Comment on The New Yorker Asks: Is the A.I. Boom Turning Into an A.I. Bubble?
bacon_pdp@lemmy.world 2 days agoNo. A state government needs to support 1/10th of its population actively using its services. Say that state has 10M people; you will want 10k cores for all state services. an 8P server has about 1536 cores and you will need about 7 of them. So it still takes a whole rack even with the COBOL programs and applications written in C and Assembly.
Blue_Morpho@lemmy.world 2 days ago
“State services” is database lookups and billing. Back in the 90’s, I supported 10k users (1.5k active at any moment) on a Pentium 3 with 512MB of Ram.
bacon_pdp@lemmy.world 2 days ago
Constraint solvers for things such as Medicaid eligibility; OCR tagging for scanned documents; Anti-AI detection for uploaded images; but yes most state services are data entry and batch processing with web front ends.
Also the number of supported users does not scale linearly with the number of CPU cores as Amdahl’s law showed back in 1967.
Blue_Morpho@lemmy.world 2 days ago
US population has grown 25% from the year 2000. Other than Anti AI detection, everything worked on the hardware of 25 years ago. Single core performance has gone up more than 25% over the past 25 years.
bacon_pdp@lemmy.world 2 days ago
25 years ago an 8P server had only 8 cores (even if you bought Alpha 21364s) And states needed whole buildings to host their servers. Scaling that down to a single rack is the progress that occurred.