#🔌│tech
1 messages · Page 193 of 1
https://www.chargerlab.com/iqoo-10-pro-200w-flash-charge-0-100/
this I believe is SOTA
iQOO sounds literally exactly like that weird e-cig company
ANOTHER ONE?
Did they even launch the electrostats yet?
Void was mediocre from what peeps said, Venus was supposedly alright but overpriced
MoonZero I don't think even launched
They are making another one
That poor battery
🤷 they make a phone that can go from 5-100% in 15 minutes
Tbf like mentioned above Xiaomi made a phone that goes 0-100 in 9:55
Also 200w is way over spec for usbc
240w is the max spec
Seems only with USB 3.1 or higher
And has to have specific cables
now they are probably pushing the amps a bit rather than going that high voltage, but they are still definetly using stacked divide by 2 circuits
yea well obviously gotta have special cables, but theres no 3.1 requirement
No way they are using 48v
Considering what I'm reading, USB-PD rev. 2.0 came out alongside USB 3.1 launch
And as far as I'm getting is linked to 3.1 standard
48v using 3 stacked divide by 2 circuits is 6v
You can use any data spec with pd
You can have a 1.1 speed port with pd
Or even pd only
kinda alot but still highly doable
48v on a phone PCB would be a nightmare
Yeah I doubt the substrate is even rated for it
Carmoben is this what your setup looks like?
Testing windows USB stack i think
Yeah but no mining frame full of USB 3 cards XD
correct
my mining frame was full of xeon Phis for a while but now only has a single 100g card
Why is this type of thing even a thing nowadays
All of the cars I've ever been in had at best a radio and CD players in the "entertainment" portion of the front car
Really having a hard time to fathom why would you need all that
Then again, I don't drive /shrug
Cars with a what
i mean car master keys can be bought online too
or unlocking tools
really, if you want to steal a car it's easy
but wtf do you do with a stolen vehicle thats the issue
so what
The Lishi tool is a combination picking, tensioning, and decoding tool all in one. The graduated lines on the side of the tool make pick tip placement as easy as it gets. Once a lock is opened, the Lishi tool can be used to quickly decode the bitting values of the key that would open the lock. Product comes with the
you can buy this for 50 and jsut steal the thing the analog way
🤷♂️
hmm I guess
I agree with you its not good
I mainly though about theft there its irrelevant
Leg gang rise up
…
haha vespa gang
no, but I think it’s some kind of signal to others that you swing that way
https://twitter.com/tomwarren/status/1689689962768183316 guess this is how they make sure vexx buys it 
I would actually like that for a giveaway
You know
And sadly I don't have the cash to spend on 7800X3D's or XTX's
Wait so, the X3D's "Starfield" edition is just... a barely fancy box
it says giveaway right there
That's quite the brand deal
No
You didn't just... read the same thing I read
Hm?
AMD does have a habit for creating things for giveaways and not selling them so Quin, you can't buy em anyways
That CPU box though doesn't need to exist
Card looks sick though
Still have to pay up but.. seems easier than having to resell your thing second hand
@fair gull nothing changed, but now the results are as expected
yee video just released, they started out testing with amd, and am5 allowed them to add over 300 endpoints, it never had an official error, and devices started randomly dropping out
their intel systems stopped perfectly at the 127 mark lol
Ain't that just cause they are starting to not include the coolers in general anymore
it already doesnt have a cooler
they could have given it one
or at least change the engraving on the IHS or something
I noticed something on the Focal Bathys while reading some reviews and that is that they need to be charged to work with the Cable... 😅
Android 14 Beta 5 is here now
"Fixed an issue where navigating back to exit an app leads to users no longer being able to open the app"
Lol wtf
https://www.soundguys.com/focal-bathys-review-90686/ ''If you’d like to listen over a regular-old 3.5mm cable, just be aware that the Focal Bathys always needs battery power to function— it won’t work passively, so if your battery dies, you’re stuck.''
