szvsw

21 ms·

211.
by szvsw 2y ago
Edit: I should say that I think a much more applicable use-case for something like this is passing in N papers and asking for synthesis, or to eliminate N-1, select the most relevant to a query, etc etc. what I am reacting to below is just
212.
by szvsw 2y ago
> appealing to nerds to believe that a long-tail disaggregated phenomenon like EV V2G is going to save the grid, but the apparent truth is that the grid will be saved by central planning and top-down command and control. I think this sen
213.
by szvsw 2y ago
Heating and cooling would be massive. Another one is dishwashers (eh) but also laundry machines (washers and dryers/hot water heaters). Pushing those to random overnight spacing or preferably midday can help at least a bit.
214.
by szvsw 2y ago
> some designs for batteries that basically involve physical pumps to pump liquid anodes and cathodes around Yeah in addition to flow batteries there are also grid scale thermal batteries, though these are targeted more towards long dura
215.
by szvsw 2y ago
OpenAI also just published similar work, though Anthropic did beat them to the punch. https://openai.com/index/extracting-concepts-from-gpt-4/ https://news.ycombinator.com/item?id=40599749
216.
by szvsw 2y ago
There’s also ARIA [1]. I actually think the one linked above seems more interesting. At the same time, doesn’t seem all that different from just rolling your own solution. [1] https://github.com/lifan0127/ai-research-as
217.
by szvsw 2y ago
I believe that generally speaking, shipping costs typically are considered a relatively small portion of overall costs for any given product - just going off discussions in Box by Levinson, in which he asserts that one of the paradigm shi
218.
by szvsw 2y ago
It’s in the 4th par. > Our 23 solar panels are rated to produce a total of 8.28 kilowatts (kW), but in reality energy production doesn’t get that high and maximum production is for a short part of the day. We peak at about 6.6 kW kWh i
219.
by szvsw 2y ago
Home PV + batteries could at least help tame the duck curve and at least moderate the cost of grid expansion which ultimately is a cost born disproportionately by low income homes which are already energy burdened. But yes I mostly agree wi
220.
by szvsw 2y ago
It can definitely make a big difference. DC’s latitude is 38 degrees. That’s about the same gap to Miami (25 degrees). Think about shining a flashlight perpendicular to the ground vs at an angle. Beyond the dynamics of direct irradiance, th
221.
by szvsw 2y ago
I agree V2G will be huge, but rolling it out will be hard, especially somewhere like the US without major financial incentives (for both the infrastructure and getting people to accept the idea that they might not fully control their rollin
222.
by szvsw 2y ago
Where do you live? What latitude? That can have a big impact. Also note that the publication was 2021, not 2011 (typo by parent)
223.
by szvsw 2y ago
They do indeed. For instance, Singapore uses the most energy per capita in the world (or top 3-4 depending on year)… and it’s mostly due to AC. Even with very high coefficients of performance (usually in the 2-4 range, meaning you can remov
224.
by szvsw 2y ago
What is the monetization scheme that would result in it being built? The beautiful part of governments spending money on this is that there is not nearly as much pressure for an RoI to “justify” it…
225.
by szvsw 2y ago
EDIT: more succinctly: are there clear reasons beyond the historical and social that fundamental research in one field receives large investment versus another? —- I think stuff like this is awesome (though I have absolutely zero experience
226.
by szvsw 2y ago
Probably would have been better if I had instead suggested figure formatting and page layout… I think I’ve come close to stroking out over that…
227.
by szvsw 2y ago
I really disagree with pigeonholing it as an LLM architecture! It is much more general than that as I mentioned in another comment in this post [1] (and of course as mentioned in the original paper which you linked). [1] https://
228.
by szvsw 2y ago
The Mamba application is my current research project so I haven’t published anything yet. But the basic idea is to create a latent representation of the static features, repeat the latent vector to form a time series, concatenate with the w
229.
by szvsw 2y ago
It’s much more than just an LLM. The mamba architecture is often used in the backbone of an LLM but you can use it more generally as a linear-time (as opposed to quadratic-time) sequence modeling architecture (as per the original paper’s
230.
by szvsw 2y ago
Choke on chalk, nerd! Or maybe have an aneurysm trying to remember some latex command.
231.
by szvsw 2y ago
> Sidebar: I'm a founder of a self-bootstrapped startup to build a better weather data broker that doesn't cost an arm and a leg. If that's something you are passionate about solving, feel free to reach out :) Would love t
232.
by szvsw 2y ago
Yeah there are lots of ways to normalize that might be better, but per capita isn’t uninteresting per se: US uses roughly 2.5-3x the energy per capita compared to china fwiw [1]. The energy use per capita can be influenced by all sorts of
233.
by szvsw 2y ago
> I've been self studying architectural engineering. Get the form and forces textbook. It’s fantastic. > I hope to some day build my own home. It may well be a tribute to brutalism, we'll see. Don’t tell any architecture stu
234.
by szvsw 2y ago
I mean, if you like state space models, then you should read the paper on Mamba if you haven’t already! Because it quite literally uses state spaces… and you will probably think it’s a really cool application of state spaces! Apologies if y
235.
by szvsw 2y ago
Very well said. It absolutely drives me crazy - the focus on the model architecture for autoregressive text generation as the “key” (or blocker) to AGI seems so misguided as a debate. It seems clear to me (and maybe I’m wrong!) that a much
236.
by szvsw 2y ago
> buildings are actually moments of force, the cement or stone or wood or steel only exists to keep everything from collapsing in on itself. This is actually kind of a beautiful sentence, especially the first half. I’m sharing it with so
237.
by szvsw 2y ago
Yep yep I know, but I was trying to suggest something diffusion-like occurring with a language model through a totally separate mechanism that does not rely on the denoising process (at least not literally).
238.
by szvsw 2y ago
Wow, really cool concept! I wonder if this starts to become similar dynamics to what we see in image generation models, where structure/detail emerges in one region of the image and then the surrounding areas start to resolve themselve
239.
by szvsw 2y ago
SHAP is pretty separate IMO. Shapley analysis is really a game theoretical methodology that is model agnostic and is only about determining how individual sections of the input contribute to a given prediction, not about how the model actua
240.
by szvsw 2y ago
> likely all these guys went to the same metaphorical SF bars, it was in the water It also is coming from a long lineage of thought no? For instance, one of the things often thought early in an ML course is the notion that “early layers

More ›