azeirah

19 ms·

211.
by azeirah 3y ago
> At first, the CEO carefully monitors the work, but as months go by without error, the AI receives less oversight and more autonomy in the name of efficiency. It occurs to the CEO that since the AI is so good at these tasks, it should t
212.
by azeirah 3y ago
The one I use is CodeGPT. It doesn't. It's a simple chat-like wrapper over the GPT4 api.
213.
by azeirah 3y ago
I don't have any super concrete examples, but I can confirm it _helps_ in writing software for me. Things it has helped with, that are now part of a codebase I work in and get paid for. Generated example css to center a text element on
214.
by azeirah 3y ago
Some possibilities are 1) Complexity of the human brain is higher than we thought, reproducing the complexity necessary for human-level intelligence is further into the future than we initially expected. 2) Intelligence is bound by size&#x2
215.
Steering GPT-2-XL by adding an activation vector(lesswrong.com)
5 points by azeirah 3y ago|0 comments
216.
by azeirah 3y ago
It's because vicuna is fine-tuned on chatGPT answers. LLaMa will not do this, but LLaMa-based models fine tuned with chatGPT answers will.
217.
by azeirah 3y ago
Privacy as well! I'd much rather run a model locally than in the cloud even if I lose out in performance in both speed and quality.
218.
by azeirah 3y ago
I'm following the discussions on GitHub as well as their PRs closely. The primary bottleneck for now is compute. They've recently made a big improvement to performance by introducing partial gpu acceleration if you compile with a
219.
by azeirah 3y ago
It has partial gpu acceleration if you compile it with LLAMA_CUBLAS or LLAMA_CLBLAST They really have come a long way since... A few weeks ago. Using cublas with my 1080ti results in a 52% speedup compared to cpu-only. Vram usage is very mi
220.
by azeirah 3y ago
Llama.cpp recently added partial GPU acceleration. Model dequantization as well as some BLAS operations have been moved to GPU. It runs a lot faster if you compile with cuBLAS (nvidia) or clblast (other). GPU vram doesn't matter much s
221.
by azeirah 3y ago
I was wondering about the integrated gpu in my desktop ryzen 7900x. I can find very little about it (or other 7000 series integrated gpus). Is this usable at all for running LLaMa in some way?
222.
by azeirah 4y ago
Influenza is a virus that is constantly mutating. I think the comparison means to the variants colloquially known as "the flu", which are the ones that make you sick for a week or maybe two. They're not comparing to all influ
223.
by azeirah 4y ago
huh? Of all things what does 4chan have to do with anything here?
224.
by azeirah 4y ago
Not entirely certain about this, but I believe that because Apple M-series GPUs use system memory you could possible run the larger models on accessible (albeit expensive) consumer hardware. Need 64GB for 30B And 128GB for the 65B I'm
225.
by azeirah 4y ago
What OS are you running, then? Linux, MacOS and Windows are all supported. I don't know a lot of commercial software that is cross-platform beyond those three.
226.
by azeirah 4y ago
He talks about this topic, it's in the transcript. I ctrl+f'd for "large" of large language model
227.
by azeirah 4y ago
It's possible to do it without their connect API for sure, but that would require users to 1. Look up their ssh password 2. Connect tablet to their computer 3. Run a (my) program on their computer to install custom sync software on the
228.
by azeirah 4y ago
It is not, it requires the connect service subscription as it uses their cloud API to access data from your Remarkable.
229.
by azeirah 4y ago
For what it's worth, I am working on a tool that syncs files from a ReMarkable drawing tablet to Obsidian. I was planning on looking into support Samsung notes and Apple notes next. An open-source project like in the OP will help a lot
230.
by azeirah 4y ago
In my opinion LLM's are easier to learn, hard to master. Anyone can use chatgpt to make something happen for them. Want something specific and amazing? You need to take some time to learn about how it works and how you can make it do w
231.
by azeirah 4y ago
This is pure speculation on my part, but I don't think LLMs in their current form will be able to talk about what they do and don't know. No matter how large the input. There is no introspection in their architecture. Introspectio
232.
by azeirah 4y ago
Is there a reason why we aren't using BLOOM's model? It's huge, 160B parameters. It's open, it's free, it's been out there for a while. Is something wrong with it? Using outdated training data? Using outdated t
233.
Ask HN: How have you applied AI in your life, work or hobbies?
26 points by azeirah 4y ago|20 comments
234.
by azeirah 4y ago
A good counterexample is Robert Sapolsky. He uses a lot of storytelling and anecdotes in his scientific teachings and lectures. His writing is amazing, his lectures even more so. Here's an example lecture: https://www.youtub
235.
by azeirah 4y ago
"It's difficult to compare a graphics card like the GTX 1070 and an aircraft like the F-16 as they are designed for very different purposes and operate on completely different scales. The GTX 1070 is a powerful graphics card that
236.
by azeirah 4y ago
I wish Wittgenstein was still around to comment about these developments
237.
by azeirah 4y ago
Imagine the nightmares possible of 30 years of layer on top of layer on top of layer of AI written legacy code in an international business O_O We're not talking about millions of lines of code anymore, that'll be hundreds of bill
238.
by azeirah 4y ago
Ran it and nothing really happened?
239.
by azeirah 4y ago
Are the 12k people all software engineers?
240.
by azeirah 4y ago
It's interesting to note that webflow was also inspired by Bret Victor. I'd certainly say this is a succesful product.

More ›