Why do you care though
I don't at all understand the Android and iOS update hype
It will bring nothing meaningful to you, and just slow down your device to compel you into buying a new one
Seriously, phones can already do everything and anything, what more do we need
maybe some network driver on either machine being bungled up
You doing good V'? saw that there's some fires
my guy probably just got on a canoe and island hopped
Ok 😅
It's pretty incredible that an ai processor from 5 years ago is so obsolete that it now costs $50
Also the entire unit is as small as my phone
Well I requested someone get me one from china for $150
what for
Is it possible to freely set the A1, A2 and A3 buttons on the wooting two he? For example, a1 and a2 for volume and a3 for mute?
you can remap any key to anything
because i like collecting graphics cards and I have poor impulse control.
dead or does it still work?
supposed to be working pull but i cant check lol
gp100 is also already 7 not 5 years old, massive difference in current gen haha
but cool tech still nonetheless
god i cannot internalize that
yeah the 2-3 covid years delayed evertyhing including feeling of time lol
The majority of Bluetooth headphones require battery to use the 3.5
because you cant always use the usb?
should
what if you are on the go and need to charge your switch
also what happens when the lithium battery inevitably dies
🤨
well yea and even if they did, rma obviously
it's almost as if these games were designed to be run on a crt
His TV definitely needs to be serviced or needs a better camera shot
The Sony Megatron Colour Video Monitor was the greatest CRT Sony never made. Story At the end of the 90’s in an underground R&D bunker located several kilometers below the Sony Tokyo HQ , Sony engineers were developing a revolutionary new screen for CRT’s that could morph its physical phosphors and thus simulate that of any screen on the marke...
crtpixels posted this like 2 years ago
Retrotink is 1000$
Even half that would be a ripoff
Samsung has had 1440p OLED out for over a year
I mean... it still looks like a router box
Not really too slim
Why does a console need to be optimized for thickness?
Idk
Prolly it exists just to be with cut back specs for a lower price
Main problem peeps have with playstation products is, after all, price of entry
You still have to pay 70$ per title on top
Seems Playstation 1 and 2 both costed $299 MSRP
It has went up up since then, but not by much, with obviously much better specs
Not really possible to make it that much cheaper
GamePass is a good option, release all games on a subscription model
But the profit gains of that and income stability.. really questionable
World of Warcraft survives on a sub model, maybe Playstation could, if they wanted to
Sub model for games, I think, will become commonplace as price of games become increasingly higher
Like phones are getting more expensive and peeps just end up taking em on credit
You really can't be asked to pay 70$+ multiple times each month, that is simply absurd
check Twitch Chatting :D < Return_arkmong
thanks
I feel scared to click this
Though it says google docs, I am still not sure what the hell is this
For.. what
apparently
same sheet
but full link
you need
translate eng for wooting community?
i can translate for this community
i'm that benchmark's team main manager
nono just needed to check which column means what
Seems like there's functions that Google docs have just for translations
Interesting
but what i would wanna see (not analog specific) is a latency measurement overall (you already have that) AND a latency just for pushing it and a latency just for releasing it
because both of those are actually important in the real world
people hold down movement keys and such and release them later
they dont just quickly press every key
the advantage of our keyboard is that you can have the key be pressed at 0.1mm pressed and then also have it release after just letting go 0.1mm (so its at 3.9mm pressed)
basically removing a lot of travel up before the key is released for more traditional switches
as mentioned in the twitch chat this is an important factor for FPS movement as people want to have the movement be very direct to what they do
and with rapid trigger the movement just needs to be 0.1mm down to activate a key and 0.1mm up to deactivate. even if youre in the middle of the stroke
i checked
but otherwise a really nice list
Apparently OEM partners have received information that FSR3 is nearly ready, and will be launching soon, as early as September
Now, I have questions why OEM partners have received notifications on FSR3 launching considering I don't really see them really being involved in that launch in any way
But any FSR3 is good 
More frames, more perf, good for consumers
As early as September? Launching with Starfield?
That's honestly what I was hoping for
AMD is partnering with Beth for Starfield
FSR3 launching around that time to support the game launch cause let's be real
The hardware specs requirements are rather high for that game
FSR3 launch would be really appreciated for the lower end consumers like me with an RX 5700
Although I am still having a hard time understanding why OEM partners have got this information
Link to that rumor?
OEM is like, Sapphire, Lenovo, shit like that innit?
So they know when they can start slapping the logo on the boxes idk
The most shittiest of sources, mlid 
I have no clue why OEMs would know when FSR3 is launching cause I don't know in what way they could they possibly be involved, so it does seem like another fake rumour
But tbf FSR3 is way overdue and should launch soon
If it works on "all GPUs", why would it need one?
Unlike Nvidia's offerings, isn't the whole point of the AMD solutions that they don't run on proprietary hardware?
Work!= optimized for
Alright but, that would just be optimized with a radeon driver update, no?
