You are probably already familiar with GPTs, these artificial intelligence systems that can generate text from keywords or phrases. They are able to converse, summarize, translate, write and much more. But did you know that there is a new technology that goes even further? It is Auto-GPT, a system that can not only make text, but also give itself a plan and execute actions to achieve this plan.
Auto-GPT, more than just a chat
Auto-GPT is a new technology that can learn to do tasks in any field, from simple instructions or examples. It can, as with other GPTs, write creative content like poems, stories, songs, pictures or parodies of celebrities but it can also search the web, connect to APIs, make a multi-step plan and himself to make "prompts" by using other GPTs.
It is therefore a major advance over the old GPTs because it is not limited to making text or images. He can give himself a plan and carry out actions to achieve his objectives. For example, he can create a website from specifications, write a contract from a list of clauses, design a suitable diet with recipes and feedback or even program an application from a scenario.
It is a system that constantly learns and improves because it does not depend solely on its initial training. He can use the data available on the web or those provided by users to enrich his knowledge and skills. He can also receive feedback or corrections to improve his performance. It seems to have few limits and boundaries. He can learn to do tasks in any field, according to the needs and desires of the users. He can also combine several tasks to carry out more complex or much more ambitious projects.
The potential is therefore enormous and can literally revolutionize the world. Technology can bring many benefits to individuals and organizations. It can facilitate access to information, education, culture or entertainment. It can also increase productivity, quality or creativity.
But Auto-GPT is also a risk-increasing technology, risks that we are only just discovering with his little brother ChatGPT. This technology will reduce jobs in several sectors or professions. It will also most likely be used for malicious or illegal purposes. Several Auto-GPT type projects are in "open source" mode and the speed at which the community participates in this project is impressive. We can remember the dazzling speed of innovation that followed the "open source" release of Stable Diffusion. The fact that these projects are not centralized makes it almost impossible to control, increasing the risk of them being used by malicious people.
Companies must ask themselves questions very quickly about how these technologies will impact their business model and how to use them quickly in order to remain competitive. Several jobs are in danger and the speed of evolution suggests that this will continue very quickly.
If you are a developer, you can experiment with the technology yourself via its GitHub link Here are some examples of early stage Auto-GPT implementation : GoalGPT : https://beta.nando.ai/goalgpt.php Cognosys : https://www.cognosys.ai/