Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
cypress66
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
11 ms
·
91.
▲
by
cypress66
3y ago
Yeah, a lot of them are very cool.
92.
▲
by
cypress66
3y ago
2 used 3090s
93.
▲
by
cypress66
3y ago
Could you be more specific about how much is GPT4?
94.
▲
by
cypress66
3y ago
Just make it opt in or have the user manually click yes/no during the install.
95.
▲
by
cypress66
3y ago
5 tokens/s on 70B 4bit seems really high for your setup.
96.
▲
by
cypress66
3y ago
What language? 70B English - Spanish is pretty strong. But that's almost surely the best case scenario. And 70B will no doubt be much better.
97.
▲
by
cypress66
3y ago
You don't need to be doing research to read an ML paper. With some general knowledge in AI you should be able to understand most papers. And even then, sometimes you don't understand or care about their procedures, and you just wa
98.
▲
by
cypress66
3y ago
Current LLMs are incabale of something like that so it doesn't matter.
99.
▲
by
cypress66
3y ago
Node is pretty performant for anything IO related, not compiled and reasonably dynamic.
100.
▲
by
cypress66
3y ago
Sorry I don't follow how this "standard" configuration is relevant. If you want more just buy more? If you don't want to pay pre-built markups build it yourself. And the point about gaming setups is very relevant because
101.
▲
by
cypress66
3y ago
Talk about over engineering
102.
▲
by
cypress66
3y ago
I'm not sure what "standard computer configuration" means. Maybe you mean a budget laptop? Your typical new gaming desktop build is 32GB, and for a workstation probably 64GB. I think you can get a 2TB ssd for like a 100 bucks
103.
▲
by
cypress66
3y ago
Finetuning is more relevant than ever now. People are fine tuning LLaMA every single day.
104.
▲
by
cypress66
3y ago
Yep. Use ooba. And people who like to RP often use ooba as a backend, and sillitavern as a frontend.
105.
▲
by
cypress66
3y ago
Not the base model
106.
▲
by
cypress66
3y ago
>AI systems shall be developed and used in a way that includes diverse actors and promotes equal access, gender equality and cultural diversity, while avoiding discriminatory impacts and unfair biases that are prohibited by Union or nati
107.
▲
by
cypress66
3y ago
With this law, you can forget about base models such as LLaMA being released, since models will be required not to say "mean things".
108.
▲
by
cypress66
3y ago
Well, yeah. Deep learning shows that approach works (as long as you have the data as well).
109.
▲
by
cypress66
3y ago
That's the 13B model. If you want something comparable to GPT3.5 you must use the 70B.
110.
▲
by
cypress66
3y ago
Apple's P/E is around 30. Which is a bit high but completely within reason. Very low compared to inflated stocks such as NVDA at 230 P/E Apple is just crazy profitable.
111.
▲
by
cypress66
3y ago
Most of that life expectancy gap is due to obesity I'd guess.
112.
▲
by
cypress66
3y ago
Yeah. That's why I much prefer an L4 car that you still own and sit in the driver seat with steering wheel, than a taxi I rent and have no control over. (obviously there's no such L4 car yet, but I'm sure tesla/mobileye&
113.
▲
by
cypress66
3y ago
People like different things. Some people like high density urban areas, and some people like low density urban areas. We should have both so people can enjoy what they like.
114.
▲
by
cypress66
3y ago
I wished they allowed non user replaceable batteries if they use chemistries that support a high number of cycles (such as LFP batteries). I'd rather have a battery that lasts the full lifespan of the device, than a easy to replace bat
115.
▲
by
cypress66
3y ago
> Self-driving cars kill people all the time. Maybe L2 systems. But those are just a driver assistance, the driver must supervise and intervene. Actual L4/L5 systems haven't killed anyone as far as I know, except for the Uber a
116.
▲
by
cypress66
3y ago
A parameter is comparable to a synapse, and the brain has around 100 trillion.
117.
▲
by
cypress66
3y ago
What's the problem with generating porn, assuming that's what the user asked for?
118.
▲
by
cypress66
3y ago
There are use cases where you'd prefer the "safe", dumbed down version that's hard to jailbreak, like chatbots for customer support. However they should definitely offer an uncensored version as well.
119.
▲
by
cypress66
3y ago
What context size is used? Can you "brute force" your way by using let's say 16k or 32k of context?
120.
▲
by
cypress66
3y ago
Carpenter and tailor are basically unique words. If you never heard of the word carpenter, you'd have no idea what it meant. It is very rare I think for completely new words to emerge nowadays. Prompt engineer is descriptive and you ca
More ›