2
1.5 hours runtime for like half a liter of gasoline?? That's unbelievably inefficient. A half-liter of gasoline is like 15MJ, should power a laptop drawing 30W for a week.
Maybe it would be better with a fuel cell.
199
"content curation"
(thelemmy.club)
cross-posted from: https://mander.xyz/post/46665693
PieFed blocks !enoughmuskspam@lemmy.world (and a few other communities) by default. At the time of writing this post, you can search for the comm on many PieFed instances and you will not find it.
- https://feddit.online/communities?search=enoughmuskspam
- https://piefed.ca/communities?search=enoughmuskspam
- https://piefed.europe.pub/communities?search=enoughmuskspam
The block is only by default. The admin can choose to override it. Many big instances have done so, including
- piefed.social
- piefed.world
- piefed.zip
lol hardcoded shit everywhere. that codebase is so bad it's entertaining. you should make a standalone post about this here and crosspost to !programming_horror@programming.dev
It is giving you exactly what you ask for.
To people complaining about this: I hope you will be happy in the future where all LLMs have mandatory censors ensuring compliance with the morality codes specified by your favorite tech oligarch.
41
Are you in the 95%?
(thelemmy.club)
Spoiler
More like 99.9995%
https://artofproblemsolving.com/community/c2532359h2760821_the_emoji_problem__part_i
solution:
🍎 = 36875131794129999827197811565225474825492979968971970996283137471637224634055579
🍌 = 154476802108746166441951315019919837485664325669565431700026634898253202035277999
🍍 = 4373612677928697257861252602371390152816537558161613618621437993378423467772036
The article over-dramatizes the story. This "deeply wrong" discrepancy is less than 10%. CMB measurements predict a Hubble constant of around 68km/s/Mpc. Distance ladder measurements get around 73km/s/Mpc.
Our current understanding of the universe the Lambda-CDM model is still wildly successful and it's more likely that the true correct model of the universe will be a correction/extension to Lambda-CDM rather than a completely new theory (although if it is a completely new theory that would be pretty cool).
The only times anyone would use the asterisk as multiplication symbol are
- they are doing some fancy math and it's not the same kind of number multiplication we're familiar with
- they are on a computer, the keyboard does not have a (×) key, and they don't know how to typeset it (
\timesin LaTex), so they just use the asterisk instead
The US government falls in the second category.
414
someone teach them LaTeX
(lemmy.ml)
A microblog post by @kareem_carr saying "as soon as i saw they were using asterisks for multiplication symbols, i knew we were in trouble", with an image from the "Office of the United States Trade Representative (Executive Office of the President)" showing the mathematical formula $\Delta \tau_i = \frac{x_i - m_i}{\varepsilon * \varphi * m_i}$. The formula show asterisks (*) instead of multiplication signs (×).
I recommend critically reading the paper. It is quite accessible to those with college-level science background.
Most importantly, it is still highly controversial whether this galaxy rotation direction bias actually exists. If you look at section 4 of the paper, the author is debating against different groups that did similar surveys and found no bias. Someone needs to actually work through this author's methodology as well as those of other groups and figure out what is going on.
If there is indeed a bias, that is super exciting! An anisotropic universe due to being in a black hole would be a very cool explanation. But given the ongoing debate, a general-audience publication like Independent presenting this rotation bias as a given fact is very poor journalism.
Stop depending on these proprietary LLMs. Go to !localllama@sh.itjust.works.
There are open-source LLMs you can run on your own computer if you have a powerful GPU. Models like OLMo and Falcon are made by true non-profits and universities, and they reach GPT-3.5 level of capability.
There are also open-weight models that you can run locally and fine-tune to your liking (although these don’t have open-source training data or code). The best of these (Alibaba’s Qwen, Meta’s llama, Mistral, Deepseek, etc.) match and sometimes exceed GPT 4o capabilities.
Yes, the earth accelerates toward the ball faster than it does toward the feather.
As always, relevant Wikipedia links:
https://en.m.wikipedia.org/wiki/Al-Khwarizmi https://en.m.wikipedia.org/wiki/Donald_Knuth https://en.m.wikipedia.org/wiki/Algorithm
The x axis is position. The y axis is energy. The blue box is a potential energy barrier. The red curve shows the wavefunction of a particle at a certain energy level coming in and tunneling through the wall. (the wavefunction actually live on a different y-scale from this plot and is only superimposed here for illustrative purpose, so don’t use the energy y-scale to read into the amplitude of the oscillatory part).
more info: https://en.m.wikipedia.org/wiki/Quantum_tunnelling
view more: next ›
BB84
0 post score0 comment score
joined 2 years ago
My oversimplified and possibly wrong understanding: this is like speculative decoding, but instead of a separate draft model (which does its own prompt processing), they use some diffusion thing strapped on top of the main model. The diffusion reuses the high-quality prompt processing result of the main model.
The 7.8x faster claim sounds almost too good to be true. But even if we get like 3x then this is still a huge revolution in localLLMing.