The State of Open Source AI: Meta's New Global AI Alliance Against Major Players

Table of Contents

• Meta's Llama 2 Lags Behind Open Source Models
• Meta Lobbying for Looser Open Source AI Regulations
• Big Tech Still Seeking Clear Business Models
• Apple Enters with Powerful New MLX Framework
• Nvidia Joins Meta's Alliance Amid Threat

Meta's Llama 2 Lags Behind Open Source Models

The AI research community was recently surprised when Meta's Llama 2 model scored very low on the Hugging Face truthful QA benchmark test. With only 33.3 points, it ranked far below numerous open source models in the 7B parameter class, some achieving nearly double the performance.

This raises doubts about Meta's claims that their models represent cutting edge open source AI advancements. The Llama 2 model is not freely accessible either, requiring approval from Meta after submitting personal information and stating intended use cases.

Meta's Open Source Claims Questioned

Meta has been making bold claims that their models like Llama 2 represent industry-leading open source AI that furthers transparency. However, the poor performance of Llama 2 compared to other open source models, along with the restrictions Meta places on accessing it, undermine these claims. The AI research community views Meta's 'open source' branding as more of a marketing tactic than accurately reflecting the closed nature of their models. Meta appears to be promoting an open source image to gain public trust rather than truly embracing open source AI principles.

Llama 2 Performance on Leaderboard

The Hugging Face truthful QA leaderboard for 7B parameter models shows over a dozen open source models outperforming Meta's Llama 2, with top models scoring around 65 compared to Llama 2's 33.3. This benchmark indicates Meta has fallen behind the capabilities of open source AI models from the research community. Despite claims that Llama 2 represents cutting edge industry AI, it lags well behind other openly available models.

Meta Lobbying for Looser Open Source AI Regulations

Recent reports indicate that Meta is lobbying government regulators to exempt open source AI systems from accountability standards and obligations that would apply to other companies' AI models.

Critics argue this is a ploy for Meta to avoid regulations for its own 'open source' models by having them classified differently than proprietary AI from competitors like Google and Microsoft.

Seeking Exemptions from Accountability

Meta appears to be arguing that open source AI should be freed from accountability measures around safety, fairness, and transparency. They want open source models treated differently than AI systems from Big Tech companies, likely to give their own 'open source' models like Llama 2 a competitive advantage. But the research community feels that open source vs proprietary status does not inherently make an AI model more or less in need of accountability. Meta seems to just be exploiting the term 'open source' as a loophole to avoid regulation.

Big Tech Still Seeking Clear Business Models

Major technology companies investing heavily in AI like Microsoft, Amazon, and Meta have yet to demonstrate AI can drive profits outside of cloud computing services.

There is no clear business model emerging around monetizing AI itself, beyond using it to entice companies into cloud infrastructure contracts.

Cloud Profits Main Goal So Far

Microsoft and Amazon's big AI investments appear focused on getting companies locked into their cloud platforms, not directly profiting from the AI capabilities themselves. Microsoft's failed attempt to gain search engine market share with Bing shows AI itself does not inherently produce revenues. The real target is cloud revenue through AI bundled with cloud infrastructure.

Apple Enters with Powerful New MLX Framework

Apple recently open sourced its new MLX machine learning framework optimized specifically for its latest M3 chips.

Benchmarks show MLX crushing Nvidia GPUs for training large AI models, presenting a major new contender in the AI hardware space.

Optimized for Apple Silicon

MLX combines the best aspects of PyTorch and TensorFlow in a framework tailored for ideal AI performance on Apple's proprietary M3 processors. This positions Apple to leapfrog competitors by offering a optimized AI training solution tied tightly to its vertically integrated hardware and software stack.

Crushes Nvidia in Benchmarks

In benchmarks, Apple's new M3 Max chip on a laptop absolutely demolishes top-of-the-line Nvidia RTX 4090 GPUs in training large AI models, showing major promise for MLX and Apple silicon. This poses a serious threat to Nvidia's dominance in AI hardware, as Apple brings optimized software and silicon together in a uniquely integrated package.

Nvidia Joins Meta's Alliance Amid Threat

With Apple's emerging challenge in AI chips, Nvidia has now joined Meta's AI alliance likely as a defensive move.

This aligns Nvidia with Meta in pushing back against Apple's integrated hardware-software AI solution that could displace Nvidia GPUs.

FAQ

Q: Why did Meta create a global AI alliance?
A: To boost its position in the AI ecosystem against major players like OpenAI and Anthropic.

Q: How far ahead are Meta's AI systems?
A: Meta's Llama 2 model lags significantly behind open source models from the research community.

Q: What are Meta's goals with open source AI claims?
A: To lobby governments for looser regulations on open source models like its own, while restricting competitors.

Q: Do big tech companies have clear business models for AI?
A: Not yet beyond bundled cloud services, hence moves like Microsoft UK cloud deals.

Q: How good is Apple's new MLX performance?
A: In benchmarks, the new M3 chip with MLX crushes even Nvidia's latest GPUs.

Q: Why did Nvidia join Meta's alliance?
A: Likely due to the threat of Apple's powerful new MLX framework outperforming its AI chips.