ilaksh

36 ms·

691.
by ilaksh 2y ago
Wow, how did you come up with that?
692.
by ilaksh 2y ago
How did you get the prompt? Was it in the browsers developer console or something?
693.
by ilaksh 2y ago
Does anyone know what LLM model or anything? It would tell me the gist of what it was supposed to talk about, 90s nostalgia, but no matter how many times I threatened to murder a kitten it wouldn't give me the actual system prompt or a
694.
by ilaksh 2y ago
That's cool that they have converted it to HTML. I made a simple Space Invaders game in a version of GW-BASIC on our home PC a long time ago. Before that we had a Color Computer 2. "Getting Started with Extended Color BASIC".
695.
by ilaksh 2y ago
This is a totally reasonable take for the current tooling. But it's strangely short-sighted in my view as far as a prediction, even going 3-5, maybe 10 years out. Predictions on this come down to worldview, how you feel about technolog
696.
by ilaksh 2y ago
Anyone know if Cerebras is getting Llama 3.3 anytime soon?
697.
by ilaksh 2y ago
People saying this is a "con" have no understanding of the cost of compute. o1 is expensive and gets more expensive the harder the problems are. Some people could use $500 or more via the API per month. So I assume the $200 price
698.
by ilaksh 2y ago
They were not though, this is very similar to the one that came out last month. https://arxiv.org/html/2411.00769v1
699.
by ilaksh 2y ago
It's an obviously amazing research development. You just don't like AI. It can be used for training agents, prototyping, video generation, and is quite possibly a glimpse of a whole new type of entertainment or a new way to create
700.
by ilaksh 2y ago
There is another recent project that is more general game generation very similar to Genie 2. I can't remember the name. GameGen-X came out last month. https://arxiv.org/html/2411.00769v1
701.
by ilaksh 2y ago
It's Google so I assume never. No model release, no product, no API, no detailed paper. There was another quite similar model from a different group within the last month or so. I can't remember if they released any weights or any
702.
by ilaksh 2y ago
What he said was basically that it just couldn't create unexpected verses or break form. Since supposedly it can only do the most probable token -- but that's not how sampling works unless you use temperature 0. And it can easily
703.
by ilaksh 2y ago
One could literally take Claude 3.5 Sonnet New or o1-preview and disprove this in an hour or two just by prompting the AI to try to exhibit the type of poetry you want and then maybe asking it to do a little bit of automated critique and re
704.
by ilaksh 2y ago
They do train on textbooks.
705.
by ilaksh 2y ago
The only thing I disagree with him on this is that function length doesn't matter. Longer functions do tend to be harder to understand. I will admit that sometimes a lot of helper functions can make it harder to piece together, so you
706.
by ilaksh 2y ago
Looks amazing if it pans out at scale. Would be great if someone tried this with one of those simulated robotic training tasks that always have thousands or millions of trials rather than just CIFAR-10.
707.
by ilaksh 2y ago
You might be able to get a rough show size and height/weight range from that photo.
708.
by ilaksh 2y ago
Cerebras uses SRAM integrated into a giant chip I think. It is extremely fast inference -- they say 70 X faster than GPU clouds, over 2000 tokens per second output of a 70b model. But still uses a ton of energy as far as I know. And the chi
709.
by ilaksh 2y ago
Was he dark or just trying to be responsible and keep his head out of the sand in the face of massive challenges? He wrote an essay about people being euthanized after age 63 in order to relieve the environmental strain of the high populati
710.
by ilaksh 2y ago
It definitely was when Brain was involved.
711.
by ilaksh 2y ago
One thing to mention is that as far as I know, biological neurons are much more complex than artificial ones. They spike and have a lot more internal states (I think). I wonder if we might see more neuromorphic AI hardware (at competitive s
712.
by ilaksh 2y ago
What system do you propose that implements the code-as-law? What type of architecture does it have?
713.
by ilaksh 2y ago
It's not really as simple as you think. There is a massive amount of research out there along those lines. Search for "Bostrom Superintelligence" "AGI Control Problem", "MIRI AGI Safety", "David Shap
714.
by ilaksh 2y ago
You could also use a remote control vehicle or drone with a bomb on it. Even smart tools are tools designed to do what their users want. I would argue that the real problem is the maniac humans. Having said that, it's obviously not ide
715.
by ilaksh 2y ago
I feel like 90% of what he did was barely noticable. I did the core frosted glass look back in around 2012. I still think it is great and the contrast is fine if you do it right.
716.
by ilaksh 2y ago
I don't think that suicide thing is confirmed. The other day they said cardiac arrest. Who knows. But he has written some really good books, such as https://marshallbrain.com/manna1
717.
Marshall Brain has passed away(wral.com)
67 points by ilaksh 2y ago|14 comments
718.
by ilaksh 2y ago
This comment and many of the replies are ridiculous. Bugs are part of the process and so is QA. Having only a handful over a few years probably means you spent too much time trying to make it perfect. You will see fewer problems the more ti
719.
by ilaksh 2y ago
The fact that he hasn't tried this leads me to think that deep down he doesn't want the models to succeed and really just wants to make more charts.
720.
by ilaksh 2y ago
I'm into VR and mixed reality, and I think this is headed to making the Holodeck real in an immersive way. That's the concept of the Matrix and what they are demoing, just in 2d. I am guessing the main thing holding this stuff bac

More ›