The contribution hard: New AI creates realistic pictures – faster than ever before by Felix Baumann appeared on Basic Thinking. You always stay up to date with our newsletter.
Researchers at the Massachusetts Institute of Technology (with) and by chip manufacturer Nvidia have developed a new AI model with “Hart” that should be able to create realistic images faster and more efficiently than ever.
Researchers of the MIT and NVIDIA have a new AI model called Hart (Hybrid autoregressive transformer) Developed that should generate realistic images faster and more efficiently than previous systems. The AI combines the strengths of two approaches: the quick but error -prone auto -compatible models and the high -quality but slow diffusion models.
The result is a hybrid system that should be able to generate images up to nine times faster than ever before – with the same high quality. Author -compressive models that are also used in voice models such as chatt creates content gradually – for example, image areas one after the other.
This method is quick, but prone to errors. Because such systems cannot correct errors again. Diffusion models, on the other hand, work iterative and refine a picture through several computing steps. They deliver relatively good results, but are significantly slower and more energy -intensive.
So the AI model generates hard realistic pictures
At HARt, a auto -compressive model initially creates a rough image version, which is converted into compact data packages – so -called tokens. Then a small diffusion model takes over the fine work by optimizing only details such as edges, faces or hair.
Because the diffusion model focuses exclusively on these areas, eight computing steps are enough – instead of the 30 or more that are otherwise common. A decisive advantage of hard is its efficiency. It needs around 31 percent less computing power than current high-end models, but provides a comparable image quality.
This is made possible by a compact architecture: While other systems work with billions of parameters, significantly less comes from – without loss of quality. This can even be operated on a smartphone.
Efficient performance for real applications
This lightweightness makes the AI model hard particularly interesting for mobile or energy-limited applications. It brings the necessary strength to create realistic images quickly and resource -saving – even outside of data centers.
In the future, the model could play a key role in so-called multimodal AI systems-i.e. in applications that process text, image and sound at the same time. The system for robotics, the gaming area or virtual assistants should be particularly interesting. Because there must be generated and combined in real time.
Also interesting:
- Ki-token: The smallest linguistic unity of artificial intelligence
- Artificial photosynthesis: New process gains energy from waste
- Artificial intelligence should improve flood protection in Germany
- Artificial intelligence: 7 AI podcasts that you should know
The contribution hard: New AI creates realistic pictures – faster than ever before by Felix Baumann appeared on Basic Thinking. Follow us too Google News and Flipboard.
As a Tech Industry expert, I am fascinated by the advancements in AI technology that are enabling the creation of realistic pictures at an unprecedented speed. This development has the potential to revolutionize industries such as graphic design, gaming, and animation, where high-quality visuals are crucial.
The ability of this new AI to generate lifelike images faster than ever before opens up a world of possibilities for artists and designers, allowing them to iterate and experiment more freely in their creative process. It also has the potential to streamline workflows and reduce production times, ultimately leading to more efficient and cost-effective projects.
However, as with any new technology, there are also potential ethical and societal implications to consider. For example, the rise of AI-generated images raises questions about the authenticity and ownership of digital content. It will be important for creators and consumers alike to be vigilant about ensuring that proper credit is given and that copyright laws are respected in this new landscape.
Overall, I am excited to see how this new AI technology will continue to evolve and how it will impact the creative industries in the years to come. It is an exciting time to be a part of the tech industry, and I look forward to seeing the innovative ways in which this technology will be utilized.
Credits