Can decentralized, open-source artificial intelligence models compete with proprietary ones like OpenAI's powerful GPT-4? The frequently asked question sparked a heated debate on Twitter after a former Google AI researcher took a side.
“If you think open source models are going to beat GPT-4 this year, you're wrong,” said Arnaud Benard, founder of Galileo AI, throwing down the gauntlet. Citing OpenAI's talent and resources and GPT-4's strong character as a product beyond LLM, he said open source projects may struggle to transition from challengers to AI champions.
Unsurprisingly, Benard's tweets have drawn a range of opinions, from vocal support to extreme disagreement.
Prominent AI enthusiast Ryan Casey, who writes the journal “Beyond the Yellow Woods,” offers a more optimistic view of the potential of open source AI, saying that “open source will not match or win.” [private models] This year, according to his calculations. “If there is demand, there will be innovation.”
On the other hand, Jeremy Traguna, an AI strategist, said that “OpenAI models continue to move.” In other words, while open source models may be dealing with GPT-3.5 in the GPT-4 era, GPT-5 will likely exist when we have a generic LMS comparable to GPT-4.5 Turbo.
Technology analyst John Howells believes resources aren't the only standard that differentiates closed-source LLMs.
“Mistral is heavily funded, has a great team, and recently released a GPT-3.5-beating open source model,” he wrote. “They or a similar outfit will release an open source model of the GPT-4 standard later this year.”
Mistral AI, a French startup, has gained recognition after releasing Mixtral LLM, which offers improved performance over GPT-3.5 in many applications.
Technium, the founder of Nous Research, made an important but philosophical point in a series of discussions. “Each capability increase in an operating system (open source) is a permanent thing that can never be removed from the world and will be used safely forever,” he said. Basically, as long as there is some progress in open source AI technology, no company can restrict its access.
Open or closed? An endless debate
The open-source vs. closed-source debate is reminiscent of the early operating system wars between Windows and Linux. The ML School's Santiago Pino writes that proprietary AI models may defeat general consumers like Windows, but open-source software offers enterprise users far more customization and control.
Pino points out how many companies start with ChatGPT experimentation but migrate to open source models, which can be fine-tuned and customized for their needs and data compliance requirements. Open source solutions remove vendor lock-in and provide transparency, he said.
“Closed proprietary models may win over individuals, but most companies don't want to send their data to Microsoft or Google. They want control. Open source models are the answer,” Barnard said on Twitter in the days before the thread went viral.
Such is the attitude of Sciumo Inc., a software development company. It was shared in a debate about Bernard Twitt, a software development company, who emphasized the great potential of open source models: “(Open source models) compete where it matters: domain-specific problems with domain-specific data and knowledge that OpenAI does not have.”
Furkan Gozukara, a computer engineer known for his YouTube channel Secours, is among the most deviant. Speaking to Decrypt, he agreed with Bernard, “On certain tasks, open source LLMs outperform OpenAI.
Gözükara gives an example of a company that “trains LLM on its own documents.” Yes, OpenAI has the ability to customize GPT based on specific guidelines and documentation, but handling sensitive data to third parties is always a concern. That concern was recently confirmed when it was discovered that personalized GPTs could provide sensitive information to third-party users.
“Open source AI foundational models destroy closed and proprietary AI models,” Yan Lekun, head of AI development at Meta and a staunch defender of open source, has repeatedly said. Google, another AI giant, also recognizes the threat of the open side. -Source AI: “Open source models are faster, more customizable, more private and pound for pound more powerful,” says a Google memo from 2023.
It remains to be seen whether open source models will match or surpass GPT-4 this year and future iterations. However, expert opinion from both sides reveals an interesting tension. Closed-source models may have an edge in resources and rapid iteration, but open-source tools are rapidly evolving, offering enduring capabilities and customization. For now, the AI community can watch the race unfold and enjoy the benefits of using the best technology available.
Edited by Ryan Ozawa.
Stay on top of crypto news, get daily updates in your inbox.
