Away from the power struggles in the leadership of Open AI, the company is planning revolutionary innovations for Chat-GPT. The AI tool should reach the masses thanks to an app store.
Soon Chat-GPT users will be able to create their own individual tools and even earn money with them.
Sebastien Bozon/AFP
The startup OpenAI has made headlines in media around the world in recent days due to the drama surrounding the dismissal and return of CEO Sam Altman. In addition, the technical innovations that the company presented at its first developer conference in November were lost. The changes are considerable and their potential is great: they could help the AI tool move even further into the mainstream.
The most important change is integrated apps, so-called “GPTs”. The function allows ChatGPT users to create their own, individual voice assistants and tailor them to their personal requirements. No programming knowledge is necessary. Instead, the ChatGPT setup is done via conversation in natural language.
If OpenAI has its way, the GPTs should not only make the use of the service more individual, but also form the basis for an entire AI ecosystem. To this end, the company wants to launch a kind of app store for ChatGPT in the coming weeks, in which users can offer and monetize their GPT creations.
AI apps generate images and analyze program code
Like many other features of the AI system, GPT creation is reserved for Chat-GPT Plus subscribers. In the “Explore” area you will find the option to create as many GPTs as you want. To do this, you tell the service what your AI bot should do. An assistant then guides you step by step through the creation of your own GPT model; later adjustments are possible at any time. The associated possibilities are diverse. The possibilities that Open AI already offers range from simple chatbots to complex productivity environments including connections to external web services.
The code interpreter, which can analyze program codes, as well as the Dall-E image generator, which was recently updated to the third version, can also be integrated into your own GPTs. Artists can use the latter, among other things, to convey their personal drawing style to the GPT, which the service can imitate in an almost incredibly credible way. Authors, on the other hand, have the option of uploading their own manuscripts to convey a style to the chatbot or to ask questions about complex content such as books. Companies could use the GPTs for employee training, for example to create chatbots based on company guidelines and manuals.
Although the app store is not yet online, the GPTs created in this way can already be shared with others. Countless specialized GPT models were found on Reddit, X and Discord just a few days after publication. For example, while Assistant-Assistant acts as a helper for entry-level programmers, God-GPT opens the discussion about various philosophical and religious ideas based on well-known literary works.
The possibilities of GPTs are impressive. However, the system also has its downsides. The biggest one is probably the nature of Chat-GPT: All data that users entrust to the GPT model is processed in the cloud on Open AI servers. By default, the company also reserves the right to reserve all input for training its own AI skills. This procedure can be objected to in the account settings, but of course this does not change the server-side evaluation of the entries.
With the introduction of the GPT store, Open AI could also disrupt the market for third-party AI service providers. Countless small startups rely on the chat GPT interfaces for their supposedly own AI offerings, including many browser extensions for Chrome and Firefox. From a user perspective, such programs are convenient, but in practice they are often more expensive than direct access to Chat-GPT. If Open AI makes the GPT app store transparent, this could potentially lead to a consolidation of the AI oversupply.
Knowledge level extended to April 2023
GPT generation is not the only innovation that Open AI has given its flagship product in the past few weeks. The tool’s knowledge base has been expanded to April 2023. Subscribers also no longer have to choose whether to have a conversation, create an image, or perform code analysis. Instead, the chat leads dynamically to the goal. Chat-GPT’s smartphone apps have also recently started allowing spoken conversations.
Behind the scenes, Open AI has introduced a new language model called GPT-4-Turbo. Anyone who accesses the service via an API interface can use it to make longer entries, for example to send complex texts several hundred pages long to the system. Both with this and with the introduction of personal GPTs, Open AI also wants to eradicate the digital forgetfulness of chat GPT step by step. In several sessions, the developers behind the artificial intelligence emphasized that longer storage of conversation information is high on the internal priority list. The development of Open AI and Chat-GPT remains exciting even outside of business-political trench warfare.