All right, but apart from the sanitation, the medicine, education, wine, public order, irrigation, roads, a fresh water system, and public health, what have the Romans ever done for us??
11 karma · joined October 28, 2022
Is a powerful way to start a blog post. 9/10
, is the worst economic take i've heard in a while and I read the FT. CoL is primarily driven by supply and demand, not willingness to pay, not to mention inflation, supply chain issues, inelastcity...
Fake diamonds are for vegans.
[Born during Clinton's first term]
Whatever $€£¥₹₽^x to BTC bridge exists, I'm all for it.
The only real revolution in/for/from 'crypto' is BTCs double spend PoW solution/it's ability to be defi p2p cash for 15+ years and running.
Huh, intresting.
Believe 15A and 19A as much as you can believe De Beers' 'Building Forever 2030 Goal's
People are being priced out of art and beauty and it's a shame economics and corruption make real diamonds dirty.
Humans search for certainty and authority. LLLMs are so 'confident' in their answers. The truth/reason issue is not unque to AI - look at those who wide up in cults.
Coding aside, their not made for theraputics, they're predictive, and as a scientist who damn well wishes it could DO science FOR ME (would make my job easier), it cant, i tried, it cannot create anything new, it cannot reason from first principles, its a glorifed spell checker that summarizes articles reall well for me.
The concept of "Computational Thermoepistemics" ambitiously seeks to unify computation, thermodynamics, and information theory into a single theoretical framework claiming to reveal fundamental energy limits of AI and intelligence. While its interdisciplinary ambition is commendable, the framework overreaches by treating computational processes too literally as thermodynamic microstates with physical quantities like pressure and chemical potential, which lack rigorous empirical grounding in current computing technologies. The core claims about logical depth and undecidability bounding thermodynamic efficiency conflate abstract computability theory with physical energy costs in a way that oversimplifies highly distinct domains. Moreover, practical quantum advantages are currently limited by error correction and decoherence costs rather than fundamental thermodynamic laws, so the paper’s distinctions are mostly theoretical, not yet substantiated by experimental evidence. Biological efficiency comparisons often ignore the vastly different operational scales and purposes between brains and silicon AI. Finally, while energy-aware algorithm-hardware co-design and reversible computing are promising research paths, the lofty promises that energy efficiency gains could approach biological levels or exploit deep thermodynamic parallels remain speculative. Overall, the framework presents an intriguing synthesis but lacks conclusive validation, so it should be seen as a stimulating hypothesis rather than an established theory reshaping AI energy paradigms.
*edit, for the record, for me, as a postdoc, a hypothesis, while more structured than mere speculation, is still an unproven explanation that requires rigorous testing and peer review before it can be considered established scientific knowledge, and thus remains speculative.