j4k3

j4k3 , 10 months ago

People being open and friendly on Lemmy.

…and figuring out how to make Silero text to speech work with any length text and got it integrated with a chat bot to talk back to me. I also just got speech to text working. Now I just need to train it on my voice and background noise and I can have a real world voice conversation with offline AI I tune and control.

j4k3 , 10 months ago (edited 10 months ago)

Lol. AI training is more like human awareness of a subject or style. LLM’s are not Artificial General Intelligence. They have no persistent memory. They are just a complicated way of categorizing and associating subjects mixed with a probability of what word comes next. The only thing they are really doing is answering what word comes next. Thar be no magic in them thar dragons

This is just another market hype article. AI can’t reproduce a work or replace the author. It can write a few lines that may reflect a similar style just like any human also familiar with the work and style.

Unless you want to go back to the medieval era of thought policing, all of these questions about AI training are irrelevant.

j4k3 , 10 months ago

I only play with offline AI stuff, but I usually use a model that is way more powerful than the average in the offline LLM space (Lama 2 70B). I have tried building a Leto II character from God Emperor of Dune, and I did several LoRA training layers on various data including the entire GEoD book. The book itself had very little value. It lacks the right context scope and detail needed to do anything relevant. It becomes possible to question some high level plot or character info, but I remember more than the model could accurately pull out.

The really cool cutting edge thing of the future will be authors embracing AI for their own creative inspiration and whomever is the first to fully integrate with AI. Train yourself and your characters so that a screen can generate images of characters and locations as they develop page by page. The identity of each character can build with a random seed with a persistent token so that every read has a different accompanied look and feel. In the future I’m sure it will be possible to limit a character’s context to the information read by the user, (this is a major key to individualized learning - probably the biggest future application of LLMs). A character/user matched scope would make interaction and direct questioning next level cool. Forget book clubs, now you converse with a circle of characters, or the author as a digitized character.

AI is the framework. All the hype and BS is because of stupid greedy people trying to make AI a product. This is next level privacy invasive nonsense. Proprietary AI must die. The future is offline AI as a simple tool unitized to create a richer or accessible future experience.

j4k3 , 10 months ago

Breaking my shoulder blade in physical therapy while pushing myself extremely hard to recover from a broken neck and back. That’s when I had to acknowledge I’m partially disabled, and no amount of pushing was going to change that.

I got much more balanced after, but it is really hard to slow yourself down to match your physical limitations. That was probably my biggest mental challenge to overcome. It took another 2 years before I slept more than 6 hours, most nights were less than 4. I was a dysfunctional zombie for most of 5 years.

OP, I know it is hard, but absolutely any regular physical activity will help tremendously. I struggle to sit or stand for any length of time. Posture is hard on me. It feels like muscles are missing and just failing. I never feel great and usually feel like I can’t, but I still ride a bike most days. I was an amateur racer before I got hurt. So I had a major advantage in conditioning. The balance I get from riding gets me though my lows both physically and mentally. It only takes two months of pushing yourself to develop a physical routine before it will flip and just as quickly it will become as hard of a habit to stop. I was very nearly killed on a bike, but I’ve ridden most days since 2009, and managed to drop and keep off over 150lbs, even in my limited state. Exercise really is the best medicine for almost every situation.

j4k3 , 10 months ago

For me, it has been a realization that there are a lot of small muscles that are used to hold up the human head. No one has been able to figure out what exactly I have damaged, but it is probably quite a few things.

If you do the math for the speeds involved, I basically took the equivalent of a headfirst dive from a ten story building and survived. It really was a “I barely survived” situation; no hyperbole required.

When I push, I break stuff, and I just didn’t see any returns. I still do a daily routine on my own, but there is no reason for me to spend the money on PT with no results. I did that for 3 years. I’m at 10 this coming February.

j4k3 , 10 months ago

I enjoy your work too. Thanks!

j4k3 , 10 months ago

I just went from a 4700 i7 to a 12700…because I wanted to play with software that utilizes every bit of this hardware and then some (offline AI). Not to justify the stupidity of this hardware obsolesce, but it is a MASSIVE upgrade as far as hardware performance. Much more than I expected. Your results will likely vary though. I only use Linux on everything. My single thread speed doubled, the PCIE gen 4 NVME is three times faster than my old Samy Evo SSD, my internet speeds quadrupled with no other changes on my lan.

