Open-source software has been one of tech's biggest-ever success stories. No surprise, then, that developers of open-weight AI models try to borrow some open-source sunshine.

But the two approaches aren't the same.

Meta Platforms chief Mark Zuckerberg on Monday became the latest tech executive to draw an equivalence between the two, calling open-weight models "open source AI" in a 6,500-word essay posted by the company. Such models are important for U.S. national security, he said, adding that Meta would be releasing more "open source" models after a hiatus.

Meta, Zuckerberg said, was "strongly supportive of open source, including open source AI models."

Here is the problem, though: none of the models Zuckerberg is talking about is actually open-source. And the distinction matters.

Open-source, as the name suggests, refers to source code that is available in full for anyone to take, use or modify. The community of people using the software can suggest changes that make it better. If the maintainers of an open-source project accept them, they are incorporated into the code.

That model works especially well in cases where a large number of people need a piece of software, but where that kind of software doesn't differentiate them from competitors. The computing infrastructure that drives the internet, for example, mostly uses the open-source Linux operating system. That saves time and money for network operators who don't want or need an operating system with lots of bells and whistles. They would rather focus on other things.

Open-weight models aren't quite as open, and don't serve the same philosophical purpose.

With open weights, a company -- Meta, for example -- trains a model using methods that aren't disclosed to its users. Unlike with open-source software, users can't see the code used to train the model, nor can they retrain it themselves or add new data. Its level of intelligence is essentially fixed by whoever trained the model.

Its weights, though, are open for users to play with, giving the model a level of adaptability . Weights are the results of training -- numerical values that influence how the AI's digital brain works. Tweaking them later can vastly alter how a model behaves.

As with open-source software, modification of weights can be useful. Users can tailor weights to their purposes, making an AI model specialize in one narrow area, for example, and reject questions outside that domain. Modifying weights can make a model more polite, sassier or more prone to hallucination. Of course, modification can be harmful, too, if bad actors with sufficient computing resources change or remove weights that block the AI from executing cyberattacks or giving instructions on how to make weapons.

That is about where the similarities end, though. Open-source software projects are the product of a community of developers and open for the world to see. Open-weight models are closely tied to the company that creates and promotes them.

Meta decides when it trains the next generation of its open-weight models, and its capabilities -- not those of the community -- determine the models' basic level of intelligence. Perhaps most important, Meta would remain in a better position than anyone to benefit from those models.

This is due to its familiarity with how they are built and its willingness to invest in the computing power it takes to deploy them to users. The balance of power in open-weight models is much more centralized.

Meta, of all companies, knows that open-source software isn't quite the same. It has contributed extensively to the open-source universe, including launching hugely popular open-source projects called React and PyTorch.

Zuckerberg in his Monday essay wrote of a world where "open-source" models allow anyone and everyone to harness AI's power and bend it to their own purposes. There may be a grain of truth in that, but it is also self-serving.

This is an edition of the WSJ AI & Business newsletter, a weekly digest to help you make sense of AI's impact on business with news, insights and data from our global team of technology journalists. If you're not subscribed, sign up here.

Nvidia and Wall Street Heavyweights Reach $500 Billion Financing Deal

Nvidia is teaming up with some of the largest Wall Street asset managers in what they hope will become a big new investment opportunity and a huge injection of capital into the AI computing market. The deal aims to create pools of capital that will be available to Nvidia's customers at "attractive rates," suggesting Nvidia will use its financial strength to support further purchases of its chips. The move comes at a time of growing investor worry that Big Tech spending on AI is getting out of hand, although outlays on computing infrastructure have hardly lost steam. Nvidia's sales are projected to be near $400 billion in its current fiscal year, which ends in January, according to FactSet.

The Number

Average asking rent for the San Francisco metropolitan area, the highest in the U.S., driven by big AI salaries distorting the housing market.

What the Humans Are Saying

AI in Charts

SpaceX's appeal to investors has largely rested on its rocket-launch capabilities and its Starlink broadband satellite-internet service. In the future, though, it looks increasingly like AI is going to take center stage.

That shift, at least in the near term, has a lot to do with the company's growing investment in AI computing. Demand for computing power is sky-high, and Elon Musk's company is renting its infrastructure out under cloud-computing deals with the likes of Anthropic and Google.

Analysts now expect revenue in SpaceX's AI division to rise to nearly $15 billion a quarter by the middle of next year, far above the roughly $2.5 billion the company last week reported for this year's second quarter. The company still has a long way to go to reach Musk's goal of $1 trillion of revenue by 2030. AI and the company's next-generation Starship rockets will be critical to that aim.

AI in the Wild

The AI boom is altering South Korea's dating scene as young engineers weigh whether to tell prospective partners they work at Samsung or SK Hynix, where they're in line for big bonuses. Once seen as nerdy and questionable marriage material, employees of those highflying memory companies are now at least on par with lawyers, doctors and accountants in the dating pool.

Other Highlights From the Week in AI

   -- Intel is raising $20 billion through a stock offering to expand 
      production and cater to skyrocketing AI-driven demand for its chips. 
 
   -- Leopold Aschenbrenner is making a new $400 million bet on a chip startup 
      after his hedge fund, Situational Awareness, sold most of its stock 
      portfolio. 
 
   -- SK Hynix is spending $38 billion to expand its memory-production capacity 
      during an unprecedented demand boom. 

About Us

WSJ AI & Business is a weekly look at AI's transformation of the business world. This newsletter was curated and edited by Asa Fitch. Reach him at asa.fitch@wsj.com (if you're reading this in your inbox, you can just hit reply). Got a tip for us? Here's how to submit.