Veedrac

79 ms·

871.
by Veedrac 7y ago
To be clear, I'm not denying that genius correlates with certain mental afflictions. I'm denying that this should be a major reason to be concerned about IQ enhancement through genetic modifications, and would add to my original p
872.
by Veedrac 7y ago
It's unreasonable to expect initial modifications to start out as swathe modifications summing to tens of IQ points. The first gene modifications are going to be fixated on the safest changes that can be made to health and, along the r
873.
by Veedrac 7y ago
This argument never made much sense to me because overall IQ is correlated positively with health and longevity. Sure, it might change the nature of the afflictions we get, but mostly this just feels like an argument from just comeuppance,
874.
by Veedrac 7y ago
Most of the world lives in countries about twice the size of Iran or greater.
875.
by Veedrac 7y ago
> It's two bits, DAZ and FTZ. (seems like "denormals are zero" and "flush to zero") You mean the standard ones for normal floats? You certainly wouldn't want to reuse that for bfloat16s. > ML is all about
876.
by Veedrac 7y ago
I appreciate you're trying to wind it down so I'll try to get to the point, but there's a lot to unpack here. I'm not evaluating these models on whether they are AGI, I am evaluating them on what they tell us about AGI i
877.
by Veedrac 7y ago
Thanks.
878.
by Veedrac 7y ago
> What exactly is incredible (relatively) about the current state of things? Compared to 1999? Watch https://www.youtube.com/watch?v=kSLJriaOumA Hear https://audio-samples.github.io/#section-4 Read http
879.
by Veedrac 7y ago
> I mean the bit to toggle denormals, not the one to identify support for the opcodes. bfloat16 doesn't handle denormals, why is there a bit to toggle it? What's it called (so I can CTRL-f for it)? > If you still think denor
880.
by Veedrac 7y ago
> Intel did add a global state bit. It just isn't useful because you can't modify it. You mean the CPUID bit? That's free. Toggling denormals isn't. > Nervana is discontinued, isn't it? Compatibility doesn&#x2
881.
by Veedrac 7y ago
I'm not interested in categorizing for the sake of categorizing, I'm interested in how AI researchers and those otherwise involved can get a measure of where they're at and where they can expect to be. If AI researchers were
882.
by Veedrac 7y ago
My comment was not on ‘continual’ but on ‘Turing test’. If you gave 100 chimps of the highest calibre 100 attempts each, not a single one would pass a single Turing test. Ask a feral child to translate even the most basic children's bo
883.
by Veedrac 7y ago
> I think continual Turing testing is the only way of concluding whether an agent exhibits intelligence or not. So you think it's impossible to ever determine that a chimpanzee, or even a feral child, exhibits intelligence? This see
884.
by Veedrac 7y ago
https://openai.com/blog/openai-lp/
885.
by Veedrac 7y ago
I'm compiling a list of reasons people doubt AGI risk, could you clarify why you think AGI is certainly far term?
886.
by Veedrac 7y ago
This makes no sense. Mice can't write poetry. Expecting a 1:1 equivalence between human and manufactured intelligence is no more coherent than denying the possibility of human-bearing flight until we have planes as acrobatic as a hawk.
887.
by Veedrac 7y ago
> I don't know, but I'd say after a definition of "AGI" has been accepted that can be falsified against, and actually turn it into a scientific endeavour. Uhh, that's the Turing Test.
888.
by Veedrac 7y ago
But then you need either new denormal-accepting instructions or, worse, a new global state bit enabling bfloat16 denormals, all to support use cases probably over two orders of magnitude less common than ML. What's the compelling reaso
889.
by Veedrac 7y ago
I'm not an expert in the nitty gritty, and I've heard conflicting information, but my rough understanding is that it would be affordable and not too difficult for Xeon processors, but relatively more expensive for the dedicated Ne
890.
by Veedrac 7y ago
AVX is SIMD, and each AVX register is a large number of bits (256 or 512) that is packed full of a given data type, which can all be operated on at the same time with the same instruction. So if you have a 512 bit register, that could hold
891.
by Veedrac 7y ago
bfloat16 is only really intended for ML purposes, and denormals don't really matter there, especially given they accurately multiply-accumulate into 32 bit floats.
892.
by Veedrac 7y ago
‘(8, 1, alpha, beta, gamma) log’ isn't a posit; it just shares the encoding of the exponent.
893.
by Veedrac 7y ago
Yes, but there's a time-accuracy trade off when you see things by scanning around. A good, wide camera trades off remembered perepheral details for lower-resolution but temporally up-to-date details, which matters more in a crash I&#x2
894.
by Veedrac 7y ago
The "neural network of neural networks" thing is a bit of lavish exaggeration TBH, because a network of networks is just a larger network. The human brain has about 100tn synapses, which I think is a less obscure statistic to marv
895.
by Veedrac 7y ago
Though post-processing from neural nets seems to close this gap. https://www.youtube.com/watch?v=bcZFQ3f26pA
896.
by Veedrac 7y ago
He misconveyed the source. The claim was about the detail the eye can percieve in a given field of view, over time and by moving the eyes over it, not the amount of information the eye captures at any one time, which is vastly lower. It&#x2
897.
by Veedrac 7y ago
Shanon pretty clearly teaches us that bandwidth is a fundamental thing you can't circumvent by this sort of argument. It might do other stuff on top (though a lot of communication in the spine is latency-critical and at least some of i
898.
by Veedrac 7y ago
As long as it's a small, non-locking pocket knife, you're fine. If it's larger, you're fine as long as you have a good reason (eg. needed for work).
899.
by Veedrac 7y ago
> Moreover, the ceiling you could get in boosting single-threaded performance with a perfect branch predictor (for conditional branches) over current state-of-the-art is around 4%. No, actually having perfect branch prediction is a major
900.
by Veedrac 7y ago
In an ideal world, this would be how hardware spoke, and C wouldn't have even had a choice. Alas, the past cannot learn the lessons of the present.

More ›