apps that already use this technology.

Bing has been using GPT-4 since launch.

Microsoft was one of the first companies to congratulate OpenAI on the release of GPT-4. And for good reason: it also made it possible to announce that Bing was already using the new artificial intelligence model. Yusuf Mehdi, Corporate Vice President of Microsoft, also specifies that this has been the case since the beginning.

We’re happy to confirm that the new Bing uses GPT-4, which we’ve customized for search. If you’ve used the preview of the new Bing in the past 5 weeks, you’ve already experienced an early version of this powerful template.

Duolingo launches 2 new features based on GPT-4

Duolingo, an application for learning a foreign language, already used artificial intelligence – to personalize lessons and tests, in particular. But Edwin Bodge, main product manager, wanted to go further: “there were gaps in a learner’s journey that we wanted to fill: conversational practice and contextual feedback on errors”.

Duolingo today launches a new subscription, Duolingo Max, currently available to English users who are learning French and Spanish. It contains two new features based on GPT-4.

• Role play: this feature allows you to converse with a fictional character from the application and thus practice exchanging in another language. It adapts to your language level to help you progress better.
• Explain my Answer: to obtain appropriate details, allowing the quality of the proposed response to be judged (and thus better understand its errors).

Be My Eyes uses the multimodal capabilities of GPT-4

This is one of the major novelties of GPT-4: it is a multimodal artificial intelligence model, capable of interpreting text and images. Be My Eyes is the first OpenAI partner to explore these new possibilities.

Be My Eyes develops technologies to make life easier for blind or visually impaired people: object recognition, help for navigating in public places… The publisher has designed Virtual Volunteer™, to meet the needs of its users thanks to their photos and the expression of their needs.

You can, for example, take a picture of the contents of your fridge and request a list of recipes based on the ingredients stored there. You can also ask the app “Which of these two shirts is red”. The use cases for this type of technology are very promising.

Stripe uses GPT-4 for support and to fight against fraud

Earlier this year, Stripe asked 100 employees to consider possible applications of GPT-4 to improve its services. Several uses have been selected and integrated by the online payment specialist.

• To facilitate the support provided to its users, Stripe uses GPT-4 to analyze websites and generate summaries of their activities, “which surpass those written by people”.
• OpenAI’s new technology also allows Stripe to provide relevant answers to developers who have technical questions, by interpreting the issues and reformulating documentation.
• Finally, GPT-4 is used by Stripe on Discord, to analyze the syntax of messages and detect accounts suspected of fraud. The specialized teams can then carry out a verification – human this timecomments made, in order to assess the risks and fight against scams.

Other use cases for GPT-4

On its blog, OpenAI presents three other use cases of GPT-4.

• The Government of Iceland: the artificial intelligence models are trained with widely used languages, including English. Language support – and the quality of the responses provided – depends in part on the data used. Icelandic volunteers were recruited to help train the model, to enable GPT-4 to respond properly in Icelandic.
• Khan Academy: in the same spirit as Duolingo, Khan Academy uses the capabilities of GPT-4 to facilitate the learning of new skills. Khanmigo is a new virtual assistant for students and trainers (creation of personalized courses, response to identified issues, etc.).
• Morgan Stanley: finally, Morgan Stanley used GPT-4 to structure its wealth management knowledge base. A chatbot-type interface then facilitates the support of customers by the establishment’s advisers.

The risks and opportunities of artificial intelligence

These first applications of the new OpenAI language model show very interesting possibilities – especially for the education and support of visually impaired people. They suggest new, promising uses that will make life easier for many users.

At the same time, we also perceive several risks associated with the rapid evolution of these technologies. Despite the initiative of the Icelandic Government, we can fear a more rapid development of the uses of artificial intelligence in the most spoken languages, including English. And above all, certain examples presented by OpenAI (Stripe, Morgan Stanley) show the risks that artificial intelligence represents for certain professions, including those of customer support. Appreciated as “partner allowing humans to be relieved of certain tasks”, AI will undoubtedly take the place of some employees in companies. Overall, the balance may be positive. Economists have always had very different views on the subject. But one thing seems obvious: the productivity gains achieved through AI will not directly benefit those who will see their function replaced by these new technologies.