If you want proof, the power of the open source movement in AI has been highlighted this week, with start-ups and scale-ups in Europe and the UK in a very competitive position despite the immediate threat. increase. EU Artificial Intelligence Lawis due to be passed later this year.
This week Meta created Llama 2, an open-source commercial version of their Large Language Model (LLM). Free for research and commercial use.
attract Broad industry supportthis release is a step forward in democratizing AI, decentralizing access to necessary tools for a few non-large companies, and accelerating the scope of innovation.
Meta previously released an early iteration, LLaMA, in February. It is a collection of models capable of generating texts under an open-source, non-commercial license for research purposes, granting access to scholars on a case-by-case basis. However, the code was leaked online with download instructions. GitHub and 4chan.
a leaked notes The paper, written by a member of the Google staff, claims that as a result, “they’ve acquired virtually the entire planet’s worth of unpaid labor.”
But it makes another important statement. As an employee writes, “I’m eating lunch,” it means that smaller, fine-tuned models can perform better than larger ones.
Success of open source AI
Open source start-ups and scale-ups in Europe and the UK are successfully transforming AI, even in the early stages. Europe has a number of startups in this space, including:
There are also non-profit research institutes Eleuther.ai.
talked to Florian Liebert general partner 468 Capital is a VC firm based in Berlin and San Francisco.
In his previous roles, he founded: and lead open source cloud infrastructure company Mesosphere (now D2iQ) has been CEO for over six years and has raised $252 million in funding.
He sees growing interest in open AI. We argue that the emerging MLOps category is an important focus area.
“We are experiencing first-hand what open source can do and what it does to transform and accelerate. It’s been replicated in the world of MLOps.
We expect the landscape in and around LLM to be much larger than what we have seen with DevOps. ”
This includes investments in German AI companies Aleph Alpha who raised $100 million this year.
Strengthen the human resources and presence of AI companies in Europe, Aleph Alpha has developed a ChatGPT competitor with an AI assistant that can create, understand and rate text.
Leibert notes that while the focus of investment is “probably still in the United States,” open source offshoots of big companies are growing significantly.
Let’s not forget that we saw a French company earlier this year MistralThe project, which features graduates from Deepmind, Hugging Face, and Meta open source solutions for enterprises, has raised $100 million in just a few weeks of existence.
The UK’s most successful open AI code repository is probably one you’ve never heard of.
In the UK, the UK’s leading open source AI repositories are: important gravity.
We are a small Scottish company founded in 2018. automatic GPTa semi-autonomous variant of ChatGPT.
With a community of over 250 contributors, the Significant Gravitas repository is currently the #1 UK repository, the second AI project to reach 100,000 stars on GitHub, and the #1 code repository in the UK.
This surpasses industry heavyweights such as StabilityAI, Google Deepmind, and Unity.ai.
Data compiled by Runa Capital Open UK A recently published report, State of the UK Open in 2023: Show Me the Money – AI Opennesstracking the opportunity for the UK to lead in the liberalization of AI.
OpenUK is an NFP organization focused on open technology – software, hardware and open data – across the UK. The aim is UK leadership and global cooperation in open technology.
We spoke with CEO Amanda Bullock. Regarding open source, he declared, “How to collaborate is what every space should be doing right now. There are lessons about how to collaborate and compete at the same time.”
For open source AI, this includes:
.. “Understanding intellectual property management, how to bring competitors to the bargaining table, how to save costs and work together to build something that can compete at a higher level.
Looking across sectors, there is room for sectors to work together to reduce much of the cost of innovation and improve the quality of innovation. There is also the capacity of communities to work together over the long term across geopolitical borders. ”
But there are two issues the open source community is wrestling with: how to define and regulate open source AI.
But what does open source mean?
One of OpenUK’s efforts is to standardize what open source really means.What Happens When Community Practices Are Accepted changeor will open source shut down?
Earlier this year, Elon Musk said, “OpenAI was built as open source (that’s why I named it ‘open’ AI) and was a nonprofit competing with Google, but now it’s closed source. I’m sorry,” he lamented. It is the largest profit company virtually controlled by Microsoft. ”
OpenAI was founded as an open source (that’s why I named it “Open” AI), non-profit company to compete with Google, but is now a closed source biggest profit company, effectively owned by Microsoft. Managed.
It wasn’t intended at all.
— Elon Musk (@elonmusk) February 17, 2023
Even at OpenUK admit Meta’s Llama 2 license is open source initiativein its current form, Definition of open source.
However, OSI is currently engaged in a consultation to define “open source AI,” seeking a set of shared principles that replicate open source permissionless, pragmatic and simplified collaboration for AI practitioners. doing.
How should open source AI be regulated?
And when it comes to AI regulation, open source researchers, practitioners, startups, and investors are concerned.
EU Artificial Intelligence Law It is expected to be put into practical use later this year. but, concern We are sticking with the open source AI community on how the regulation will be applied and how it will affect European companies.
2022 investigation Over 100 European high-tech AI start-ups and 15 European VC firm participants developing diverse AI systems to drive AI innovation reveal their concerns about the new law.
50% of start-ups surveyed are concerned that the AI law will slow down AI innovation in Europe, and 73% of VCs surveyed believe that the AI law will slow down the growth of European start-ups. We expect our competitiveness in the AI field to decline or significantly decline.
Additionally, approximately 33-50 percent of the start-ups surveyed would be subject to EU AI law obligations, as the current classification would classify their AI systems as high risk.This is beyond the EU’s expectations Impact evaluation AI law provisions (5-15 percent).
Understandably, GitHub CEO Thomas Dohmke has concerns about the company’s authoring. position paper About this bill.
In a speech earlier this year, EU Open Source Policy Summit In Brussels, he said open source is at the core of AI development, with tools such as: Copilot X on GitHub :
“Open source frameworks such as PyTorch and TensorFlow power nearly all AI, and open source tools have long contributed to increasing transparency and removing bias.
And now AI models are built and shared in open source. The OSS community has been and will continue to be important to the progress of AI. ”
He argues that open source developers should be exempt from AI law.
“The open source community is not a community of entities, it is a community of people.
The burden of compliance should be placed on the company that ships the product. OSS developers are often volunteers. Many people have two jobs. They are scientists, doctors, scholars, professors and college students alike. They usually do not intend to profit from their contributions.
And they certainly don’t have a big budget or a compliance department!”
OpenUK also highlights the larger challenge of inconsistent country-to-country ML regulations and judicial decisions.
“This is bad for open, which benefits from a global community. If the community wants an open model for the US, for Europe, for China, etc., many of the collaborative advantages of open are lost. It will be.”
Despite the imminent threat of EU artificial intelligence law, European and UK start-ups and scale-ups are now in pole position as global leaders in open source AI.
Going forward, maintaining momentum in the open source AI movement and ensuring a vibrant and innovative AI ecosystem in Europe and the UK will require collaboration, competition, and practical (but not overly onerous or punitive) ) promoting regulation is essential.
