Apple and Nvidia recently announced plans to join Microsoft as investors in OpenAI, the developer of the generative AI chatbot ChatGPT. The Wall Street Journal (WSJ), which reported the news, warned of the risks of this investment, as OpenAI will face tougher competition in the AI market than ever before.
According to the WSJ, the threat comes from other startups offering services at lower prices, as well as their own AI models that are more effective at solving narrower problems.
In particular, Meta Platforms sees opportunities in a new generation of AI startups. Its CEO Mark Zuckerberg positions his company as a champion of ordinary people, offering its advanced AI model Llama for free to third-party developers to use. Google has also released an AI model as open source.
In contrast, OpenAI charges end users and companies for access to its most powerful AI models. Many experts believe that users will eventually rely on a variety of AI models — some from closed-source vendors like OpenAI, others from open-source competitors championed by Zuckerberg.
Open-Source AI vs. Proprietary Models
In the meantime, open-source AI is already catching up with the AI pioneers, especially in everyday business applications that require consistent performance and low costs.
Meta announced this week that versions of its Llama AI model have been downloaded nearly 350 million times by software developers and users, up 10 times from a year ago. Meanwhile, OpenAI says its ChatGPT service now has 200 million weekly active users.
For many everyday purposes, AI models trained to do just certain tasks will be better and cheaper to use, according to Julien Launay, CEO of startup Adaptive ML. Adaptive ML is using Llama’s model to train smaller, customizable AI models for companies. They’re much easier to customize than giant, proprietary AIs like ChatGPT, Launay says. And the number of users of such models is growing. Companies like DoorDash, Shopify, Goldman Sachs, and Zoom, for example, have said they’re using open-source AI for tasks ranging from customer service to meeting debriefing.
Another example is Procore Technologies, a developer of a platform for managing complex construction projects. Earlier this year, Procore launched features that use large language models for the first time, relying on OpenAI’s ChatGPT, accessible through Microsoft’s cloud platform. Then Procore created software that makes it easy to use almost any AI across its entire system, eliminating the need for OpenAI’s ChatGPT service.
Challenges and Security Concerns
In cases like these, customer retention can be a challenge for OpenAI. Most companies don’t want to be tied to a single AI provider, and at least for now, switching between AI providers is easy. That makes OpenAI’s AI models competitors to the AI solutions the startup’s customers can now build themselves.
An OpenAI spokesperson, however, suggested that competition from open-source AI is increasing the pool of people who know how to use AI, which is also driving demand for OpenAI’s services.
When it comes to security concerns, opinions also differ sharply. Proponents of closed systems say they have the resources and controls to help prevent misuse of the technology, and that open-source tools can be exploited by bad actors. In turn, proponents of open source solutions argue that their systems make it easier to detect problems and deal with unintended harm that would be difficult to detect in closed systems, since they are available for public scrutiny, notes NIX Solutions.
We’ll keep you updated on the evolving landscape of AI investments and the competition between proprietary and open-source models as new developments unfold.