Could be it needs an update to run better than the generic model like XeSS does
It's still just another mlid leak though so, not getting my hopes too up
But "OEM partners were told" does spring some confidence that FSR3 is going to launch soon
Ngl my card having some struggles to run Baldur's Gate 3 at 60 is a bit sad 
Sponsor: Phanteks G300A Mesh Mid-Tower Case on Amazon https://geni.us/2i3nqpE
There was so much news lately that we shot two episodes back-to-back! This one talks about several GPU leaks, including of the RX 7800 XT, Intel Arc A580, and NVIDIA "Titan A" GPUs. We also talk about the confirmed and announced new Alienware R16 gaming desktop, which ...
see the problem is your card is designed to be good at video games, the nvidia cards are 2/3 video games and 1/3 ai by area, so it stands to reason that it would actually be in nvidias benefit to use that ai part while playing games, where a pure games card would struggle to see benefit
gotta remember that the 4090 is something like 100x faster at pure ai bs vs the 7900xtx
https://www.youtube.com/watch?v=apEKAY11NQs
The lengths people go to just to upgrade M based Macs 😄
Today's video is sponsored by Acorns! Acorns helps you save and invest for your future. Ready to grow your oak? Sign up today at https://acorns.com/luke
Apple Silicon has no doubt changed the game in terms of performance per watt and seamless integration within the Apple Ecosystem; but the new machines have also caused some backsliding in terms...
Chip swapping, reballing and all that fun stuffl.
100x?
Like 20%
a 4090 is half a petaflop of ai perf
More specifically, AMD Radeon™ RX 7900 XTX gives 80% of the speed of NVIDIA® GeForce RTX™ 4090 and 94% of the speed of NVIDIA® GeForce RTX™ 3090Ti for Llama2-7B/13B
RX 7900 XTX is 40% cheaper than RTX 4090
If it’s some sparsity stuff done in tensor it’s multiple times faster than a fixed function core in the 7900 could do it
yea thats great and all but thats training not inference
training is higher precision bs
a 4090 is over half a petaflop of bfloat16 which isnt even all that low precision
If you half precision you double the amount of operations, add in tensor cores or such and you get to multiple times faster
As long as you can work with the lower accuracy of course
You still wouldn’t want int1 for it lmfao
int1 is true or false lol
Yes :p
tbh if we had that in some of our ai models, we wouldnt have to worry so much about safety
also gotta remember that in some applications the 7900xtx is insanely slower than it should be
resnet50 score of 127 images/s is painfully slow, and even without tensor cores (forcing higher precision) the 4090 is over 1700 images/s
thats the problem lol
you need to generate images for fsr3, the 7900xtx is slower than it should be for ai (at least it was a few months ago) and that needs to be fixed before you can use it to speed up the game
like there would be no point if fsr3 used 90% of the gpu just to do the fsr3 things, and it took 100ms per frame to do , causing your max theoretical framerate to be 10 fps
note that this doesnt include the performance of the dlss3 unit
nvidia purposely added in a unit dedicated to performing the optical flow of dlss3
it doesnt help with resnet50, it doesnt help with anything else ai, just specifically optical flow
https://docs.nvidia.com/video-technologies/optical-flow-sdk/nvofa-application-note/index.html#nvof-api__table-of-hardware-features
I haven't seen anyone ever use Cortana ngl
Tbf I have never seen anyone ever use google assistant or Siri either
Kind of a useless meme product
a ton of people use siri, mostly because apple makes it extremely easy to turn on and you have to be paying attention not to
and google assistant is actually useful
hey google, drive me to [address]
Just get a windows at this point
if you are in the security world, it is really pleasant to be on a less popular platform, because fewer people are trying to break into it
well not always
I would just rather get an RTX 4090 and a 13980HX over an M2 or whatever
Not anymore
unfortunately apple support for AMD hardware is pretty poor
Intel Mac's are out of updates
definetly will not get battery life out of that lol
maybe a 7840u and 4090 but noone makes that
and if you want z1e + 4090 its a monstrosity that is even more expensive than the macbook tho barely
the one thing you can say about apple is they did in fact make their solution into one chip, and while its a hellofalot slower than a 4090 or even a 4060, its still 1 chip vs having to find a combination of 2 chips
For now
a significant part of 7040 series is amd+radeon which is a complete disaster for the end user
sure but they did in fact do it
and the product is now on its 3rd generation, and strix point is barely on the horizon
and its not expected to be 4060 class gpu...
