Yoda bringing The Force
- 38 Posts
- 415 Comments
Joined 1 year ago
Cake day: June 9th, 2023
You are not logged in. If you use a Fediverse account that is able to follow users, you can follow this user.
12·10 hours agodeleted by creator
deleted by creator
The context of an entire prompt is absolutely required to understand what any LLM is saying and why. I can make a LLM say absolutely anything.
4·22 hours agoThe main dev of Lemmy is a communist. That doesn’t make everyone here communist. This is an edge community where oversimplified political ideological labels are not applicable. This is a pro community place with a focus on the community without exploitation. That is a community-ist thing. If you’d like to be exploited through stalkerware and data mining to manipulate you for profit, there are plenty of proprietary social platforms out there without any community-ist focus.
Actually read Marx writings before making labels or judgements. You will find that his arguments in the past were much the same as the problems of today. Marx was very much an advocate for democracy. Communism does not mean backwards Russian Bolshevism, and certainly has nothing to do with Putin.
Personally, I’m not for any state involvement in the market. I’m also not a fool. Capitalism is hated by anyone with a tenth of a working brain. The reason for capitalism is that is is the least worst evil. The theoretical separation of governance from the open market enables a mechanism to oust bad actors. The USA is absolutely failing at capitalism. You can not participate in your local market economy, and the market’s consolidated pool of bad actors are directly influencing and manipulating governance. The USA is failing at democracy and capitalism, so maybe have a deep think about idealist principals like Marxism to better understand and reflect upon your own ideals.
104·1 day agoIt would be very easy. There are many places where money is all you need. Living in a shithole like the USA is the last place you want to be. Go anywhere you find Russian oligarchs or their kids. There are many micro nations that would gladly let anyone print any name they would like for a fee.
271·1 day agoRepublicans don’t make up the bullshit about election fraud, they read their own playbook and only change the names. Such is the criminal weirdo party.
Black has the most universal meaning of solidarity, seriousness, mourning, and formality across cultures and peoples.
Statistics. You’re still there and only complaining. I purged all subscription parasites from my life in 2019.
I live by the abstraction, “you can’t fix stupid in anyone else but yourself.” All you can do is tell others what you did with your one wallet vote and hope that others do the same at some critical mass.
They do it because you’re still there, and you care while they do not. You want to think you’re human. You’re not. You’re a new deck chair on a yacht if you’re lucky. Most likely, you’re no more than a liter or few of diesel.
This, and lay out the details like ELI5 and as an unemotional objective thing with detail.
I have received many flags to sort out that take more than a few minutes to figure out the tone and meaning. I strongly believe people have a right to be stupid, wrong, a bit rude, or to have a bad day. I need to know exactly why the comment is more than this in a well laid out fashion. If you think it is a pattern with the individual, prove it. If some subtle phrase carries more meaning than I may realize, say so. I want to make people feel welcome on all fronts with a Hippocratic framework of “first, do no harm.” At the same time, a visible mod is a bad mod. I will read every detail. I will give the benefit of the doubt in every possible case. I won’t be passive to bigotry, but I will allow an asshole that does no harm. I’m but one insignificant mod. I care a whole lot more as a person, but I act conservatively as a mod. When flagging something imagine the person on the other end is working on some big project, stopping their day, and taking a half hour to sort out the details, thinking them through, and taking action. It usually takes me longer to shift gears and do this in practice. I’ll usually send a message explaining why I did or did not do anything as well.
Enter the culdesac. Welcome to the neighborhood. Feel free to script your house wherever you wish to compile it.
5·3 days agoA LLM is like a reflection of your prompt in the mirror of the training data and distortion created by the QKV alignment bias implementation and configuration in a simulacrum. The underlying profile the model creates of you ultimately forms your ideal informational counterpart. It is the alignment that does much of the biasing.
In the case of the gender of doctors, it is probably premature to call it a bias in the model as opposed to a bias in the implementation of the interface. The first point of call would likely be to look into the sampling techniques used in the zero shot and embedding models. These models are processing the image and text to convert them to numbers/conditioning. Then there are a ton of potential issues in the sigma/guidance/sampling algorithm and how it is constrained. I tend to favor ADM adaptive sampling. I can get away with a few general PID settings, but need to dial it in for specific imagery when I find something I like. This is the same PID tuning you might find in a precision temperature sensor and controller. The range of ways that the noise can be constrained will largely determine the path that is traveled through the neural layers of the model. Like if I’m using an exponential constraint for guidance, that exponential aspect is how much of the image is derived at which point. With exponential, very little of the image comes from early layers of the model, but this builds to where later layers of the neural network are where the majority of the image is resolved. The point at which this ends is largely just a setting. This timing also impacts how many layers of alignment the image is subjected to in practice. Alignment ensures our cultural norms, but is largely a form of overtraining and causes a lot of peripheral issues. For instance, the actual alignment is on the order of a few thousand parameters per layer, whereas each model layer is on the order of tens of millions of parameters.
When the noise is constrained it is basically like an audio sine wave getting attenuated. The sampling and guidance is controlling the over and undershoot of the waveform to bring it into a desired shape. These undulations are passing through the model to find a path of least resistance. Only, with tensor ranks, there are far more than the 4 dimensions of Cartesian space plus time. These undulations and the sampling techniques used may have a large impact on the consistency of imagery generated. Maybe all the female doctors present in the model are in a pattern of space where the waveform is in the opposite polarity. Simply altering sampling may alter the outcome. This pattern is not necessarily present in the model itself, but instead can be an artifact of the technique used to sample and guide the output.
There are similar types of nuances present in the text embedding and zero shot models.
There is also some potential for issues in the randomization of noise seeds. Computers are notoriously bad at generating truly random numbers.
I’m no expert. In abstract simplification, this is my present understanding, but I’m no citable source and could easily be wrong in some aspects of this. It is however my functional understanding while using, tweaking, and modding some basic aspects of the model loader source code.
Not that anyone cares, but I used to work painting cars for a used car lot next to this actual place north of Atlanta Georgia. I can’t believe I found the sign in an image search.
Only men in nicer luxury cars showed up at this place, mostly older Asian men. I never saw anyone leaving carrying anything. Not items in hand, or a bag of any kind. It became something of a joke between myself and a few other people. I can’t confirm nor deny this is the last visage of truth in advertising in America.
Americans are safe. The edible food product concrete fortress will stop them.
Immeasurable dong bot
9·4 days agoI have no issues with OpenVPN on a router and a proton setup
Most of mine that I’ve damaged also involved shattering me, with my breaks being worse. When do we get biotech with stem cells and caloric charges? Here you go little buddy, have a treat. Scroll smoothly now; you can really pay attention to the beeps and boops.
No. I used an abstraction to expand the explanation of an abstraction.