Text-to-video Is Here; Text-To-Image Models Were Considered Popular Up Until Last Month.

by | Mar 7, 2024

AI is evolving so quickly. The most cutting-edge text-to-image models were released only a few months ago, but engineers are already showcasing text-to-video systems.

Make-A-Video, a multimodal algorithm introduced by Meta, allowing users to provide a text description of a scene as input and receive an animated clip that generally depicts the scenario as input. As an input prompt, you may also utilise other forms of data, such a picture or a video. According to a non-peer reviewed publication [PDF] detailing the programme, the text-to-video system was trained using open datasets.

The samples provided by Meta demonstrate that these faux AI films don’t have the same level of quality as some of the photos produced by generative models. However, text-to-video requires more computing and relies on the generation of a series of pictures in order to capture motion. The Make-A-Video service from Meta is presently not open to the general public, but those who are interested in testing it out can sign up for access.

The owner of Facebook said in a statement, “We are transparently disclosing this generative AI research and results with the community for their feedback, and we will continue to utilise our responsible AI framework to develop and adapt our approach to this new technology.”

In order to create deepfake footage of the Die Hard actor for any upcoming films, Bruce Willis gave his image rights to Deepcake, a firm that creates videos using artificial intelligence.

A bogus Digital Duplicate Of Willis Has previously made an appearance in a MegaFon commercial in Russia.

AI has been used to mimic an actor’s voice and look, but according to Gizmodo, Willis may be the first to formally sell the rights to use his likeness in any future deepfake media projects. Following a diagnosis of aphasia, a medical disorder that impairs language comprehension and communication, Willis announced his retirement from the entertainment industry.

According to a quote ascribed to Willis and published on Deepcake’s website, “I appreciated the accuracy of my persona.” “It’s a wonderful chance for me to go in the past. My persona is reminiscent of the visuals from that era because the neural network was trained on the material of “Die Hard” and “Fifth Element.””

“Modern technology has made it possible for me to work, communicate, and take part in filming even while I’m on another continent. I am appreciative to our team for providing me with such a novel and fascinating experience.”

NLP Is Being Used To Target Paper Mills:

Publishers can use natural language processing algorithms to determine whether a scientific submission was likely produced by a fraudulent scientific paper mill.

Paper mills are dubious companies that create phoney research for writers who wish to look credible. Science articles are ghostwritten by people who are paid, and they frequently contain plagiarism by misrepresenting the findings of prior work. These fraudulent articles are frequently distributed by journals with a poorer reputation, which cares more about collecting publishing fees than a paper’s quality.

According to Nature, six publishers, including SAGE Publications, are now interested in exploring AI-enabled software to automatically identify publications that seem to have been made in a paper mill. Adam Day, a director and data scientist at Clear Skies, a business in, created Papermill Alarm.

The programme calculates a score indicating how likely it is that a paper was written by a forger by comparing the phrasing of the title and abstract to those of manuscripts from paper mills. One percent of the titles of publications with citations in the PubMed system, according to Day, appear to be phoney research created by paper mills.

The sum was “too high for comfort,” according to David Bimler, who goes by the alias Smut Clyde and is referred to as a research-integrity sleuth. “These unreliable papers are cited. People take use of them to support their own flawed theories and continue fruitless research projects “said he.

Controversial Project Maven Contract Is Expanded:

Leaders at Google cancelled a deal with the US Department of Defense for Project Maven in 2018, making it possible for businesses like Palantir to continue up where it left off by using AI to analyse military drone imagery.

A $229 million one-year deal with the US military forces, joint staff, and special forces was announced by the big data analytics company as part of its expansion of services. According to Bloomberg, some of that money comes from maintaining Project Maven.

“The Department of Defense continues to retain a leadership edge via technology and by providing best-in-class software to those on the frontlines by introducing cutting-edge AI/ML capabilities to all members of the Armed Services.