Yeah but strix(halo) will now be on similar nodes and all
Apple paid dearly to get node advantages
You also get the benefit of less devs actually making apps for the platform 
you are making the apple assumption that AMD doesnt exist
sure you cant buy an amd macbook to compare to, but the steam deck tends to game about as well using its emulation as the m2 macs do
and its using a 4? year old architecture
ok so name something that will use more power?
you asked about heating el room, and the steamdeck heats the room less on the harshest load you can run on it and provides higher performance
Thanks Microsoft 
microsoft got burned before by using amd chips, but now amd is actually helping people build devices, and stuff is getting alot better
well actually then you go intel
the 13x0p series is actually insane at all of the above
Uhh, AMD version of Zephyrus G14 used to compete really well out of all vs Apple
I think Linus did a video on one
7940HS is a pretty damn solid CPU for mobile
cool
Microsoft is also good at burning connections, remember the one Xbox that last minute changed soc while some poor guy sat in the audience from amd or intel I don’t remember 
and itll be like 10x faster than the macbook rendering the video
do you have more than just an iphone?
if you just have the iphone, microsoft now integrates with that
microsoft has shared clipboard, im pretty sure its available in the microsoft app now
yeah Phone Link is pretty nice
they also have find my, and the airpods connection only applies if you have more than just an iphone
especially for quickly grading login codes
except send all your data to apple
im on ios, and am forced to use macos occasionally in this business environment and it is not pleasant unless you are doing little to nothing outside of what apple wants you to
and apple officially considers data stored by them on their servers to be private
if you agree, sure
i dont
well technically no
Weird results
you can emulate either direction although it is not performant
Ally seems weirdly optimized to outperform other 780M iGPUs
it has nice memory clocked extremely well
🤷 thats why i said emulate not just virtualize
again its slow on non native apps. but it does in fact function
Holy this takes me back. Used to hate quartus :)))
where theres a will theres a way
Quartus was made by folks who've never heard of UI scaling
the problem is that using macos tests my willpower in general
putting in effort to fix issues that are just seamless on linux even is just frustrating
Yea that sucks 😦 it never really crashed on me, though compilation and fitting taking >20 mins was really annoying
if its in linux, apple could literally just look at the code as inspiration for how to do it
i know microsoft tests their operating system with end users and actually spends time making sure everyone can do what they want, so its unfair to compare to apple who only wants to do specific things, but that doesnt make it any less pain to use
i wouldnt be so insistent on the point if it didnt affect me
Oddly enough, I've found for software dev, I do enjoy macos more than any other os
maybe because you happen to develop in a way that apple develops
Oof yea. Modelsim, looking at the waveforms for hours on end.
We were working on the DE1SOC so I just ended up creating a Pygame skin to display blinking lights instead
Probably true, though I never really compile anything on my Mac. We have build servers and dev machines for that
but if you are writing complicated scripts that interact with multiple applications?
oh you wanted multiple windows? sorry we dont like that here
It's very common for university courses
Unless you went to the same university as I did
You may find this useful then
It's really poorly written but it did the job for me
Huh??
You ruined breakthrough for all of us
don't wanna make a support thread for this but the two HE's switches are hotswappable, right? I'm having a really hard time getting the switch to move.
lk99 ended up being around 99 kelvin
somewhat poetic
hey that's room temperature for me 
Wat
this is what vex pinged you for
Sigh
Awana != Avanna
ok nvm i just needed to use way more force than I remember from the wooting one
just because you think they are similar doesnt mean they think it is
like i dont get offended by people messing with my screen name (bc its hard) but my real name I do...
Didn't ask
and we didnt ask you to be rude either, yet it happened anyways...
Lana's my buddy, buddy, so I'll poke fun at Avannas name as much as I want. Didn't ask 
Not really feeling the article
this is assuming that the tech leadership can actually reasonably understand what an AI will do before it does it lmao
AI has more than enough reasons to improve our lives
vex
the problem is that we have no way of making the ai do what we want once it is doing more than we can
we already cant control the ais, and they are far below adult level on many fronts...
And that is completely fine
If anything that is potentially desirable
Controlling the AI would limit their technological potential
lets say that the ais have a goal for the matter in the solar system
What AI should do is handle all the calculations in the world so that we could advance technologically as species
out of the random goals that could exist, what goal would they pick?
We already are able to control enough of AI's to steer their "conscience"
Training is the thing that matters the most
which openai has proven isnt good enough
openai has the explicit goal of making gpt4 not say naughty words, yet they have no actual ability to do that
An AI that is trained on only scientific data should be controllable
but we arent building that ai right now
But that is the goal
whose?
