Goal It does not give rest with its projects linked to generative AI, and it would be about to play one of its most important chips. According to The InformationMark Zuckerberg’s company works on a new solution open source designed to suggest code and help users program. Would be called Code Flamewould be released as early as next week and could be in direct competition with Copilot and Copilot X, the popular GitHub tools.
The sources of the aforementioned medium affirm that Code Llama would be based on LLaMA 2, the most recent artificial intelligence model developed by those from Menlo Park. In any case, several details remain to be known regarding its implementation, main characteristics and distribution.
Despite still not knowing its finer details, it makes sense for Meta to point its guns at an AI capable of writing or correcting code. Keep in mind that conversational chatbots like ChatGPT (OpenAI) and Bard (Google) already offer support for a wide variety of programming languages. And while it is true that they are still far from perfect, they have proven to be very useful. to speed up workflows, especially when detecting errors and suggesting corrections. In fact, 46% of the code published on GitHub is already generated with artificial intelligence.
That said, everything indicates that Code Llama would not be offered as a bot capable of programming already “canned”. On the contrary, it would be a set of tools for companies, researchers or engineers to can create their own programming assistant based on artificial intelligence. This could be especially useful for those individuals or organizations that are trying to cut costs on their projects. Let’s not forget that GitHub Copilot, the most popular tool among developers, costs $10/month for individuals and $19/month per user on the enterprise plan.
It will be interesting to see how Meta intends to address one of the major problems of AI capable of generating code: plagiarism. This stands out as one of the hottest moral debates around these tools, considering that many times they are trained with proprietary code and reused without permission. And today it is impossible to know if Code Llama can be the exception to the rule.
Everything indicates that Code Llama is Meta’s new great bet to become an industry leader in generative AI. Something for which Mark Zuckerberg has preached in recent times. And the Californian firm seems to have taken this possibility very seriously, chaining together important announcements in a very short period of time.
After all, LLaMA 2, the AI model that Code Llama will be based on, was officially unveiled in mid-July. While in August, those from Menlo Park introduced AudioCraftwhich is capable of creating songs and sound effects from text.
waiting for more details
It remains to be seen which programming languages Code Llama supports and how it evolves over time. The truth is that, regardless of whether the proposal is successful or not, Meta would seek to demonstrate that it is at the same level —and even above— OpenAI and Google, which in recent months have attracted attention for their developments in terms of generative artificial intelligence. And having your projects promoted as open source gives you even more exposure.
It is true that there is a lot of debate about whether Meta’s AI models are really open source. And scrutiny is likely to fall on Code Llama as well. Recently, the Open Source Initiative asserted that the LLaMA 2 license it is not open source. This, because it restricts its commercial use to certain users, as well as its implementation for certain purposes. However, it is an approach that the other big players in the sector have not yet turned to.
Will Code Llama be Meta’s next big AI bombshell? Can it stand up to GitHub Copilot and chatbots like Bard and ChatGPT? We will be attentive to the news.