qumpis

15 ms·

121.
by qumpis 3y ago
I wonder why it's slower at inference time then (for members using their web UI), or rather, if it's similar in size to gpt3, how gpt3 is optimized in a way that gpt4 isn't or can't be? I'd expect that by now we wou
122.
by qumpis 3y ago
I understand the joke, and am only saying that context wasn't there. Maybe Chopra proceeded to clarify his argument after the question asker had left, but the clip doesn't show that. It also doesn't show that Chopra ignored t
123.
by qumpis 3y ago
I've never heard of the person, but the video you linked doesn't seem to do a good job of exposing charlatism at all, it's a rather strange meme of a moment taken out of context. Edit: this random post on reddit seems to ques
124.
by qumpis 3y ago
It's a spectrum and OP offers to get closer to one end of the spectrum when it comes to authenticity. What's the fuss about?
125.
by qumpis 3y ago
In the "sparks of AGI" paper, authors noted that the unicorn shape degrees as more "alignment" is injected to to. If openai adjust the model (say by training more), the picture should reflect it. If they make the model b
126.
by qumpis 3y ago
Interesting that it's so insufficient for you, maybe indeed stuff you do is novel and would require lots of instructing before helping you. Personally, my usecases involve quite standalone applications. copy pasted from another thread
127.
by qumpis 4y ago
Why would Bard get axed? Or do you mean Bard in its current form? Last I heard they mentioned about rolling out higher-scale models, since apparently current models were 'efficient' ones, to see if the demand can be met, according
128.
by qumpis 4y ago
Yes I think since it requires to look up your own comments to see if you got any replies, it's quite common to not get any replies, and I'm very guilty of this myself. My personal use of gpt4 (also daily) is: correct, rephrase spe
129.
by qumpis 4y ago
Can you give some of your usecases? Is it involved stuff or mostly boilerplate? Curious how a team lead uses this tech.
130.
by qumpis 4y ago
Dreamer v3 in model-based direction had some interesting scaling plots showing a pattern of faster (per-sample) learning using bigger models. In terms of generalization, Ada by Deepmind was also quite impressive, but operated within simplis
131.
by qumpis 4y ago
How do you define extrapolation at the scale LLMs operate? Even if you work with unseen-to-model software, it seems sufficient to "understand" the documentation and code examples to orient itself to helpful context. That's wh
132.
by qumpis 4y ago
I felt similarly before I started using GPT 4. Then I got scared
133.
by qumpis 4y ago
I wonder how good the neural engine with the unified memory is compared to say intel cpu with 32gb ram. Could anyone give some insight?
134.
by qumpis 4y ago
Inference, even fine-tuning a few layers would be difficult since one needs to use non-quantized model, I'd imagine
135.
by qumpis 4y ago
Has anyone used spaced repetition in their research/job for a problem they are trying to solve? I sometimes feel like there are many aspects I need to hold in my brain at once and keep having the need to revisit them. If I could chunk
136.
by qumpis 4y ago
Nice to see progress on this end. I've been hoping for some time for a continuation of AI generated shows (like the previously-famous Nothing Forever) that can 1) interact with the open world and 2) keep history long enough (e.g. by re
137.
by qumpis 4y ago
Yes, I have no idea why people are comparing apple's excellent hardware to gaming laptops with wobbly hinges and inefficient CPUs that are intended for different purpose altogether. All tech in Europe is costlier. Check Dell 2023 XPS l
138.
by qumpis 4y ago
How do you know what the representations they infer contain? Why are these void of a model? Why the way of their learning is the answer of their abilities?
139.
by qumpis 4y ago
Why does it require spatial reasoning if it can learn the (logical) rule of how the mirroring and glass doors behave?
140.
by qumpis 4y ago
Learn how? I think having infinite context is perfect - no need to learn on my data online and risk exposing it to others.
141.
by qumpis 4y ago
Sure, but the fraction of people who would go out of their way to access these is presumably small(er), especially those who havent yet tried them and who would nt jump through so many hoops to acquire them if they dont know their effect. I
142.
by qumpis 4y ago
How is it relevant here?
143.
by qumpis 4y ago
When it comes to laptops, I don't see how apple is overpriced. I'm currently in search of a well-built CPU-performant laptop with a decent bettery, and the likes of XPS and Thinkpads cost about the same or more than a similarly de
144.
by qumpis 4y ago
Well the inference time of gpt4 seems to be far greater than gpt3, so it could hint a difference in parameters count.
145.
by qumpis 4y ago
Offtopic, but for what purpose are you running llms locally (especially everyday)? My understanding was that the prompting requires to make them work at all was too great.
146.
by qumpis 4y ago
Why do you care what others think of you, even if you show your true self? If they're not repricocating, then maybe your behavior isn't as innocent as you think? It would be great to see an example of how your behaviour is in the
147.
by qumpis 4y ago
It requires lots of manual control but from my experience I can match or exceed battery life in comparison to running windows, with some tweaking. Out of the box it was terrible as well.
148.
by qumpis 4y ago
Or to just imagine a 1000 boxes with the same problem formulation
149.
by qumpis 4y ago
What's your ram usage for this?
150.
by qumpis 4y ago
I'm also confused by this. If everything was done properly, test results on the holdout set would've been shown. Wasnt that the case?

More ›