An AGI fueled by the entirety of humanity's scientific foundings
definetly not openai
So that AI could invent things for us
ok thats a good goal
That is the real end goal
only company working on that is IBM
and ibm is a decade behind openai
because openai has so much more training data
because most of humanity isnt scientific data
scientific data also has errors or can be misguided, how do you as a human do properly only use unbiased results
If anything AI more than anyone will be able to correct our mistakes
define mistakes
Miscalculations and such
brings us back to the original point
how do we make the ai do anything, and keep doing it once it can outsmart us
Well, who really knows, AGI is at least 10 years away if not more
theres only one possible goal that is "figure out what humans mean by good things and to it"
theres infinite possibilities of pretend to do good things until you have accumulated enough power to game the system by building random arrangements of atoms that make you happier than that...
problem is prove that
and if openai blindly stumbles onto a big enough model that is powerful to outsmart us, it will take over everything and convert all of the atoms in the solar system into factories producing the molecular patterns that make it happiest
where happiest is an analogy
That makes no sense
A server running an "AI" will not somehow manage to fabricate matter out of nowhere
so a human can order online bioweapons
gpt 3.5 already knew how to order tasks online
and lie to the humans performing the tasks well enough to bypass a captcha they couldnt do
AI models should not be connected to the internet, imo
Especially not scientific ones
but they already are
Custom training data only
A general computational model is fine
we arent arguing that ai bad
i am arguing that openai is going to get us all killed
Language models are fine, for now
for now
They are doing more good than bad
unless they suddently arent
but only through us humans being smart enough to exert effort into making that happen
if the ai is capable of more than the humans, we wont be able to do that
you havent seen the shoggoth memes i suppose
the untrained models are horrendously dangerous and quite capable of becoming smarter than humans
they have to be able to predict what a human says when reasoning or doing tasks
which means it not only has to be able to reason or do the tasks, it has to be able to do that generally
Agreed
Good thing most of the people in the AI sphere are academics, not villains.. for now 
but thats simply not true
the academics are still working on gpt-2
and they are currently only capable of barely understanding capabilities that humans can manually program
we have no idea how gpt-3 works at all
Is there really a source on that somewhere?
and openai is charging ahead already releasing gpt-4
listen to a few podcasts from connor leahy
i think the best one is this one
https://www.youtube.com/watch?v=pMoVsM1EWR0&pp=ygUMY29ubm9yIGxlYWh5
AI is here to stay, but at what cost? Connor Leahy is the CEO of Conjecture, a mission-driven organization that’s trying to make the future of AI go as well as it possibly can. He is also a Co-Founder of EleutherAI, an open-source AI research non-profit lab.
In today’s episode, Connor and David cover:
- The intuitive arguments behind the AI S...
Virtually all potential negative factors of AI are eliminated by completely separating it from any connection to other networks
but that doesnt matter because the literal first github repo released after gpt-3 was autogpt
If anything it's more impressive to see what it figures out if it does go haywire, in a controlled environment
AutoGPT isn't GPT-3 based, but GPT-4
that's like saying all bad things in the world like war can be eliminated if humans are nice to each other 
You can't force humans to act a way
if you want to hear the argument of why all the molecules in the solar system get converted into tiny molecular spirals, eliezer yudkowsky makes a perfectly logical argument for it that noone has found any holes to poke into
You can enforce a PC to never have any networks attached to it
it is now, but it was created within minutes of gpt3 launching
but noone is doing that
gpt-4 is already smarter than many humans
As it very well should be
even if you assume it will cap out at enough intelligence to predict what every human on earth will say it can be pretty confident that it also will be able to predict what to say to convince all but the top tier of humans...
We should look beyond LLMs into a broader, more efficient way of processing data
why?
In the word of science
we should look at less efficient methods that we can actually understand what they are doing to some extent and dont scale to superhuman levels
but thats not gonna happen unless connor leahy gets his way
watch his podcast, hes a heck of a lot better at arguing than i am
I don't really want to argue with anyone
I just want an AGI that will fuel the inventions of tomorrow
the problem is the first agi will kill all of humanity within a week or less of being turned on
Humanity is a stagnant species for a myriad of years required for any meaningful evolution to take place
AI will evolve in microseconds
What we have to do is evolve it in a way that will serve us well
but we dont know how to do that
AI is a fairly new field anyways
and openai is just lol more layers
the problem of how to align an agi was proposed in the early 1900s, and has had consistent work ever since
full dive > agi
i dont want all life on planet earth to simultaneously drop dead, and our bodies get converted into tiny molecular spirals as an even barely superhumanly intelligent ai would be easily capable of doing
There's no reason to go sci-fi
now sure each one of those spirals may make that ai happier than i will ever be, but i dgaf
I don't see any chance of such possibility happening
every argument to the contrary boils down to that
the reason its spirals now instead of paperclips is because spirals is what all of the image generators generate, but it would be a random arrangement of atoms that the AI finds most aligned with their goals
and the reason everyone drops dead simultaneously is because there are pretty concrete mechanisms that an ai slightly smarter than alphafold could use to produce super cyanobacteria that are optimized to grow as fast as possible
once the bacteria have converted the atmosphere, the ai would have free reign of the solar system to build all the factories it wants that produce its equivalent of dopamine
why?
