Facebook’s Most Powerful AI Yet

When it comes to innovation, Mark Zuckerberg is no slouch.

This week, the Facebook founder took his boldest step yet… releasing a powerful new AI model that has engineers at top AI firms like OpenAI and Anthropic scrambling.

The newest model, Llama 3.1, builds on the capabilities of Facebook’s previous Llama 3 model, with improvements in its abilities to use common sense, follow instructions, and answer math questions.

According to common industry benchmarks, Facebook’s newest model surpasses the accuracy of the latest models from competitors like ChatGPT and Claude 3.5.

As with previous LLama models, Facebook is giving LLama 3.1 away for free – developers are free to use and improve 3.1 to meet their specific needs.

For companies with AI-intensive tasks, the flexibility of 3.1 is a financial godsend.

According to Meta, the costs of operating Llama 3.1 on a company’s infrastructure are 50% less than using the latest models from OpenAI.

Perhaps most excitingly, Llama 3.1 comes with a new legal license that allows developers to use the model to train their own models.

Previous LLama versions had specific restrictions on how developers could take content produced by Llama and use that as input into new AI models.

This concept, called “synthetic data,” is becoming an increasingly important topic in the next generation of AI development.

Current AI language models have been able to advance by absorbing as much information as possible in order to get smarter.

However, scientists are quickly reaching the limits of the amount of information available to train AI models.

The latest models from Google, OpenAI, Anthropic, and others have already been trained on nearly the entire public internet and every document ever published.

This means that in order to make AI smarter, scientists are now using AI to create content to train the next generation of models.

With Llama 3.1, researchers and businesses are now free to use Meta’s model to create new content.

The policy change represents the latest signal from Zuckerberg that he does not believe that the latest AI tools pose a risk to Meta’s profits.

Also worth noting – the release of 3.1 comes just three months after Meta released its last model update, Llama 3.

This release speed is unheard of in the world of slow-moving big tech companies.

Zuckerberg predicts that at the current rate, Llama will be able to easily surpass the abilities of all competing AI models by sometime next year.

Of course, this sort of development speed comes at a price.

In an earnings call this week, Meta executives indicated that they plan to spend about $2 billion more on capital expenditures (capex)… including pricey Nvidia servers for developing new AI models.

That pushes Meta’s estimated spending into the range of $37 billion to $40 billion for the year.

Despite the expense, Meta is optimistic about the outcome.

Meta’s sales in the quarter increased 22% over the past year, a growth rate nearly double that of sales at Google.

Zuckerberg attributed that increase in sales as one of the benefits of the company’s AI investment…

More Than Just a Tech Flex

In addition to developing the Llama family of models, Meta has quietly been hard at work improving the AI that powers its recommendation and advertising algorithms.

Meta’s success in monetizing AI could give the company greater freedom to invest in even more AI projects.

This comes at a time when other big tech companies are struggling to make the same case to Wall Street.

Just one day prior, Microsoft told investors that growth in its cloud computing business had slowed in the most recent quarter. 

Microsoft’s cloud business has been a major source of growth – and expense – in recent months as the company pushes to achieve sales growth from its investment in OpenAI and other AI projects.

All of this is to say, Zuckerberg's aggressive push into AI with Llama 3.1 is more than just a tech flex… it's a strategic move that could reshape the AI landscape. 

By offering a powerful, cost-effective, and open-source model, Meta is positioning itself as a key player in the AI arms race.

For investors, this signals that Meta isn't just riding the AI wave – it's aiming to create a tsunami.

The company's ability to rapidly innovate, coupled with its growing revenue from AI-powered advertising, suggests Meta could be a dark horse in the AI market. 

While other tech giants struggle to monetize their AI investments, Meta's approach could lead to sustained growth and market dominance.

This development matters because it shows Meta's potential to disrupt the AI industry, potentially offering better returns on AI investments than its competitors while literally giving away its AI tech. 

Investors should keep a close eye on Meta's AI initiatives, as they could be a significant driver of future value and market position in the rapidly evolving tech landscape.