pull down to refresh
So you're saying that if right now I open Claude, the reason that it is capable of doing what I want is because of USG datasets? How about GLM? It is more capable than GPT-5.2. Made in China. Is this because GLM has access to the USG's datasets too?
Remember that "Build World-Class Scientific Datasets" was stated in future tense last July:
High-quality data has become a national strategic asset as governments pursue AI innovation goals and capitalize on the technology’s economic benefits. Other countries, including our adversaries, have raced ahead of us in amassing vast troves of scientific data. The United States must lead the creation of the world’s largest and highest quality AI-ready scientific datasets, while maintaining respect for individual rights and ensuring civil liberties, privacy, and confidentiality protections.
So you're claiming here with a straight face that between the release of Claude 4.0 and now, 4.6, this model is good because it was post-trained on a new dataset that didn't exist in July, made by the USG not industry?
So yes if the AI gets cut off from the cutting edge data sets it is going to degrade. If the AI isnt able to be trained on the highest quality data for specific tasks like what the DOE and DOD wanted to use it for then yes it will not be as high of a quality. The reason Claude works so well is because of the data. With the US Government now opening up these data bases to companies they have the ability to dramatically improve their models. If Anthropic doesnt get access they will fall behind.
The DOE data is entirely different than regular run of the mill scraped from the internet data. Our adversaries have not gotten ahead of us our National Labs continue to produce state of the art cutting edge experiments that no one else has the resources to be able to. China is trying to catch up in certain areas but we still maintain leads in fusion, fission, space, quantum, etc.
Well the US Government does have the most scientific research in the world many times over. So for scientific discovery if they do not get access to that data than Claude will degrade over time.
With DOD activities yeah I would say that the DOD was the primary source of Claude DOD purposed capability.
These are both use case examples of course but they do lead to a cascading issue of OpenAI, Google, even xAI getting access to this data while Anthropic wont.
This is going to cost Anthropic $200 million on top of it a year with just DOD. If it goes into the other government departments and agencies then the cost is going to skyrocket for Anthropic