LlaMA: Meta's AI joins the competition against ChatGPT
Earlier this month, we announced that Sundar Pichai, CEO of Google and Alphabet, made a public and official announcement about the start of the release of the Google's ChatBot Bard. With which, of course, they hope to be able to compete directly with OpenAI's ChatGPT.
While today, a few days after the end of this month, we echo the Mark Zuckerberg official announcement, President and CEO of Meta, on the existence and official launch of «LLAMA» the new Meta AI. Which, will seek to compete with the AI-based technology that gives life to ChatGPT from OpenAI, Sidney from Microsoft and Bard from Google. In other words, this technology will not be focused on the end user, but rather on the researchers and developers of these AI technologies.
Google's ChatBot Bard: The Next Big Competitor to ChatGPT
But, before starting this post about "Calls" the new Goal AI, we recommend that you explore, later, the previous related post:
Calls: Goal AI released to compete with ChatGPT
What is LLaMA, the AI of Meta?
Citing the Meta AI official announcement about your AI FLAME we can anticipate the following:
As part of Meta's commitment to open science, today we are publicly launching LLaMA (Large Language Model Meta AI), a next-generation large language foundational model designed to help researchers advance their work in this subfield of AI. Smaller, higher-performance models such as LLaMA allow others in the research community who do not have access to large amounts of infrastructure to study these models, further democratizing access in this rapidly changing important field.
How will it work and what will be its benefits and differences?
And, among the highlights of this new AI technology available to researchers and developers, we can mention the following points:
- LLaMA as part of smaller base models is ideal for training and testing large language models. Which is because it requires far less computing power and resources to test new approaches, validate the work of others, and explore new use cases.
- will facilitate the full access to AI research, making it possible for more researchers can understand how and why, behind the operation of these great models of language. In such a way, to increase the progress of efforts to improve the robustness of said technology. And thus, mitigate known problems, such as bias, toxicity and the possibility of generating misinformation.
- LlaMA works by taking a sequence of words as input, and then predicting the next word or phrases. Trying like this, generate text recursively. For this, she has been trained in texts based on the 20 most spoken languages worldwide. Focusing on those with Latin and Cyrillic alphabets.
- It is designed to be versatile, and to be applicable to many different use cases, compared to other lean AI models. That is, those that are designed for one or more specific tasks.
- For now, and in order to maintain the integrity and avoid misuse of LLaMA, Meta has decided to release said technology under a non-commercial license (General Public License GNU version 3.0) focused on research use cases. Sabove all, to those who are affiliated with US government organizations, civil society and academia. Y of course, industry research laboratories around the world.
More about this AI technology
For those interested in learning more about LlaMA, the AI of Meta, we recommend you explore directly the PDF document by Meta Research. And if necessary, your website on GitHub.
Summary
In short, let's hope that this new Artificial Intelligence technology available, that is, «LLAMA», the AI of Meta, contribute so that the entire Community of researchers and developers of AI technology at a global level, can work more together. In such a way, that they manage to develop clearer and more efficient guidelines on the creation and use of responsible (ethical) AI in general. And, of course, with everything related to it.
If you liked this post, do not stop sharing it with others on your websites, channels, groups or favorite communities of social networks or messaging systems. And finally, remember visit our home page en «DesdeLinux» to explore more news, and join our official channel of Telegram from DesdeLinux, West group for more information on today's topic.