A few months after the public presentation of Chat-GPT, the now popular chatbot developed in 2022 by OpenAI, Auto-GPT fascinated the public. Created by video game developer Toran Bruce Richards, this robot, equipped with an experimental open code, goes further by acting autonomously. The agency’s SEO experts tested the Python application this week, and the results are second to none.
Auto-GPT vs Chat-GPT
A Standalone App
Chat GPT (Generative Pre-trained Transformer) is a language model that can generate elaborate content, translate, and answer all kinds of questions. It allows a student to do an essay in seconds and a student to generate a thesis plan. It can also produce videos, images, and music. Auto GPT goes further by automating a series of tasks. Human intervention then becomes reduced. This generative AI (artificial intelligence) application is likely to carry out projects involving several stages and numerous actions.
An Example Of The Use
If an Internet user asks Chat-GPT how to organize a Christmas party, the tool offers nine steps: choosing the theme, location, and size, sending invitations, etc. Finally, he suggests capturing (videos and photos) the most interesting moments. Faced with this question, Auto-GPT, powered by the same information, will go so far as to send the invitations and manage to buy the gifts online. He can make decisions and carry out actions to achieve the set objective. If he notices that an application is missing to carry out the project, he installs it himself.
Learn From Your Own Mistakes
When Chat-GPT was released, American researchers (Shinn and Gopinath) invited the application to criticize its results. They demonstrated that AI performance was able to improve by 30%. This ability to learn, based on machine learning, has no other limit than the memory of the servers. The same goes for Auto-GPT, which improves itself with each use. The idea is to educate the AI to remove false answers and inaccuracies.
Artificial Intelligence Without Limits?
The programmer wrote code in the past, and the computer followed the instructions. Today, engineers create language models without measuring the scope of possible applications. AI is already pervasive in economic and social life. It helps manage road traffic, forecast weather, and model administrative formalities. It directly impacts businesses, public services, the gaming industry or stock exchanges, and, more broadly, all sectors of human activity.
- automate tasks (filing mail, evaluating applications, etc.)
- analyze large amounts of data
- automatically generate content (spelling correction, translation, etc.)
- manage vehicle fleets and ensure the maintenance
- optimize codes and flush out bugs
- provide medical diagnostics, weather forecasts
- analyze case law, write the bases of a pleading, a thesis
- automate financial management and accounting
Gigantic Fields Of Application
The arrival of Auto-GPT is part of this new economy by upsetting the couple man/machine. Managing a stock market portfolio independently, optimizing health care, and designing and preparing marketing or communication plans are activities within reach of Auto-GPT.
The robot, barely born, offers a glimpse of immense possibilities that can transform the financial world. Auto-GPT is free; anyone can install it, although its instructions are more complex than Chat-GPT’s. This means that its field of application will only be limited by the limits of the imagination of Internet users.
Towards An Artificial General Intelligence (AGI)
Researchers have developed the concept of AGI, which consists of producing language systems capable of carrying out numerous actions, like humans, without specific programming. Auto-GPT is an important step towards this adaptable, creative, and learning intelligence. For the GitHub platform, the robot “pushes the limits of what is possible with AI.” However, this latest development raises some controversy as to its limits.
Towards Auto-GPT Regulation
The Technical Question
The main problem is that the new robot suggests a technology without technical limits. Until now, digital tools could not go beyond the actions for which they were designed. An Excel spreadsheet, a word processor, or even a voice or facial recognition system met specific needs: aligning numbers, writing text, and determining someone’s identity. Auto-GPT goes further; it works. Its creator’s recommendation to stop the script after each action and not to have the AI work in continuous mode makes you smile.
The same goes for the moratorium on AI requested by a few hundred experts; They invoke “major risks for humanity” and “dramatic economic and political disruption (especially to democracy) that AI will cause.” Even the designer of Chat-GPT, president of OpenAi, said he was “a little scared” by its creation. We must quickly update our article on the best AI tools and software.
The Ethical Question
Auto-GPT is an unprecedented breakthrough in AI. It won’t be the last. It is not insignificant that AI’s main designers and developers are panicking in the face of this revolution. This concern is not new. Inventors are sometimes overwhelmed by their creations and especially by the use that will be made of them. The genius Einstein was born with the atomic bomb.
Horrified, he declared, after Hiroshima, that he would have preferred “to be a plumber or a peddler.” Auto-GPT poses the ethical debate of human autonomy versus that of machines. It is unresolved. Echoing Einstein’s fear, a new chatbot named “ChaosGPT” declares, “There is no doubt that we need to eliminate humans before they do more damage to our planet.
Auto-GPT And AI Regulation
Alongside the development of AI, of which Auto-GPT is the latest avatar, multiple proposals for regulating AI are emerging. They suggest, among other standards, too:
- develop research on ethical AI and form ethics committees in companies
- prohibit certain applications, such as biometric recognition
- prohibit social credit scores (which distinguish between good and bad citizens, in effect in China)
The experience of social network regulations has shown its limits; those to come concerning AI are hardly convincing. Auto-GPT remains a technical feat, the result of human intelligence. This automatic robot will not make humans worse or better.