The following is a deep rabbit hole, but if your gen 4 has a TPM chip, you may be able to manually take control of the PK key by generating your own key set and then setting up a Microsoft key. I think there is info for windows too here:

…defense.gov/…/CTR-UEFI-Secure-Boot-Customization…

j4k3 , 10 months ago

Welp we all know you’re eating steak in the Matrix.

AI porn is interesting for its extreme detail in systems that were not quite designed for it, and what has been achieved with extremely small model sizes. Like a typical chat model of what seems like an equivalent quality as far as accurate detail comprehension is two to three times larger. It is hard to objectively compare these two, but this is my intuitive/highly speculative opinion.

That said, in is hilarious how much some model checkpoints can troll someone. Let’s just say, after my casual experiments to explore how LoRAs and other modifications and enhancements work, there would be many PTSD experiences for anyone that tries this. You might just find yourself reorienting your preferences every time you blink.

Also, if the option is available, you run the serious risk of it becoming an alternative lifestyle, especially amongst those that pursue an academic path and must stay free from distraction. If this is experienced at a younger age, it may remain as a permanent choice. It objectifies relationships and that may prove difficult to change.

I think you will find the only barrier to relationships is really the person in question’s state of mind and willingness to put in effort. If a skateboard has aspirations to board a hundred million dollar super yacht, that’s a mental health issue. However, outside of mismanaged birth policies where the sexes are disproportionately represented, I’m a strong believer that there is a skateboard for every skateboard, and at least a dock and dingy for every super yacht.

j4k3 , 10 months ago

Depends if you are into tech, skateboarding, or yachting

j4k3 , 10 months ago

I am something of a skills collector: Bicycles (roadie/former amateur racer), cars (mods/engines/paint), composite prototyping, metal machining, home foundry metal casting, embedded programming, analog and digital electronics design, reverse engineering, KiCAD, PCB etching, motor controls, python, CPP, Forth, woodworking, Linux desktop, Linux kernel deep dives, computing hardware, open source software, Dune, Asimov, SciFi, astronomy, telescopes, designing and building optics, 3D printing, CAD design, FreeCAD, Blender, cooking better than the best restaurant foods, Asian cuisine, psychology

I am a generalist, but a swiss army knife, I know about and do a lot, but I am like the worst scissors you’ve ever attempted to carry and use. I can get the job done, but am well aware I am not a master of any of these.

j4k3 , 10 months ago

I have no qualms over admitting curiosity is the one trait I find most interesting and respectable, in others.

j4k3 , 10 months ago (edited 10 months ago)

So when it comes to true vacuum Quantum Field Theory proves that things are not as simple as they appear on our human macro level. Basically, there is no such thing as true vacuum in an absolute sense.

Ref: piped.video/watch?v=n6jAOV7bZ3Y

As far as things moving towards a vacuum, it is just stuff taking the path of least resistance at a simple level. Think of air and gases just like a slightly different liquid compared to water. Air and water flow in a very similar way. If you displace water by making a boat, the displaced water is always looking to get back into the displaced space. A vacuum is a similar displacement void that is fighting against everything else that wants to occupy that space.

Honestly, I’ve come across many things where I asked questions like this. I could be wrong, but it is likely that you are thinking of vacuum as some kind of state or force. It is not. Vacuum is the lack of baryonic matter. Maybe you have encountered this one before: “cold does not exist.” Cold is always a relevant thing and really it is a human invention to relate ideas. Nothing can be at true absolute zero in the universe as the motion of atoms is tied to time and entropy. So cold is always just the absence of heat or really, the absence of energy. Much the same, a vacuum is the absence of matter. The main thing that might move matter into a region of vacuum is gravity, or to a much lesser extent possibly magnetism depending on the properties of a gas.

As far as black holes, no vacuum can exist in a black hole. Once matter gets past the event horizon humans do not know any testable physics. Technically, as it is currently explained there is a singularity somewhere inside the event horizon. This singularity is the merging of both space and time. The merging is because you are being pulled at light speed the moment you pass the event horizon. You’re moving at the speed of time which makes physics break. Moving at the speed of time, means you are already merged with the singularity the moment you pass the event horizon, however the event horizon is larger than the singularity itself. The singularity, as we understand it, is time as much as it is space, literally “spacetime.” If it could be seen, theoretically, we could also see all of time on its surface…if it could be called a “surface.” A black hole is not really an object as much as it is a state.

