Join top executives in San Francisco on July 11-12 to hear how leaders are integrating and optimizing AI investments for success. Learn more
Hugging faceThe fast-growing New York startup that has grown into a central hub for open-source code and models reaffirmed its status as a leading voice in the AI community on Friday, drawing more than 5,000 people to a local gathering celebrating open-source source technology at the Exploratorium in downtown San Francisco.
The meeting was born by chance three weeks ago, when the charismatic co-founder and CEO of Hugging Face, Clement Delanguetweeted that he planned to be in San Francisco and meet others interested in open-source AI development.
Within days, interest in the informal meeting grew. Enrollments numbered in the thousands. In the last week before the event, Delangue booked the Exploratory museum, one of the few venues still available that could support thousands of people.
He turned the informal gathering into a huge showcase and networking opportunity for those who were fascinated artificial intelligencefrom real-world researchers and programmers to investors, entrepreneurs and the curious.
“We just passed 1,500 registrations for the Open-Source AI Meetup!” Delangue said in a text blast to the RSVP list a few days before the event. “What started with a tweet could lead to the largest AI gathering in history.”
The event took place against the backdrop of a growing debate about large language models (LLMs) and their applications. Critics have raised concerns about the potential monopolization and commodification of closed LLMs by Open AI and other companies, such as Google and Microsoft.
Open LLMs, on the other hand, are trained on common web data and serve as a substrate for downstream applications to build on. The open source community views LLMs as a public good or community resource, rather than a private product or service.
Open-source AI is having a breakthrough moment
Attendees began pouring into the Exploratorium around 6 p.m. Friday and continued to come for hours. They were a striking mix of ages, races and backgrounds, including retirees, parents, engineers and large groups of twenty-somethings dressed in a wide variety of clothing – from ball gowns to baggy jeans – a broad mix of high fashion and streetwear. The atmosphere was full of energy and the crowd was buzzing with excitement, similar to a music festival.
In brief remarks, Delangue addressed those in attendance and said the turnout was a testament to growing mainstream interest and excitement open source AI development. He said Hugging Face’s mission was to make state-of-the-art AI accessible to the widest possible audience while increasing transparency across the entire ecosystem.
“We expected maybe a few, 100 people to show up,” Delangue said in an address to those in attendance. “We have 5,000 people tonight. That is amazing. People call it the ‘Woodstock of AI’.”
“I think this event is a celebration of the power of open science and open source,” said Delangue. “I think it’s really important for us to remember that in the field of AI we’re where we are today thanks to open science and open source.”
“If not for the ‘Attention is all you need‘paper, for’The BERT‘paper, and for the’Latent diffusionpaper, we might be 20, 30, 40 or 50 years away from where we are today in terms of capabilities and possibilities for AI,” he said. “Without open source libraries or languages, without frameworks like PyTorch, TensorFlow, Keras , Hugging Face, transformers and diffusers, we wouldn’t be where we are today.”
“Open science and open source [are ways] to build a more inclusive future, with less concentration of power in the hands of a few, more contribution from underrepresented populations to fight prejudice, and overall a much more secure future with the involvement of civil society, from non-profit organizations , from regulators to all the positive impact we can have with AI and machine learning,” added Delangue. “And that’s what we’ve seen on Hugging Face: the impact of open science open source. All of you in the room contributed to over 100,000 open models on the platform.”
The battle between open and closed LLMs
In recent weeks, a high-stakes debate has emerged over whether new major AI models should remain proprietary and commercialized or instead be released as open-source technologies.
On the one hand, researchers argue that transparency reduces risk and commercial pressure to deploy AI before it is ready; on the other hand, companies say secrecy is necessary to profit from and control their technology. The issue has come to a head in recent weeks LLMs are starting to sound the alarmbut there is still no consensus on whether open science or commercialized AI will yield more reliable systems.
On Wednesday, three days prior to the open-source AI event, there was a highly contentious open letter calls for a six-month pause for large-scale AI development were circulating in the AI community. The letter was signed by high profile names such as Elon Musk, Steve Wozniak, Yoshua Bengio, Gary Marcus and several thousand other AI experts, researchers and industry leaders.
“I think OpenAI has done an incredible job of advancing the state of the art. I think they develop large language models first via GPT-2 and GPT-3 – and then the InstructGPT or ChatGPT-like model following the instructions. So I think those are at least two big breakthroughs that OpenAI has been responsible for. Andrew Ngone of the most influential voices in machine learning over the past decade, said in an interview with VentureBeat.
“At the same time, I feel like I’m also excited about all the open language models being released,” he added. “But I think it is very reasonable for different companies to choose different policies for different reasons. I’m excited about the very open models and thankful for all the researchers who publish open models, but I’m also thankful for all the work OpenAI has done to get this out there.”
The path to ethical AI probably depends on the balance between scientific openness and trade secrecy. But that balance clearly remains elusive, and the future of AI is at stake. How tech companies and researchers work together – or not – will determine whether AI uplifts or endangers our lives. The stakes are immense, but so are the challenges of navigating this debate.
VentureBeat’s mission is to become a digital city plaza where tech decision makers can learn about transformative business technology and execute transactions. Discover our Briefings.