This is a good interview with Mozilla's CTO Raffi Krikorian. Mozilla is the company that distributes the open Firefox browser, which is what I use daily. He "explains why businesses are turning to open-source and open-weight AI models for greater control over data, costs, and customization." I think there's going to be a lag in open weight models (much lss fully open source models) because the cost of running them is still too high. The Canadian Cohere model called Command A Plus requires at least one B200 GPU set, which costs $6 per hour to run (or $45K to buy). Now I remember when Compuserve cost $6 per hour to access, back in the 1980s. So we can imagine a world in which this is attainable for most people. But it won't be soon. Meanwhile, though, it makes sense to focus on the creation of the open source infrastructure we need to make sure that when the time comes, it actually is available to all. Image: Nvidia. See also: What's going on with open weight models?
Today: Total: [Direct link] [Share]
