Interestingly enough, both the GitHub repository + user apparently been deleted from GitHub. People who ran this project might want to take a look at the code/dependencies if they didn't do that beforehand.
> will undoubtedly influence the opinions of millions.
Is this actually true, based on just one statement?
Usually the ones who change their minds so easily, are following a person like a cult almost, and go with whatever they say, whenever they say.
There aren't huge masses of millions of people idly watching TV, coming across an interview with some random baseball star and trusts his opinion on stuff.
Personally, I think in generally the problem is "celebrity worship", and all the related things, rather than just about the specific person who said something, or even about the platform. People are so quick to just go "Well, I like that person, so whatever they say, is what I think more or less" which is absolutely bananas to me.
Well in fairness so long as you describe the goal as "largely destroying the frontal cortex and all the functions that go along with it," then lobotomy is incredibly sound.
I mean, if we wanted every human on Earth to be over 6' tall, we could just cull the short ones (or stop them from reproducing), and this is scientifically valid, but, of course, evil.
As far as lobotomy, I don't know, but it seems to be widely discredited inasmuch as it didn't achieve its aims (or if it did, the aims themselves were highly questionable)
Given the title, lobotomy is an apt reference "The originator of the procedure, Portuguese neurologist António Egas Moniz, shared the Nobel Prize for Physiology or Medicine of 1949 for the "discovery of the therapeutic value of leucotomy in certain psychoses"," - https://en.wikipedia.org/wiki/Lobotomy
> an informal term for the embrace of strange or scientifically unsound ideas by some Nobel Prize winners, usually later in life. It has been argued that the effect results, in part, from a tendency for Nobel winners to feel empowered by the award to speak on topics outside their specific area of expertise
Sounds pretty much like every founder out there who strike gold once in the past now believing their genius now clearly applies to everything everywhere, including country governance, economics and sociology. It probably doesn't help that others around them are affected in similar ways and become more sycophantic because of the "award" these two types of people win.
I thought this was how Nobel Prizes sickened fields.
Mathematicians believe we're lucky to have been spared the Nobel Prize. While some repeat a discredited story this was because Gösta Mittag-Leffler was sleeping with Nobel's wife, it is generally felt that mixing ego and the possibility of a Nobel Prize has toxic effects on a field.
We now have both the Fields Medal and the Abel Prize, but both pass over the midlife period when ego eclipses talent.
I like how China and Russia are mentioned as adversaries that the US might be less likely to deter with their dwindling supply of long-range munition, but the country that so helpfully "received" (being bombed) all that supply and is the reason for the stockpile being minimal right now, isn't mentioned as an adversary that could take advantage of this.
Think Ukraine is the reason the US is low on missiles? Do the numbers actually support that? The numbers don’t seem to support that. From a very cursory search, offensive missiles seem to be around 2:1 Iran. And that’s just raw numbers; someone other than me who might do a little moral editorializing would point out the use in Ukraine was to fight off an illegal invasion while the most effective strike in Iran killed a hundred schoolgirls.
When I was a vision engineer at a previous role back in 2022-ish, laptop-class Intel chips used in their SFF NUCs were powerful enough to run small finetuned convolutional nets by dancing across the different on the package: first decode using QSV, and then downsize using the CPU, run the images through an OpenVINO model running on the iGPU, and aggregate results and append metadata using the CPU core. You could have 50% of the hardware in today's high-end security cameras for 10% of the cost.
I have to assume that hardware advancements in the past four years have only pushed the efficiency envelope further.
Yup, this is how I see teams who care about quality, design and architecture seems to use LLMs, not to "produce more and faster" but to retain same speed but with a lot more confidence and reliability. Hoping this will spread eventually, some companies seem to take a more... hazardous approach to the whole thing.
> Poor coders have always existed. You're not the first and won't be the last.
Did you get so emotionally hurt by another person saying they didn't want to actually type code anymore, that you felt the need to tell them they're likely a "poor coder"? You can still continue to type code if you'd like, no need to put others down because you prefer a different way of working.
> work with types of knowledge that are inherently non-text
What are those things exactly? AFAIK, everything we can "know" can be written down, one way or another, even analog circuits.
Also, what SOTA LLMs are you referring to? GPTs been handling analog circuits fine for quite some time, I want to say for at least one year? I've been "pair programming" a bunch of working circuits with GPT models since probably GPT 5 or so.
> AFAIK, everything we can "know" can be written down, one way or another, even analog circuits.
writing down circuit diagrams is like cooking about music.
> what SOTA LLMs are you referring to?
I have done a survey among analog electronics designers just a couple weeks ago and they all said that their forays into LLMs were great for digital electronics, code, and firmware, but for analog they were pretty terrible, with a variety of LLMs, according to everyone.
I think that's different though, that's "doing" rather than just "knowing". You can ride a bike without knowing how it works, and obviously vice-versa too. I don't see circuit diagrams as "doing" though, but the soldering part of building circuits definitely is that way though, you can't just read about it and excel first time you pick up an iron, you have to practice and understand it with your body, like bicycling.
FWIW on a 6000 Pro it takes 68 seconds for 10 seconds 480p video, (cold) same demo workflow as you used. Set "megapixels" to 2.0 (1920x1088) and same video seems to take 5+ minutes, not sure if everything is right/correct at the moment.
As far as I can tell, the current ComfyUI nodes don't even do compilation, and I haven't looked into what attention mechanism they're using, but I'm sure with time these durations will come down even more.
People on reddit have definitely pointed out that sageattention will speed up the renders.
And it's literally the first day. Someone will make a distilled 4-8 step LoRA and we're off to the races.
Edit: did a couple of 10 second long 864/480 i2v videos on my RTX Pro 6000: sageattention bumps them up 33%, that is to say, 140.89 seconds without sageattention becomes 105.69 with sageattention on (if using the KJ Sageattention node, "allow_compile" doesn't seem to affect it, just "sage_attention" set to "auto" works fine).
EasyCache also appears to work, but does affect quality, at least with the default threshold or even down to 0.10. Still, at 0.10 threshold the same render above, with sageattention, is down to 71.33 seconds, so depending on your use case the quality hit might be worth it. Also it seems that with EasyCache the video still matches the un-EasyCached video (with the same seed), so you could use it to do seed hunting.
reply