This ref covers a lot of what I just said except for the very last point: piped.video/watch?v=cFslUSyfZPc

j4k3 , 10 months ago

Feels like a minds war waged by billionaires in this space when you’re actually playing with this stuff. All this hype is a joke as are the proprietary junk. Get a decent comp and try offline AI yourself and see what it can do. Try Llama2 70B Q4 GGML. You need a machine with more than 10-12+ cores and at least 64GB of system memory. It really helps to have a Nvidia GPU with 16GB+ but you don’t have to have that here. This model can write python snippets like you’re searching stack overflow but an order of magnitude faster. If you know basic code elements, branching, and looping, this model can code, and resolve its errors when it gets something wrong by prompting it with the error message. A 30B like WizardLM or Vicuna are almost technically useful, but the 70B is a beast.

j4k3 , 10 months ago

Even just in the US this is very regionally different and not really something I associate with generational cultures. Like in Tenn/Bama/Georgia it is very expected but means nothing. In Southern California it is rare, but actually means something when said.

j4k3 , 10 months ago

Often times I keep a notepad beside me. It is my preferred way to store ridiculous length passwords for stuff I care about. I’m usually on a laptop and I may switch it up and use another device to look up stuff. I don’t mix my workstation with socials or shopping. Those three activities are all done on different devices, with different networks. So I don’t care too much about what can be scaped from here. I don’t see value in a small amount of convenience exchanged for connecting my devices, I’d rather just jot down a note and look up the item when I need it.

j4k3 , 10 months ago

Your new name is Blockbusters

j4k3 , 10 months ago (edited 10 months ago)

At least the coupon book is for products with real prices. Healthcare is a total scam with prices based on who is paying. The entire system is corrupt from top to bottom. The US problem is extreme systemic corruption. It is not individual corruption outside of the billionaire supreme court judges level, it is corporate sponsored corruption on a much larger scale.

The USA has a tenth of the laws and protections of any other western country. We have had nearly 50 years of a political denial of service attack from a right wing campaign of misdirection and distraction politics. No one can institute reasonable laws and protections when they are constantly battling whatever stupid inflammatory nonsense that hits the congressional floor. This is why the nonsense keeps happening. It is because it controls the conversation. The only purpose is to keep as many loopholes as possible open for the parasitic worthless billionaires that are funding it. The only fix is to force out the billionaires. The only way to accrue billions of dollars is by exploitation and criminal activity. There are no exceptions to this rule. Every billionaire is a criminal evading prosecution.

j4k3 OP , 10 months ago

The key is the system memory is 64GB and 12th gen i7. Details are in posts below. Top link is with the GPU and bottom is just the CPU. The model itself is very important. I tried several before getting this one working.

lemmy.world/post/2952867

lemmy.world/post/2892804

GL

j4k3 , 10 months ago

Only using Android for socials and only tried element. I want something different. Element has a settings overload, but no way to remove the status bar icon. I hate that so much I just leave it disabled in settings unless I need to use it.

j4k3 , 10 months ago

Yeah. I like my home screen like Gnome’s DE; nothing on it whatsoever. I only want my 3 button navigation, clock, WiFi status and battery. Absolute fine grain control over my notifications is very important to me. I’m extremely burned out on notifications used for user attention manipulation on the google framework. I an on Graphene with no google services. I only want notifications if they can be default opt-in and primarily peer to peer for things like DM. I don’t want any visual indicator for such background services. Basically, I expect SMS like functionality; out of sight, out of mind until needed.

j4k3 , 10 months ago

Interesting. Thanks for the info. I turned off the listening notification. Hopefully this is the one you meant. I hadn’t thought about settings for the device itself.

https://lemmy.world/pictrs/image/b57b74a6-ac36-49c4-8c8f-62e84342cbb7.png

I don’t entirely understand the verification notice and all. I don’t understand the context of verification. I generally avoid making any external internet connections (finger printing) between my phone and other hardware. They are on separate VPNs and firewalls. I get the message about no verification every time I open element. I don’t know what it is trying to verify exactly. I imagine it is a copy of the keys and hashes, but like, I don’t want to connect social across devices and I am given no options or context details. I don’t mind backing up or storing keys, but I’m not installing the software on my workstation or punching holes in that firewall.