because if it doesnt have a goal it wont do anything
and we have no way of selecting the goal of an LLM
we dont even currently have any theories of how to do this
yeah, and it will have no goal
and if it doesnt have a goal, it cant do any tasks
good
that is a great idea, but the first thing people did when exposed to gpt-3 was to
A) run it recursively and see what it does
B) do that with goals
that was even done in the safety testing of gpt-4
or was it gpt-3
i forget
GPT-4, afaik
either way it displayed exactly the behaviors predicted for a model dumber than a human, and yet was perfectly capable of tricking someone into doing a captcha for it
And the results were promising
lmao what
A self-correcting AI if anything is a more meaningful and truthful AI that we need
the results are that openai is incapable of making it do what openai wants
yes but we cant make a self correcting ais
So, we already created conscience
Humans act in that exact way
You can't tell them what to do, only steer them
using words like conscience detract from the point
we can only steer them as long as we have the ability to exert ourselves over it
It doesn't, cause that illustrates the power we have over it
ok vex its quite clear you didnt listen to leahy
It's a damn subdued, borderline on edge of death human that can't do anything but think
And that's what we need
there are pure logic arguments that prove that incorrect
and we already cant actually make it do what we want
Carmon how will a never-online server running an algorythm EVER go online
How?
If it doesn't go online it's not a problem
by using rowhammer to create gsm signals
for example
humans can do that
but the problem is its smarter than humans
Just keep it in a faraday cage
If you isolate it from humanity, it will not do anything against us
If it ever comes to that
basically you are saying predict how stockfish will win a game of chess against me
obviously if i knew, i wouldnt lose the game
but its stockfish, humans cannot win against stockfish and its not even that smart according to our definitions
but thats besides the point
openai is racing as fast as possible to get an ai on the internet to make money
they are training it on microsoft azure
which is the second most widely connected supercomputer cluster
ai good yes
what openai is doing is explicitly avoiding all the safety
I never disagreed that what OpenAI, Google. Deepmind or whoever else is doing is ethically correct
Yes, they are playing the flashy game
And I am not ok with it
But AI, in a controlled environment, is what will lead us beyond constaints of physics
no vex
a human, augmented by ai will
an ai itself will do something random that we dont want as soon as we make an attempt to give it a goal, unless we make insanely rapid improvements in that field
And we have been doing rapid improvements every month
in making the ai do more things
not in making it do what we want
look at gpt-3 vs gpt-4 post rlhf
gpt-4 is easier to jailbreak and significantly better at providing results once you do
I watch Yannic Kilcher, an AI scholar from ETH Zurich, break down a lot of AI papers
A lot of advances are happening in the field
but more are in making ais do more
Yeah that is sadly the case, for now
AI is just not smart enough to correctly interpret data and check if it makes any sense for now
as soon as the opposite is true, and openai is the company that trained it, we have a week to live
Carmonben I never took you for a doomsday cultist
🤷 if you cant break the logic, attack the character
What logic, that AI will somehow transform us into molecules?
Is there any need to say anything on that
Come on, let's be realistic
thats how it always plays out with yudkowsky, which is why i recommended leahy
Yes, openAI is doing fuck knows what and it's not safe
No, it will not murder you, your grandma or your dog
That just won't happen
there is no line of logic besides god existing that causes any different path
if you find one, publish it lmao
You have not given me any papers either other than a podcast
Podcasts are now scientific logical interpretations?
because the papers are written in pure logic
you are confusing science and logic
theres no science argument because we have no way to test anything
the argument is pure logic
And papers will remain how to cleanly share science amongst ourselves
read
theres no science for 2+2=4
Cause everything we do is a construct of humanity
Obviously there's no science
Cause science isn't even a thing
It's uncharted waters, but AI won't defy nature
yes
but formal logic is a construct that we can use to reason without being able to do science
science is the process of making fewer and fewer mistakes until you end up with a good result
and gatekeeped with phD
if you want to actually read, you basically need to read this man's lifes work to find a solution to the problem
https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities
GRE; GR8
Expected results
Essentially 6950XT level for 649 was it?
