[Tokyo Tech Translated] The Algorithmic Divide
private ai infrastructure and the social media funnel, two sides of the same coin
today's selection orbits a single tension. on one side, a sober industry push to reclaim compute from the cloud. on the other, a weary recognition that the same algorithmic logic funnels everyone else toward the casino. the common thread is control, or the lack of it.
@gihyosd, private ai infrastructure
our july 2026 issue's "ai data center feature" has been getting good feedback. thank you. we planned it partly as a kind of grown-up field trip curiosity, but it actually contains a lot of practical information.
with cloud ai service costs spiking lately, many people using ai for work have probably wondered at least once, "can we keep paying for this?" or "could we run ai on our own infrastructure?" this feature is also an attempt to explore the possibility of a dedicated ai environment.
the three companies that wrote for this feature, fixstars, ntt pc communications, and getworks, all build and provide custom ai infrastructure for enterprises and organizations. the reasons for wanting a private ai environment vary. we initially focused on cost reduction, but it seems that "we handle confidential information so cloud ai services are difficult to use" is another common reason.
the content is also useful for anyone thinking about setting up and running a local llm. please take a look. (and we're planning a future feature on local llms as well, so in that sense this issue is worth reading as background.)
source: https://x.com/gihyosd/status/2071068637499375843
@EaKbkl8TpSbVxrJ, the funnel
so many people on social media, right? their daily engagement numbers far exceed their actual trades. and yet everyone just uses it like it's normal. content filtered for you, consumption behavior steered, funneled toward porn, gambling, dating apps, videos, games... all by the same algorithm.
but somehow it's a playground for the rich.
source: https://x.com/EaKbkl8TpSbVxrJ/status/2071031389475184990
the first piece describes a world where enterprises are quietly building their own compute, driven by cost and confidentiality. the second points out that everyone else is being algorithmically herded into a different kind of infrastructure entirely. one path is about owning the stack. the other is about being owned by it. the gap between them is not just technical. it is a question of who gets to be the user and who gets to be the used.
more at falsifylab.com