

Darlington Akogo
5.3K posts

@darlingtinho
In search of Raftel. Founder & CEO @minoHealth AI Labs, @karaAgro AI & @RunmilaInstitut. Building real world AI & AGI4Health, doing AI research & AI policy.



Building the next infrastructure for discovery! We’re proud to announce our partnership with The MRC/UVRI & LSHTM Uganda Research Unit, advancing the frontier of generative AI in biomedical research. Read More: …intillating-khapse-49be31.netlify.app/blog/exploring… @LSHTM @The_MRC



Scam Altman … x.com/i/grok/share/4…



@darlingtinho @Wofa_Jeffrey @minoHealth What was the strategic logic for positioning Ghana toward the frontier group rather than focusing on readiness for the capabilities that will diffuse as costs fall?

@darlingtinho @Wofa_Jeffrey @minoHealth Reading the AI Strategy, the misalignment between the ambitions and the announced finance is hard to miss. National AI Fund: 5B Cedis (~$420M USD) for 2025-30. The 10^21 FLOPS compute target alone needs billions of USD in hardware at current prices.


AI strategy is industrial policy for the digital age. Ghana has the talent. Execution and regulatory clarity are what separate builders from consumers.


That’s because you used H100s, which would actually be more about 3 OOMs more expensive than the calculation you did. But if you switch to B200/GB200, for the same FLOPs, it translates to ~2.8 billion. If you switch to Rubin, it translates to ~1 billion. Very realistic.


This is one of the few National AI Strategies anywhere in the world that specifically strategizes for AGI. This was very important to me that we focus on that for actionable strategies and timeline, because that's going to be the most consequential thing in our history.


Nodes needed is (9x10²⁰) ÷ (32×10¹⁵) = ~28,125 Price per node is ~$250k(best case scenario) Total price = 28,125 x 250k = ~$7 Billion



A 10²¹ FLOP/s compute build out in 12 months would genuinely require a miracle to pull off