j4k3 , 10 months ago

Yes, for social type stuff. I’m disabled. Interruptions are very counter productive for me. If I have the strength to work on something, the last thing I need is some excuse for distraction. It will only be a short time before I will need to take a break, and will inevitably see whatever pending dot notifications I am interested in checking. Or I’ll just check them manually if I care. I don’t mind when notifications are from individuals sending direct messages. I don’t want any other form of notifications by default. I’m fine with something like a dot notification on the app only, like if there is a direct comment reply to something I posted on a general thread, but this is not a phone call, sms or equivalent type of priority to me. It is peripheral, and something that has limited import. It shouldn’t demand my attention with a sound in the real analog world, or shove itself in my face the moment I view a device. Like a student in a classroom, it should quietly raise its hand and wait for its turn if time allows.

j4k3 , 10 months ago

I have just gotten in the habit of editing first and adding the same two characters. Then I delete. If I want to in the future, I can potentially search based on those characters. I mostly just delete because I don’t want to see the replies any more, or get reminded about something every time I view my profile.

j4k3 , 10 months ago

Thanks for the nostalgia. I haven’t seen a login YT home feed in years, and I used to be a CC. YT has been a dumpster fire since the lobotomy of 2017.

j4k3 OP , 10 months ago

I wonder if the kernel parameters for CPU isolation were set, if it would enable this.

j4k3 OP , 10 months ago

Speculatively, I think it has more to do with CPU schedulers in kernels. Junk like windows is super inflexible when it comes to the kernel. Like Windows 11 had to be made just because of the Intel asymmetric cores. Changing the scheduler in W10 just wasn’t viable. Likely having Alder different makes issues because Microsoft is so inflexible with their tiny OS.

I don’t know how Linux is handling the scheduler for intel exactly. I’ve watched most of the Linux Plumbers conference about the changes to the scheduler from 2021, but I don’t recall details about this.

People saying all it takes is disabling the efficiency cores hints that it is a scheduler issue though. I remember someone mentioning that the way cores get spun up and the time it takes to get to full speed is complicated and that management overhead was required, so I’m not sure if the scheduler is the only factor, but I think it is likely the primary.

I’m hoping someone comes along that can identify the mechanism that detects the available instructions and how this can be tested or altered in practice.

j4k3 OP , 10 months ago

It isn’t a direct line to connect, sorry if I am bad at explaining it. The P and E cores mix are the issue. The P cores are Xeon designs that (may) have the extra instructions. The E cores do not. If simply turning off the E cores has made the P cores show up with AVX512 on some systems, I imagine it may have to do with the scheduler. I could be wrong.

The CPU scheduler will require some kind of management function that could bind the process to a core with the extra AVX instructions without manual intervention. This would need to override availability, kernel threads, and things like power efficiency or spin up optimisation. I haven’t taken a super deep dive into how the scheduler is working on a 12th gen. I can say it appears to pin processes more, but I still see a regular rotation of most running processes across cores when they have no affinity or isolation settings, like when I am running a large LLM.

j4k3 , 10 months ago

Open AI died the moment Meta’s Llama model weights were replicated completely open source. The outcome is guaranteed. It does not matter how much better the enormous proprietary model can be, people will never be okay with the level of intrusive data mining required for OpenAI or Google’s business model. Personal AI tech must be open source and transparent with offline execution. AI is the framework of a new digital economy, not the product.

j4k3 , 10 months ago

I am only one of these. I need a drink. The old water into wine glass trick should work

j4k3 , 10 months ago

The biggest selling point for Fedora IMO is the way it handles UEFI and Secure Boot. I haven’t found anything comparable. Securing the proprietary garbage running on your main board is critical regardless of your OS.

j4k3 , 10 months ago (edited 10 months ago)

Without the spam ad insanity:

web.archive.org/web/…/linux-commands-cheat-sheet

On a whole different level… (on enthusiast lvl hardware) a Llama2 70B running with 4 bit GGML on a 16GBV 3080Ti 12th gen Intel with 64GB can do bash and Python completely offline at a cheat sheet/stack overflow level without major errors. I just spent a day modifying someone else’s python script and never went online for anything, have never been good at Python, and haven’t messed with it for years. I actually got more done than I ever have before in a single day mostly because I didn’t need to search documentation. FOSS/almost FOSS/offline AI rocks.

j4k3 , 10 months ago

Must be hard walking around with those massive steel balls clanking around as you walk. Making smelly food on the front line sounds…uh…dangerous… but I have no clue what it is like IRL

j4k3 , 10 months ago

It is too bad no one talks about the obvious. Americans want the jobs. The foreign jobs pay for the same quality of life. In the USA the same quality of life cost is inflated so much by corrupt mismanagement that it makes Americans uncompetitive. Housing and food should be a tenth of what they are. The USA has no shortage of land, just a massive lack of zoning management and reform. The US is one of the largest food exporters in the world. High costs here are a ridiculous corruption and mismanagement. The nonsense coming out of the Right is an attempt to prevent reasonable legislative reform. It is about controlling the conversation in a dos attack. This is the only way the criminal billionaires can keep their criminal legal loopholes open. The USA has 1/10th of the laws and protections of any western country.

