causal

40 ms·

841.
by causal 2y ago
Can't read the article, assume is points out the obvious problem of excess carbon in the atmosphere remaining? Ocean acidification is one big issue that a solar shield won't solve.
842.
by causal 2y ago
> Governments need to do better at reducing the burden of choice for end users. Arguably that's what Texas is doing here. Requiring permission to apply facial recognition feels like a very sensible default.
843.
by causal 2y ago
Many places basically expect employees to keep work slack on personal phones too
844.
by causal 2y ago
If anything I think it's easier to have an "I can fix this if it breaks something" attitude when you know more
845.
by causal 2y ago
AI in general amplifies capabilities. In some ways it offers new ways to challenge the consolidation of power. The need for courage has not fundamentally changed.
846.
by causal 2y ago
This is really cool. Also a healthy reminder that anything we post publicly is likely to be analyzed. With a little analysis it's probably easy to know me better than I know myself.
847.
by causal 2y ago
> Depending on how you twist the words this could crush a satire completely. This can crush freedom of artistic expression. This is ripe for abuse because it's so emotionally charged is why you get a unanimous but vote. Who's f
848.
by causal 2y ago
Not a lawyer; agree it seems pretty reasonable and not-overbroad. IMO the most subjective aspect is what constitutes likeness. I presume that's left to judge/jury's discretion? The range of face-recognition ability is pretty
849.
by causal 2y ago
Anyone know of any decent SFW alternatives to CivitAI?
850.
by causal 2y ago
It feels like Reddit is approaching an inflection point anyway where bot-made content is concentrated enough to spoil the whole experience. Closed servers like Discord and Slack may be the last haven of online human interaction.
851.
by causal 2y ago
Yeah a lot of people here seem to not understand that PyTorch really does make model definitions that simple, and that has everything you need to resume back-propagation. Not to mention PyTorch itself being open-sourced by Meta. That said t
852.
by causal 2y ago
You could run a 4bit quant for about $10k I'm guessing. 10x3090s would do.
853.
by causal 2y ago
You are definitely allowed to train other models with these models, you just have to give credit in the name, per the license: > If you use the Llama Materials or any outputs or results of the Llama Materials to create, train, fine tune,
854.
by causal 2y ago
I think you can use an HF account as well https://huggingface.co/meta-llama
855.
by causal 2y ago
"Open weights" is a more appropriate term but I'll point out that these weights are also largely inscrutable to the people with the code that trained it. And for licensing reasons, the datasets may not be possible to share. T
856.
by causal 2y ago
Common practice because accounts get points for editing posts. Others are supposed to review the edits but when a certain critical mass of the population is all trying to game the system the whole thing fails.
857.
by causal 2y ago
Certainly living up to the name
858.
by causal 2y ago
Yeah it will be interesting to see if we ever settle on standard sizes here. My preference would be: - 3B for CPU inference or running on edge devices. - 20-30B for maximizing single consumer GPU potential. - 70B+ for those who can afford i
859.
by causal 2y ago
1) Rule of thumb is # of params = GB at Q8. So a 12B model generally takes up 12GB of VRAM at 8 bit precision. But 4bit precision is still pretty good, so 6GB VRAM is viable, not counting additional space for context. Usually about an extra
860.
by causal 2y ago
Literally said "AI is so hot right now" lol
861.
by causal 2y ago
This looks good. I created something similar using blade a while back, but I found that using English to express what I want was actually really inefficient. It turns out that for most commands, the command syntax is already a pretty expres
862.
by causal 2y ago
Good point. And even if the Apple case was a "samples" issue, I think it's unrealistic to expect it to avoid ever producing a design that resembles any prior work.
863.
by causal 2y ago
Perusing Slack has become this for me
864.
by causal 2y ago
Agreed.
865.
by causal 2y ago
Benchmarks are pretty flawed IMO, in particular their weakness here seems to be that they are poor at evaluating long-tail multiturn conversations. 4o often gives a great first response, then spirals into a repetition. Sonnet 3.5 is much be
866.
by causal 2y ago
I wish Github READMEs had native PlantUML support. My go to is MermaidJS for that reason.
867.
by causal 2y ago
It should be corrected but the interesting aspect of this release is the architecture. To stay competitive while only needing linear inference time and supporting 256k context is pretty neat.
868.
by causal 2y ago
The cosmic beings running the simulation may not be constrained to discrete simulations at all.
869.
by causal 2y ago
The problem is that security making things difficult results in employees resorting to workarounds like running rogue webservers to get their jobs done. If IT security's KPIs are only things like "number of breaches" without
870.
by causal 2y ago
+1 I'd add versioning and diff tools as another critical advantage for text. If your visual tool can't provide a superior diff experience, then it's dead on arrival for most serious projects.

More ›