Facebook, a Meta Platforms company, turned 20 on Sunday, which seems like an eternity for a technology company. Perhaps it is because Mark Zuckerberg, Meta founder, continues to invent new types of services to keep the company young. At least investors think so. 

Sunday marked the day in 2004 that Zuckerberg, a 19-year-old entrepreneur and student, went live with a new website, thefacebook.com, which at that time was available to Harvard University students, and then expanded to students at other colleges.

Last week on Thursday, the company reported 3.07 billion monthly users on Facebook — driving up profits to $39 billion in 2023. It sent the stock per share price up 20.3% on Friday. 

Meta reported fourth-quarter revenue grew 25% to $40.1 billion, and said it will start paying dividends, the first distribution in its history.

Massive layoffs of about 20,000 employees during the past year and vast investments in artificial intelligence (AI) by the company continue to wave the “shiny object” in front of advertisers. Advancements in its Large Language Model Meta AI (Llama), a code-generating AI model, will be expensive. 

Llama, however, makes ads more relevant to users across all of Meta’s companies, such as Facebook, Instagram, WhatsApp, and Reels.

Last week the company announced an advanced version of Llama, which comes with increased processing power, greater accuracy and support for specific programming languages, such as Python, C++, Java and PHP.

Code Llama 70B is a derivative of Meta’s open-source Llama 2 and designed to create code based on natural-language prompts, which is text that describes the task AI should perform.

Mass-producing code makes software development more efficient and accessible and lets developers get more creative. Code generation is complex. The AI models require large amounts of computing power. This is where the latest Llama, 70B, excels, according to Meta.

Meta open-sources its AI. During the company’s earnings call the company explained that open sourcing improves its models.

It is typically safer and more secure, and more efficient to operate because of all the online feedback from those who use it. Efficiency improvements will lower the cost to compute. Open-source models often become an industry standard.

Engineers at Meta also provided a guide that explains several basic steps related to LLMs.

The tips are published under Llama Recipes on the company’s GitHub page, Prompt Engineering with Llama 2, which means using natural language to produce a desired response from a LLM.

There are six steps for getting the best results from Llama 2.