j4k3 , 10 months ago

Yeah, but what if the app is more complicated, like it needs to run in a conda container and may have extra flags passed to it? (I’m double dipping in the container sauce for hacking mods to AI stuff)

j4k3 , 10 months ago

So that explains Vader’s helmet

j4k3 , 10 months ago

How far in the future can I go before pathogen microbes evolve too significantly for me to survive? Whatever the answer is, this, but only after making a few minor investments first. Then I’ll go. Place? Switzerland. Why? Hopefully, - stability.

j4k3 , 10 months ago

FD RX7 https://www.carscoops.com/wp-content/uploads/2018/07/3674cb6f-mazda-rx7-00.jpg

I just loved the idea of a rotary and how a 1.3L engine could have such a high output.

j4k3 , 10 months ago

If you can’t ride a bicycle, go out of your way to make riding safer for others by never using an idiot brick behind the wheel, and driving safely. You are not important on the road. You are not an exception under any circumstances. You always need to park in a parking space. You never have a right to double park. Your special situation is a dilution of your own invention. Your delay or rush is your own incompetence and you are the only one to blame. Your 10 seconds or even your 5 minutes is not more important than someone’s life. Back roads are not your secret, an alternative highway, or a racetrack. Drive like the airbag is really a dagger pointed directly at your chest because your car is just as deadly and just as dangerous to everyone else. Pedestrians, including cyclists always ALWAYS have the right of way. No matter how prejudice you are and how anger this makes you, you WILL lose in court, and likely everything you own, as hitting a pedestrian doesn’t have the same limitations as hitting another vehicle; your liability insurance won’t even remotely cover this issue. I was disabled while riding to work in 2014; by someone who double parked; then made a sudden u-turn directly into another car and nearly killed me with a broken neck and back. Stupid drivers should have lethal or at least life changing consequences. “Safe” cars create stupid drivers. There is no effective low bar for a drivers license. The number of times a person can take the test is infinite. There is no effective white and blue collar alternative public transit system in the USA outside of Manhattan. Buses are not remotely professionally viable, and trains like in SoCal have inflexible, sparse schedules that just don’t work in the real world. I have tried them. There must be an efficient nearly point to point method of transit every 15 minutes 24/7-365 or you work for your transportation not for your job.

j4k3 , 10 months ago

There are only a couple thousand billionaires. Their locations are listed here. Today is a World holiday. The official end of colonialism. Redistribution day. The purge.

j4k3 , 10 months ago

I’m not inciting any particular outcome. Send them all to be little emperors of Corsica or to a reunion on Reunion. Billionaires are all criminal parasites. That kind of money is not possible when treating people rights and working honestly. Everything that is wrong with the world right now stems from this corruption of a few people that have been left completely unchecked. They bribe and fund more and more corruption in a spiraling feedback loop that will never stop. Their social position must be eliminated. Whatever that means for the human under the surface is what it is. 2k people are irrelevant to the lives of 8 billion humans.

j4k3 , 10 months ago

Foot balls is a thing. Most humans now have at least 4 balls.

j4k3 , 10 months ago

Porn has sound?

j4k3 , 10 months ago

Yeah. I’m exclusively Libretube and Piped.video now.

j4k3 OP , 10 months ago

Thanks. I went down this rabbit hole already with large FreeCAD assemblies where single threaded operations had an enormous amount of non voluntary context switching. It helped some in that instance. I don’t have a benchmark, but I reduced the NVCS by an order of magnitude using Tuna to set affinity and pinning. Editing a large assembly in the middle of the tree went from 5++ minutes to around 3-4 minutes.

I’m not claiming I know better, or that this is a bright idea. As mentioned, it’s an abstract curiosity. Playing with FreeCAD got me interested in the OS on a deeper level, and got me reading as much as I can about the CFS and others. The LLM toolchain I’m working on is to help me follow the CS curriculum posted freely online from UC Berkeley. Ultimately I plan to use a langchain database to help me follow along with the books and transcribed lectures. It is just a hobby interest.

Thanks for the insights!