Why 'open' AI systems are closed, and why this matters

This is a worthwhile read (especially the conclusion) for anyone worried about AI safety, the concentration of power, privacy abuses, censorship, and the likelihood for a few megacorps to take all the gains from this ongoing technology revolution. I am not sure if I agree with everything they say and conclude with, but think this is an important quote:

“Unless pursued alongside other strong measures to address the concentration of power in AI, including antitrust enforcement and data privacy protections, the pursuit of openness on its own will be unlikely to yield much benefit. This is because the terms of transparency, and the infrastructures required for reuse and extension, will continue to be set by these same powerful companies, who will be unlikely to consent to meaningful checks that conflict with their profit and growth incentives.”

We need a total rewrite of antitrust laws, privacy laws, copyright laws, and aggressive enforcement to avoid the coming concentration of power and information. Until then, it’s pretty disappointing to see everyone from Yann LeCun of Meta to Clem from Hugging Face misuse the term “open source” for mostly closed systems that only share the weights (the output of the “compilation” process that is training). Meta/LeCun are basically open washing for their own gain. In contrast, AI2’s OLMo is an example of what real open source looks like:

https://venturebeat.com/ai/truly-open-source-llm-from-ai2-to...