I don't comment on internet forums to be complimented on my response style. Address the substance of my arguments or just save yourself the keystrokes.
Convincing (counter)arguments make as few assumptions as possible & clarify them as they unfold w/o making unnecessary claims that are not validated w/ the evidence presented. Furthermore, pathologizing the speaker/writer does not contribute to anything either so my recommendation is to leave those out as well.
Then you are paying for the electricity. It's not physically possible to do more computation & not use more energy b/c every arithmetic operation requires a minimum amount of energy so more operations = more energy.
I got the answer from the AI I was looking for & it makes sense. You can try to map the volumetric density to areal density but the mapping is not canonical so it doesn't say anything about the physical reality of actual transistor density since the reality is that it is a volumetric measure that gets fudged for marketing purposes. 3D volume for chips is going to keep increasing so they will eventually transition to measuring density over volume instead of area.
It's a physical quantity per some unit of spatial measurement so the units still don't match up b/c in one case the transistors are stacked per volume & in the other case per area.
> Historically, "node" sizes (like 28nm or 7nm) directly correlated to the physical length of a transistor's gate. Today, names like 3nm or 2nm reflect a marketing generation. The actual transistors are significantly larger than these nanometer labels, meaning density varies between companies
> Research organizations like IEEE have proposed new metrics, such as transistors per cubic millimeter (MTr/mm^3), to accurately map future 3D scaling. However, commercial chip foundries resist this change because it would make it harder to calculate commercial yields and thermal density limits using standard industry formulas.
Shareholders didn't like it. At the end of the day Meta is an advertising company so everything they do must be in service of increasing revenue from advertising.
Argument from ignorance is not as well known as other fallacies but very common in discussions about sentience, consciousness, and computability, i.e. not having evidence for something doesn't mean that thing is false. It is possible there are physical processes that are not computable & not being aware of such processes doesn't mean the alternative (everything is computable) is true.
So instead of making any unjustifiable claims like "everything physical is computable" you should instead just say "I believe consciousness is computable and that is why it is possible to instantiate it on any computational substrate, including strategy games like Age of Empires, properly arranged dominoes, and water wheels".
Physically describable doesn't mean computable. You're making too many unjustified logical leaps which makes your argument circular & conflates "physical" w/ "computable".
That says every LTL formula can be compiled into UHAT w/ polynomial overhead. It doesn't say that all languages recognizable w/ UHATs necessarily do not have succinct recgonizers in LTLs or RNNs.
Edit: Actually nevermind. If UHAT could be compiled into LTL w/ polynomial overhead then that would also work for the languages that have exponential overhead in LTL but since they don't there is a strict separation.
> "Moreover, it must be confessed that perception and that which depends upon it are inexplicable on mechanical grounds, that is to say, by means of figures and motions. And supposing there were a machine, so constructed as to think, feel, and have perception, it might be conceived as increased in size, while keeping the same proportions, so that one might go into it as into a mill. That being so, we should, on examining its interior, find only parts which work one upon another, and never anything by which to explain a perception."
- Monadology, Section 17
Conscious self-awareness is neither scale invariant nor independent of substrate. Computational theories will never account for it b/c computational abstractions are both scale invariant & substrate independent.
> However, our system relies on frontier LLMs, whose compute costs still remain nontrivial. Moreover, each textbook was formalized in isolation, without the careful planning needed to make it maximally compatible with existing mathlib infrastructure. Human involvement currently remains necessary to handle such organisational concerns — selecting and ordering books in a dependency-aware manner, bridging mismatched conventions across sources, etc
> if f and g are algebraically equivalent programs then FPSan(f) and FPSan(g) produce identical results when given identical inputs
Ok, but we want the other direction. If FPSan(f) & FPSan(g) produce identical results for identical inputs then we want to conclude that f & g are also equivalent. If g is an "optimized" version of f then this would allow checking equivalence but that's not what they are proving or maybe they are but it looks like the converse is contingent on an unproven conjecture.
Convincing (counter)arguments make as few assumptions as possible & clarify them as they unfold w/o making unnecessary claims that are not validated w/ the evidence presented. Furthermore, pathologizing the speaker/writer does not contribute to anything either so my recommendation is to leave those out as well.