Open-source LLaMA may completely change the current landscape of large-scale model competition.
The OpenAI large model competition is heating up, with Meta joining the fray by commercializing open-source large models.
According to the Financial Times on Tuesday, Meta is preparing to release a commercial version of its large language model LLaMA, which will allow startups and companies to build custom OpenAI software based on this technology.
This move will pit Meta against OpenAI supported by Microsoft, Alphabet-C, and others, with the latter two companies currently leading the competition in developing generative AI.
Yann LeCun, Vice President and Chief OpenAI Scientist at Meta, stated at a conference in Aix-en-Provence last Saturday:
The competitive landscape of artificial intelligence will change completely in the next few months, perhaps even in the next few weeks, and we will see open-source platforms that are just as good as closed-source platforms.
Open-source LLaMA model to be commercialized soon
As previously mentioned, in February of this year, Meta released its open-source large language model LLaMA, which helps researchers develop proprietary OpenAI software at a low cost, causing a sensation in the OpenAI field. Three insiders said that the new version will be more widely available to enterprises and can be customized by them.
In stark contrast to competitors like OpenAI, LLaMA is an open-source model. Essentially, Meta treats its OpenAI technology as open-source software (computer code that can be freely copied, modified, and reused), providing everything external parties need to quickly build their own chatbots.
Commercializing LLaMA not only benefits Meta's competitiveness, but also provides OpenAI developers with an alternative to the proprietary software sold by competitors Alphabet-C and OpenAI, indirectly benefiting Meta's OpenAI development as well.
The advantages of open-source models include higher user acceptance rates and the ability to input more data for AI processing. The more data LLM has, the more powerful its capabilities become. In addition, open-source models enable researchers and developers to discover and address vulnerabilities, while improving technology and security. Meta and other tech companies have faced scrutiny over various privacy and misinformation scandals over the years.
Although Meta's technology is currently open-source and free, two insiders said that the company has been exploring commercialization. One insider stated that there are currently no plans for charging, and Meta will not do so in the upcoming release.
Joelle Pineau, Vice President of OpenAI Research at Meta, declined to comment on commercialization but stated:
These things are indeed open source, but you won't completely give up the intellectual property of this work.
The competition landscape of large models may face huge changes
Although providing software for free seems to contradict making money, experts believe that companies can also use this strategy to open up the market. An insider familiar with Meta's ideas said:
Meta has realized that they are lagging behind in the current AI hype cycle, but this has given them a way to open up the ecosystem, and they seem to be doing the right thing.
Earlier this week, Meta's Global Affairs Chief, Nick Clegg, praised the benefits of open source in an article, saying:
Open source is the best antidote to dispel fears about OpenOpenAI. Of course, open source also helps Meta catch up with competitors because it allows companies of various sizes to improve this technology and build applications on top of it.
However, Meta will also face risks. In the process of commercializing the next version of LLaMA, if it is found that copyrighted materials were used to train the software, the company may face legal action.