For weeks, the entire technological world had their eyes on Openai, which already had everything learned to launch a new open -source artificial intelligence model, accessible so that anyone can use it, modify it or train it by itself.
But a few days after the premiere, Sam Altman has thrown the handbrake: «I’m sorry to be the bearer of bad news, but we need more time,» he wrote on X, before Twitter, recognizing that there are still loose ends in security.
The truth is that the model, of which the final name is not yet known, was going to be a huge change in the strategy of the AI company. While its previous versions – as chatgpt – work closed, fully controlled from the company’s servers, this new model would deliver the ‘pesos’ of the system to the community.
That, evidently, opens the door to infinite good possibilities … and bad. Therefore, Altman has stressed that Once an open source model is published there is no turning: «Once the weights are out, they cannot be removed,» he explained.
It is precisely for this reason that the team has decided to stop the launch to do more security tests and thoroughly study the potential risks. Or at least that’s what they count. Just remember that this is not the first time that they delay it and experts begin to suspect that perhaps the company is not prepared for such a step.
The great dilemma of the ‘Open Source’ of the AI models: goal has a lot of advantage
The truth is that in this sector there have always been two sides. On the one hand, giants such as Openai, which offer very powerful solutions, but that, to access the most top, you have to pay monthly for it in the form of subscription.
On the other, a large community of developers that does not stop growing and that bets on open source, argue that it is the true way to democratize AI. Precisely here mention the goal figure, which, so that you get an idea, is the business model that is carrying out when it comes to artificial intelligence, allowing anyone to access their models, touch them and improve them.
In essence, it is about making AI accessible to all, not only for large companies, governments or certain people. However, the path to this is not easy.
The most closed approach, of course, has its advantages, especially in very critical sectors where something is supported by a large company and is 100% reliable is vital. However, with this also comes the questions about accessibility. Is a digital gap in the world of AI be created?
On the other hand, the open source offers something totally different. They eliminate economic barriers and encourage global collaboration, which can accelerate innovation. Aitor Pastor, CEO and founder of Disia, told us in an interview that «open source models are innovation and transparency engines. And allow the continuous improvement of these solutions by the development community. These models give rise to a faster and faster innovation.»
The big problem is that this freedom also scares. Because if the model is powerful enough, it is possible to train it for less positive things: massive misinformation campaigns, generation of toxic content or automation of cyber attacks. Altman has insisted that these possibilities are on the table and that they cannot be taken lightly.
With this novelty, which we hope will come out soon, the OpenAi CEO walks down a tightrope, since it intends to open something enough to keep your community happy, but without running the risk of the model getting out of control. That is why it seems to decide to put the pause.
Know How we work in NoticiasVE.
Tags: Artificial intelligence, software