Since 2022, America has had a solid lead in artificial intelligence thanks to advanced models from high-flying companies like OpenAI, Google DeepMind, Anthropic, and xAI. A growing number of experts, however, worry that the US is starting to fall behind when it comes to minting open-weight AI models that can be downloaded, adapted, and run locally.
The US does not have the talent pool to pull from to compete with China. This generation of China is the generation that grew up with record investment in education and infrastructure. Our generation is the generation of mass education institute attacks, de-funding and total neglect of infrastructure. We’re at a mismatch from differing policies and culture. We can’t bridge this gap. It’ll take several generations IF we do the right things.
Hard to fall behind what? None of them is making anything interesting. Best they can do is provide some text that sound superficially plausible, is statistically correct and yet have 0 reasoning.
Nobody is “ahead” of anybody except is managing to do so with even more data while wasting even more resources.
Maybe more importantly of the participants in that race demonstrated that to keep on doing so will actually solve any of the problems that have been discovered along the way.
the us needs to invest in education, healthcare, and infrastructure if it wants to “beat china”
No, the US needs to really heavily regulate AI before it destroys every line of work other than slinging a mop or working a skilled trade, and before it destroys everyone’s water supply and potentially the power grid itself.
In fact, if the EPA wasn’t gutted, AI companies would be in deep shit right now for how they’re ravaging the environment with their datacenters currently.
Open-source AI is the data, not the weights. Weights alone isn’t useless, but it’s not a platform to truly understand and modify a network. No one wants to open their data though, because that’s both what gives them a competitive edge and it would be an admission of widespread copyright violations.
Having the weights is a big plus though. Especially seeing how people can even train models using the best models as synthetic data.
Using the best models as synthetic data is mostly pointless. You’re just going to recreate all of its biases and failures in a degraded copy. The whole point of open source software is being able to analyze the source code to learn how it works and understand and ideally remove its weaknesses.
Open weights doesn’t let you do that, and what research it enables is mostly just tinkering around the edges. If someone trained a network, but it kept saying racist stuff, you can’t figure out why it’s racist or rebuild it without the racism from weights alone. Just the weights is like having a binary. Maybe nice to have a gratis app to use, but not really open.
