More accessible artificial intelligence systems are proliferating, posing a tough challenge to the creator of ChatGPT.

Apple, Nvidia, and Microsoft are in talks to invest in OpenAI, the creator of ChatGPT, at a time when it is already clear that the company will face tougher competition than ever in the booming artificial intelligence (AI) market.

A large part of this new competition comes from emerging companies that promise to undermine OpenAI’s services with alternatives that could be more cost-effective to use and also better at certain specific tasks.

At least one tech giant sees promise in the new wave of AI startups. Mark Zuckerberg, CEO of Meta Platforms, the parent company of Facebook, is positioning his company as a champion of small creators by allowing independent developers to use Meta’s advanced AI model, “LLaMA,” free of charge. Google has also launched an open-source AI that is not as capable as Meta’s. In a letter in July, Zuckerberg argued that this open-source approach “will ensure that more people around the world have access to the benefits and opportunities of AI” without concentrating power in the hands of tech giants. Open-source software can be used commercially by nearly anyone. Examples include the Android operating system, developed by Google but available for any creator to use on mobile devices without payment. This contrasts with the more typical “closed” approach adopted by companies that control who can use their software. Microsoft, for example, charges creators a licensing fee to install its Windows operating system on their computers. Apple does not allow other companies to use its iPhone or Mac operating systems.

To a large extent, OpenAI falls into the latter group, charging end users and businesses to access its most powerful models.

Many experts believe that eventually, we will all have access to a variety of AI; some from closed providers like OpenAI and Google, and others from open-source competitors that Zuckerberg is advocating for. The nature of this mix will determine whether it was worth it for companies to spend tens of billions of dollars on creating advanced AI.

The latest example of this investment: Apple and Nvidia are in talks to join Microsoft in its initiative to invest in the next funding round of OpenAI, which would value the company at $100 billion.

Meanwhile, open-source AI is catching up to the major pioneers, especially in everyday commercial uses that require consistent performance and low costs. Meta announced that software creators and enthusiasts have downloaded versions of LLaMA nearly 350 million times, 10 times more than a year ago. A direct comparison of these figures with ChatGPT is not possible, but OpenAI claims that the ChatGPT service now has 200 million weekly active users.

For many everyday applications, AIs trained to perform specific tasks can be better and cheaper to apply, says Julien Launay. His startup, Adaptive ML, uses LLaMA to train small, customized AIs for businesses. Users can more easily customize these smaller AIs than large, closed AIs like ChatGPT, he adds.

DoorDash, Shopify, Goldman Sachs, and Zoom are among the companies that have stated they use open-source AI for tasks ranging from customer service to meeting summaries.

Procore Technologies, which creates a platform to manage complex construction projects, is a good example of how a variety of both closed and open AIs can be used for real-world tasks, from estimating costs to coordinating actual construction work.

AI can be useful at many points in that process. Earlier this year, when Procore first deployed features using large language models, the company relied on OpenAI’s ChatGPT, which it accessed through Microsoft’s cloud platform, says Rajitha Chaparala, Vice President of Product and AI at Procore. This type of AI used to be expensive, but prices have plummeted over the last 12 months.

However, Procore has now developed software that makes it easier to use almost any AI throughout its system.

This illustrates one way in which customer retention could be a problem for OpenAI. The majority of companies do not want to be tied to a single AI provider, and for now at least, switching from one to another is relatively easy. This pits OpenAI’s models, even as they become more affordable, against those that its customers can now develop on their own.

Open-source AI may make sense when applied to individual devices like the new PCs and smartphones that enable AI, says an OpenAI spokesperson. In general, the company welcomes competition for any of its services, he adds, because it is confident that it is better positioned to offer the capabilities, prices, and performance that software creators want.

All this open-source AI competition increases the pool of engineers who know how to use this technology, which also translates into a growth in demand for OpenAI’s services, the spokesperson explains.

An even higher level of transparency than what most open-source AI models offer will be needed when it comes to AI systems for sensitive fields like medicine and insurance, says Ali Farhadi, CEO of the Allen Institute for Artificial Intelligence.

In February, the Allen Institute, a nonprofit research group dedicated to solving world problems through AI, launched its own open-source AI and took the unusual step of also releasing all the data on which it was trained and all the steps involved in fine-tuning the model to provide better responses.

When it comes to broader concerns about security, opinions vary on which approach is more likely to prevent the worst-case scenario of AI becoming the malevolent, all-powerful intelligence like Skynet in the “Terminator” movies. Proponents of closed systems argue that they have the resources and control to help prevent the misuse of technology and that malicious actors could exploit open-source tools. Open-source advocates claim that their systems are subject to public scrutiny, allowing them to more easily detect issues and address unintentional harms, which would be harder to identify in closed systems.

However, long before we reach the point of battling against powerful AI overlords, those creating systems with more modest power will have to prove that their enormous investments are worthwhile.