What Llama 3 can teach us about AI: Why Open AI is better than open AI at Meta, Why OpenAI is better?
Jerome Pesenti has a few reasons to celebrate Meta’s decision last week to release Llama 3, a powerful open source large language model that anyone can download, run, and build on.
Pesenti used to work for Meta and would often push for the release of the company’s technology to other people. His new startup will be able to get access to the machine learning model that he says is very close in power to the one OpenAI has, but it will be a lot cheaper to run and be much more open to modification.
Sizzle’s story may augur a broader shift in the balance of power in AI. OpenAI changed the world with ChatGPT, setting off a wave of AI investment and drawing more than 2 million developers to its cloud APIs. Developers and entrepreneurs may stop paying for access to the latest version of OpenAI or one of the more powerful open source models if they see that open source models are competitive.
Open models were dropping at a very fast clip. I went to Databricks to see the final stages of the attempt to build a language model that was briefly the best open one around. That crown is now Llama 3’s. Ali’s comments about the game-changing qualities of the larger model of the GPT 4 are similar to those of the CEO of Databricks.
According to Pesenti, running Llama 3 on a cloud platform can save you 20 percent of the cost of accessing GPT-4 through an API. He adds that Llama 3 can be configured to respond to queries extremely quickly, a key consideration for developers at companies like his that rely on tapping into models from different providers. “It’s an equation between latency, cost, and accuracy,” he says.
Llama 3 also showcases the potential for making AI models smaller, so they can be run on less powerful hardware. Meta released two versions of its latest model, one with 70 billion parameters—a measure of the variables it uses to learn from training data—and another with 8 billion. The smaller model is compact enough to run on a laptop, but has plenty of capacity in WIRED’s testing.
A French artificial intelligence company founded by alumni of Pesenti’s team at Meta was open source two days before Meta’s release. It has 141 billion parameters, but only 39 billion of them can be used at once, a design called a mixture of experts. The model is more capable than other models that are larger because of this trick.
Of course all this openness is not purely altruistic. Meta CEO Mark Zuckerberg says opening up its AI models should ultimately benefit the company by lowering the cost of technologies it relies on, for example by spawning compatible tools and services that Meta can use for itself. Meta may benefit to prevent OpenAI, Microsoft, or Google from dominating the field, as he didn’t mention.
Alphabet CEO Sundar Pichai is confident that Google will find a way to make money selling access to generative AI tools. The CEO of Microsoft said that his company was already doing it.
The companies reported better-than- expected results on Thursday. With the results, the stock prices of both soared, with the stock prices of Alphabet jumping on plans to buy more shares and pay a dividend.
Microsoft Grows $200 Billion$ Using OpenAI, Openai, and Copilot: Data Collectivity and Machine Learning for Generative AI
In a call with financial analysts on Thursday, the Microsoft CEO claimed that his company now has over 1.5 million customers for the GitHub Copilot, a generative artificial intelligence tool. There were 1.3 million customers a quarter ago.
Sixty five percent of Fortune 500 companies use Copilot for Microsoft Office to help workers write emails and documents, while 65 percent use a Microsoft Azure Cloud service to access OpenAI’s generative AI software. “Azure has become a port of call for pretty much anybody who is doing an AI project,” Nadella said. Microsoft has invested $13 billion dollars in Openai, which helped win some clients.
The buzz of interest in AI services helped drive revenue for Microsoft’s biggest unit–cloud services–up by seven percentage points compared to a year ago, and Microsoft’s overall sales rose 17 percent to nearly $62 billion. It also gained cloud market share, Nadella added. The amount of deals Microsoft landed during the quarter increased by 80 percent compared to the same period a year ago.
Pichai didn’t offer any details about how many people had signed up for the $20 a month plan that came out in February.
The revenue figures related to the experiments on the core business of search were not shared by Pichai. If people spend less time doing other searches and more time searching for answers, they might not see more search ads. The types of ads Google does show also could have to shift.
Both fronts had little concern expressed by Picahi. He said they were confident they could manage how to serve the queries. I’m confident that we will be able to manage the monetization transition here too. Over time, it will play out.
Source: Google Thinks It Can Cash In on Generative AI. Microsoft Already Has
Microsoft has spent $mu$-$times$ $mathrmSecurity$ in the past few years
About the same amount was spent as Microsoft invested in infrastructure last quarter. Microsoft’s results and comments on Thursday confirm that the company is closer to delivering a payoff.