Important news for AI-based generative models

In the last few days, some important news has come out regarding AI, and in particular generative models. 

Generative models based on AI are used primarily for images, as well as for videos and texts, and are becoming increasingly realistic.

More powerful AI generative models: the latest news

An important piece of news in this regard is that of Stability AI, which has announced the release of Stable Diffusion 3.5, an open-source AI image generation model. 

This is an open version that includes multiple highly customizable variants based on their size, executable on consumer hardware and free for both commercial and non-commercial use, according to the permissive Stability AI Community License.

Stable Diffusion 3.5 Large and Stable Diffusion 3.5 Large Turbo can be downloaded from Hugging Face, including the code on GitHub. Stable Diffusion 3.5 Medium, on the other hand, will be released on October 29th.

“`html

Stable Diffusion 3.5

“`

Stable Diffusion 3.5 is the most powerful generative AI model created so far by Stability AI. 

In June, they had released Stable Diffusion 3 Medium, the first open release of the Stable Diffusion 3 series, which, however, had not fully met the expectations of the community. 

Thus, after listening to the feedback from the community itself, Stability AI developed a new version to advance their mission of transforming visual media. 

Stable Diffusion 3.5 aims to provide creators with widely accessible, cutting-edge, and free tools for most use cases, and offers a variety of models developed to meet the needs of scientific researchers, hobbyists, startups, and companies. 

This version is one of the most customizable and accessible AI-based image generative models on the market, while at the same time delivering high-level performance in terms of prompt adherence and image quality. 

AI news and updates in the field of generative models: autonomous management of the mouse and keyboard 

But there is more. 

Anthropic has announced the launch of a new beta version of its AI-based model, Claude, which will allow developers working with the API to even take control of the mouse cursor, to click on buttons and fields and enter text autonomously.

This update effectively allows developers to instruct Claude to use computers like people do, namely by looking at a screen, moving a cursor, clicking buttons, and typing text. According to Anthropic, Claude 3.5 Sonnet is the first AI model to offer computer use in the public beta, although at this stage it is still experimental, often resulting in cumbersome and error-prone behavior. 

Asana, Canva, Cognition, DoorDash, Replit e The Browser Company are already experimenting with this new feature. 

The update of Claude 3.5 Sonnet is available for all users. 

Negative news

However, there are not only positive news regarding the development of generative AI models. 

For example, Penguin Random House (PRH) has introduced a new statement in the copyright pages, both of new and reprinted books, stating that no part of its books may be used or reproduced in any way for the purpose of training intelligenza artificiale technologies or systems.

PRH is the largest commercial publisher in the world, and therefore it could set a precedent from this point of view. 

The declaration cites a directive of the European Parliament, and aims to prohibit the use of its texts by artificial intelligence companies for the development of chatbot and other digital tools.

On the other hand, in the USA several cases of copyright infringement have already emerged in this regard, so much so that many books have already been used without authorization to train artificial intelligence tools. 

PRH is the first among the Big Five of Anglophone commercial publishers to modify its copyright information for such purposes. 

Accelerations and braking

As it practically always happens, a new rapidly expanding sector produces strong accelerations and big slowdowns. 

The problem in these cases is that the accelerations could take it very far forward, so much so that the braking might only slow down the acceleration without being able to stop it. 

However, not all that glitters is gold, and in all evolutions, there is always also a component of destruction. 

In this case, the risk is that of destroying the business model of some of those who produce information, and doing so by resorting to what is essentially the theft of intellectual property. 

It is not even certain that legislators are able to keep up with these rapid evolutions, therefore it might take some time before all the more delicate issues are examined, and perhaps resolved. 

What is clear, however, is that it is not just a trend, or a flash in the pan, but one of those great leaps forward that has led human society to dominate this world unchallenged. 

Source: https://en.cryptonomist.ch/2024/10/23/important-news-for-ai-based-generative-models/