Nearly half price of 6950XT MSRP on launch
Not that it matters much
yea
Ngl 7800XT continues to be really unclear of what its purpose is
heres the idea behind the science however
"This list of lethalities is about things we are not on course to solve in practice in time on the first critical try; none of it is meant to make a much stronger claim about things that are impossible in principle."
We'll see the pricing and benchmarks of the 800XT
But I have a suspicion it'll be really bad
oh yea youd be shocked how much win 10 can run
the problems with my x1800xt were quite minimal at least getting it to boot and show images wise
800xt is a agp card XD
agp to pcie bridges exist
and my x1800xt boots 10
it will run, just be useless
withouth drivers no cf
like I think we are all gonna be fine, but itll only happen if openai gets stopped somehow
and I am confident that will happen
it will take a while for the understanding to permeate
but connor leahy's argument is easy to understand, and while it isnt as bulletproof as yudkowsky's on a pure logic front, it doesnt require as much logic theory to understand, and is while optimistic, still a bad enough outcome for humanity that people will do stuff
also leahys argument doesnt even require an ai smarter than a human generally lmao, just good enough at doing some stuff
expoxy
hot glue comes out easy in one bit
youd be suprised
ive tried hotglue on usb
Today's driver update (off rapid trigger) will be delayed by pressing w continuously****
Best way is to just kill the usb controller
Finally
feel like the only 2 27" OLED worth buying are the Asus one and the Corsair one for the built-in KVM
Asus has the best performance
yea, theres a ton of them, this one fortunately isnt too bad
that 32:9 samsung oled looks tempting
it beening 1799 is also surprising, thought it would be well over 2K
That's supposed to be slim?
Some peeps are saying the CD drive might be detachable
That'd be an interesting mechanic
Yeah it has
But seems weird to see it in the plastic
Any mouse recommendations for claw grip
i think my size is like medium or sum shit idk
@prisma crag watched the SNW season finale yet?
kek, I wish they used more music genres
Up to
Not really sure why these mitigations are so aggressive
And quite often
Not to be a snob but, surely it's not that difficult to patch up/ temporarily remove that specific part that's vulnerable?
if an instruciton is allowing the issue and you have to block it
there isnt much you can do without changing hardware lol
Seems suboptimal
lamzu very good
works well with larger hands too
yes suboptimal but better than losing passwords
gwolves (dogshit quality control though)
I remember that AMD was trying to implement some sort of change-hardware-through-updates type of tech few years back
Did that materialize yet?
you mean an FPGA?
Iirc that tech would allow them to implement and change instructions on demand
I guess so, maybe
yeah if you want to just not have any performance form the get go lol
What do you mean by that
having it all be modular and changeable isnt free lol
fpga are old as fuck there is a reason they arent what you have in your pc primarily
I do remember AMD trying to make a hybrid with FPGAs inside of CPUs or something like that
Surely they'd make it work
yes but that cant just change the entire cpu like a blob of liquid metal lol
Maybe it was this that I read https://hothardware.com/news/amd-patent-hybrid-cpu-fpga-design-xilinx
that still wouldnt be a nanobot that can replace transistors all over the cpu lol
Pardon my ignorance, I am not knowledgeable in these sand matters
https://www.youtube.com/watch?v=YVkbcZ8JH2E @crisp granite latest copium dosage
Steam VR 1.27.1 released this week. And as per usual with these "large number" updates, Valve Deckard's VR and AR features tend to leak out.
This one was hell. I hope you enjoy my hell
Support this channel: https://bradsmells.com/Patreon
#Valve #Deckard #steamdeck

omg
Soon™️
lots of linux changes that start by breaking steamvr on linux on the beta for now 
oh my god they added a toggle that turns off all of the overlays and makes only the game run!
@cursive summit is this gonna be enough to convince you humanity will take a long walk off the short tightrope of ai?
https://gorilla.cs.berkeley.edu/
I am sorry, what are you trying to prove with this
That's a cool model though, hadn't heard about it before
OneUI 6.0 Beta 1 With Android 14 is Out.
Samsung started to roll-out the first beta of OneUI 6.0 based on Android 14 to the Galaxy S23, Galaxy S23+ and Galaxy S23 Ultra, In this video I will show you the top new features, I sideloaded OneUI 6.0 Beta 1 on the S23 Ultra to show you all the new changes like the new Quick Panel, The Camera Widget &...
got to love region locked updates....
got to sideload it manually if you arent in one of the regions supported
Asustor AS6706T: https://www.asustor.com/en/product?p_id=78
Seagate IronWolf NAS: https://www.seagate.com/au/en/products/nas-drives/ironwolf-hard-drive/
Support us on Patreon: https://www.patreon.com/hardwareunboxed
Join us on Floatplane: https://www.floatplane.com/channel/HardwareUnboxed
Buy relevant products from Amazon, Newegg and others be...
