Meta wants to compete against ChatGPT and Google Bard with its AI “LLaMA”

Few times in our lives have we seen something popularize as quickly as the AI ​​ChatGPT. From the moment it was launched and in just two months, it already had 100 million users. But this was just the beginning, since by the time Microsoft implemented it in Bing, its popularity went even further. Most people have already heard about this AI and to a lesser extent its Google Bard competition. What we did not know is that Goal also wants to enter that sector dominated by ChatGPT and for this you will use a AI linguistic model called LLaMA.

AIs are the present, but also the future if we talk about saving time and money to perform different tasks. The automation Using artificial intelligence allows us to be more productive, although it also creates uncertainty as to whether we are going to lose our job in the future. Also, as much as we want to deny it, AIs are going to be a part of our lives and this is only the beginning, considering the rate at which they are progressing. Of course, there is a huge market opportunity behind AI like ChatGPT, something that is certainly taking advantage Microsoft wanting to include ads in Bing.

Meta presents its model for AI LLaMA that could compete against ChatGPT

We already have an AI like ChatGPT on Bing and Edge and Google Bard has already been introduced as its main competition. Now we have a new challenger who will compete announcing their new LLM. simplified as Callswe have a large language model for Meta AI. This will be available in 4 different sizes, ranging from 7,000 to 65,000 million parameters. Here we are not referring to an AI chatbot like ChatGPT, because as a linguistic model, its uses will focus on other areas.

LLaMA is now available for research work with non-commercial license and they can use it university laboratories, NGO or companies upon prior request. The company admits that only publicly available data has been used to train this model. Of all these, the 67% come from CommonCrawl, 4.5% from Wikipedia and another 4.5% from GitHub. Of course, Meta assures that her FLAME achieves better results than the GPT-3 language model of OpenAI in many cases, although it is true that the AI ​​ChatGPT employs now is GPT-3.5 improved.

The previous Meta model was called Galactica and it lasted only 3 days

The LLaMA language model is not the first to be announced by Meta, since in November 2022 there was another. In this case we talk about galacticwhich was intended for answer questions and help in writing texts. However, its useful life was very short, it was only available for 3 days before its closure. The causes of this were his incorrect and unobjective answers, something that did not make it optimal for use in research. Also, this was the time when ChatGPT came out and everyone noticed it.

Now that Meta has seen how ChatGPT-type AI is able to reach millions of people, is interested again in this segment. We have not yet seen what its new LLaMA model is capable of, but what we do know is that it will not be free of problems.

“The risks and harms of great linguistic models include the generation of harmful, offensive or tendentious content. These models are often prone to generating incorrect information, sometimes called hallucinations. We don’t expect our model to be an exception in this regard,” said Mark Zuckerberg.