OpenAI, this December 2024 launched an initiative for the first time. It was a 12-day program during which they made announcements and they presented several events that took place in December. It could be described as a “tech marathon”. The aim of this initiative was to bring to the surface new technologies, product updates and other OpenAI research developments. During the last day of this project, in 20 December 2024, OpenAI announced her new AI model, the “o3”. O3 is more like a family of models, constituted from two main parts.
- The first version is the “o3-mini” and plans to be available by the end of January 2025. It is very compact, with a very low cost but it doesn’t concern all type of tasks.
- The second and complete model is the “o3” and it is supposed to be accessible at the end of March 2025.
This model seems to be more of an Artificial General Intelligence (AGI) model, without being there yet.
Definition of Artificial General Intelligence (AGI)
To begin with, we are going to explain what AGI is concretely. AGI is an artificial intelligence with the aim to perform without any human help and to develop human cognitive abilities itself. It is supposed to be able to understand and perform like a human. In other words, AGI should have the ability to learn and adapt to new situations in a more autonomous, successful and profitable way.
Differences from existing artificial intelligences
The “o3”, the “o1” successor, strives to be different from other artificial intelligences such as ChatGPT.
Firstly, it aims to have a larger capacity of generalization. Unlike other artificial intelligences, this one will be able to learn, create, analyze and respond based on logic and not just on the large databases it holds and treats, like the other AIs (machine learning).
Then, another objective of OpenAI is to create more of an autonomous and adaptive AI with the aim of its advanced logic.
Due to these characteristics, “o3” could be applied to a much wider range of tasks, such as research, strategy and decision making, and other tasks usually requiring human intelligence. Some domains where “o3” could be more accurate and efficient are fields like physics, math and programming.
Relative concerns and challenges
- Job market: Due to the attempts to reach a human-level intelligence, it is believed that this AI could probably impact the job market.
- Ethical: If the capabilities of this model are, in fact, so significant and imitate human logic, it becomes even more essential to regulate new artificial intelligences and ensure their compliance with ethical principles.
Nevertheless, if handled correctly and under the proper regulations regarding these two major concerns, Open AI “o3” has the potential to be a helpful tool, providing valuable solutions, while minimizing risks.
- Cost: As far as the cost, “o3” seems to be really expensive with around $30 to $6000 dollars per request, regarding the difficulty and complexity of each query. This high operational cost could limit access to only persons that can afford a really expensive subscription. So these high costs could create accessibility barriers.
It is true that, at the moment we don’t have much information about this new model. We will likely begin using it at the end of this month. So only then we will be able to understand its real impact and determine if it performs as well as it is presented.
Sources:
Etudiante en Master 2 Droit de l’économie numérique (2024-2025).
