cpill

10 ms·

121.
by cpill 2y ago
k3s for on premises deployments. I'm also using it for local self hosting side projects.
122.
by cpill 2y ago
so how much GPU RAM does need to get the 70B going fast (ish)?
123.
by cpill 2y ago
the biggest thing he's overcoming is the rent?! how's he doing that while goofing off with projects like this?
124.
by cpill 2y ago
I feel like IF tech could be use for more than just fiction. Imagine government forms that could use it to help people navigate a complex bureaucratic system of forms, or something like the python cookiecutter for generating project layouts
125.
by cpill 2y ago
You'd need an example set, like all FT'ing
126.
by cpill 2y ago
I think the branching factor for LLMs is around 50k for the number of next possible tokens.
127.
by cpill 2y ago
the recent episode of Machine Learning Street Talk on control theory for LLMs sounds like it's thinking in this direction. Say you have 100k agents searching through research papers, and then trying every combination of them, 100k^2, t
128.
by cpill 2y ago
I wonder if it would help with the brain fog of long Covid...?
129.
by cpill 2y ago
I do it too but the trouble is that it's a fire hose of information. Some intelligent filtering would be nice, mixed with serendipity (some how)...
130.
by cpill 2y ago
I'll be looking for this in Hollywood movies
131.
by cpill 2y ago
you don't happen to play factorio do you?
132.
by cpill 2y ago
I guess this will give birth to a new kind of film making. Start with a rough sketch, generate 100 higher quality versions with an image generator, select one to tweak, use that as input to a video generator which generates 10 versions, cof
133.
by cpill 2y ago
not machines learning Devs
134.
by cpill 2y ago
k3s is very simple to setup and and nodes to. if the machines are on the same LAN or the internet then it's not such a complex job albeit you need to know the basics of kubernetes.
135.
by cpill 2y ago
yeah, you get to choose the cutlery when all you really want is to eat.
136.
by cpill 2y ago
so what learning were learnt for this lesson?
137.
by cpill 2y ago
If the presence of Google ads on a site is a negative signal then a search engine that down ranks sites with ads might be a good alternative? is this what Kagi does?
138.
by cpill 2y ago
I imagine it's not much fun with YC on you constantly telling you to make more, scale, or pivot to you make our investment worth it. I find the clients that pay the least demand the most
139.
by cpill 3y ago
man, I'm glad someone find this interesting so that I don't have to.
140.
by cpill 3y ago
it would be nice if there was some way, like a map of what concepts you needed to know, that you could follow, so if you did want to get into this you could ram up from nothing (or at least high school maths)
141.
by cpill 3y ago
haha, this track is hysterical https://www.udio.com/songs/jGjYfsRosZjYTkSBdFgEyF
142.
by cpill 3y ago
yeah, I had an idea for a new search engine and was hoping to get an idea of how to do it :(
143.
by cpill 3y ago
could somebody tell me why all the ROMs need updating _every time_ a new version of MAME is released? it would seem like a major cockup: updating the data, the ROM dumps which I imagine are always the same as they come from hardware(?), ins
144.
by cpill 3y ago
yeah, the lawyers will have the whole case on those two words: "where applicable"
145.
by cpill 3y ago
yeah, imagine what this will do for lawyers
146.
by cpill 3y ago
yeah that to the Ai doomers who are making six figures off of it
147.
by cpill 3y ago
which modern intellectual supports Marx these days?
148.
by cpill 3y ago
oh the arrogance! love it this is why I read the comments on hacker news before (or mostly in place of) the article :P
149.
by cpill 3y ago
"better things than a chat interface" like what? what interface comes more naturally to humans than the spoken word?
150.
by cpill 3y ago
wouldn't it be better to use something like tiktoken's tokens to represent words and then use the RGB values as indexes into that encoding? you'd get higher compression still

More ›