When I joined the US semiconductor industry, I learnt one hard truth: the best discoveries are rarely published. In short, depending on your nationality and that of your boss, your boss may not be cleared to know what you are doing! Then bring in export controls, you will realize that journals do not host the earliest best efforts in this world, unless they emanate from the halls of universities where the main work is to publish.
As a Nigerian citizen, I was cleared while some of my peers were not that lucky. During meetings, you need to check even in the same team who could know what. People, if you make mistakes, the FBI will send you an invitation.
But while most of those were for national security reasons, Google is taking it to another level: “Google’s head of A.I., Jeff Dean, used to encourage researchers to publish academic papers. But in February, the policy changed. That’s because a foundational part of the latest A.I. tech, including ChatGPT, originated in a Google study. Google now plans to share research only after it’s been turned into products”.
Simply, the papers on “transformers” were pivotal for modern generative AI systems, and Google engineers authored those. The problem was that others like ChatGPT’s OpenAI picked the ideas and commercialized them before Google. Going forward, Google does not want to reveal too much until it has commercialized.
When you see Aneke the bird dancing by the roadside, please check, someone is beating the drum. Google does not want to prepare its future for the slaughterhouse of technology.
So, going forward, only boring and worthless articles will be approved for journal publications. The real deal stays in-house! Capitalism 101!