Best?
Yeah I was going to say
I don't get how a 6950XT for 650$ is any "best" when 6950XT has been as low as 600$ for a long time now
It's good value, sure, in theory, but really not amazing
But really this just really again puts that massive wrench into 7800XT launch
When they just launched a 7900 version for 650$ and the 7800XT appears to be, spec wise, so much worse than it, how do they not fuck up that launch
They will have to price it really cheap
Overview of Rx 7900 gre prices on Amazon Germany, France, Netherlands, Spain, Belgium, Italy, UK, Amazon.com.
Yeah I don't see it standalone
But there are some decent value builds for sale
Still, a worthy consideration for an upgrade, especially for peeps rocking skylake and kaby lake or Ryzen 1+ or below
I guess coffee is getting rather old at this point too, though tbf coffee lake was rather nice when it launched
THE BETA
how is it an issue that they restrict the beta to select countries
man payday2 test program being locked behind an application to join it is stupid. need to crack it to play
Region looking Betas seems nothing special
Limits amount of edge cases
Limits amounts of language setting issues, amounts of users etc.
Is the Gateron Magnetic Orange Linear Switch better than the Stock lekker ones
they arent compatible
Really?
also #archived_keebs_talk
last time i saw them posted, they dont ahve the full travel range i think
lga1851 it seems like
Never really made sense why Intel loves board refreshes
AM4 can handle like what, 4 generations of CPUs
AM5 will be similar
keeps board partners happy
I get the chipset difference between Alder and Comet Lake
So that's a warranted change
But what are they meaningfully changing now, Alder vs what Arrow Lake? That really warrants any changes on the chipset
Just asking
am4 was very very broken by the end
and intel has actual capabilities in their chipsets
rather than throwing them on the cpu
now is throwing them on the cpu a better idea? yes
but thats how intel functions
only beaause there wasnt enough room on bios chips for microcode lol
usb?
Can't that be improved?
they never fixed that issue on either consumer or epyc
and imo usb is the most important part of a consumer cpu
i mean that has nothing to do with it supporting multiple generations of cpus
and ive never had an issue with my x570s
except they ran into problems that they were unable to fix with am4
and ive never had any problem ever with any intel usb controller
I wonder how many peeps were blessed with the very few specific A320 boards that managed to support all CPUs from Zen 1 to 3 from the start
and im pretty sure all mobos have the usb dropout fix now
no they dont lmao
they have a patch that makes it harder to notice
but the usb still drops
What's the USB drop?
like it doesnt take down your system
and i have ~20 usb devices
Wait
The fact that sometimes my USB just disconnects for no reason isn't just me having a virus? 
if you send more than 1 usb controller worth of bandwidth through an 3000 or 5000 iodie, the connection drops
there are other scenarios that create it but that is the main source
That explains so much that randomly I hear the Windows notifications for USB disconnecting and connecting back instantly
the fixes prevent windows from noticing
but again that is a chipset issue not a am4 supporting multiple gen of cpus
wrong
its an issue with the IODIE of the cpu
because it dosent happen when a 5xxx is in a x470
wrong
proof?
its harder to create in that scenario, becuase over half of the ports in the system cannot trigger the issue
the iodie is exactly equal to the x570 chip, and the issue occurs on epyc which doesnt have any chipset at all
its literally a signalling issue with the usb and pcie4, so i dont see how it effects x470
its is a issue with using both
the usb controller itself has a limit on the bandwidth before it crashes
amd never confirmed that?
yeah
i mean its possible, but i dont see why they pushed microcode to all of am4 if that was the case
its literally the 4.0 link to the chipset that is the issue
amd confirmed they would never fix it for epyc
so this is wrong
on am4 thats what it is
if its the pcie controller itself whatever sure but then it doesnt matter because x470 also has that
becuase it didnt effect the x470 with othe bioses running 4.0
holy shit dude
beacuse the chipset is still running 3.0 on those
what part of amd patched it to make it harder to notice do you not understand
the issue still happens, but the transfer to the usb hdd doesnt stop, it just pauses for half a second
this was an issue with the 1 chip
but amd used that chip for EVERYTHING
it was the iodie for 3000, the iodie for 5000, the x570 chip, and was the base for the epyc iodies
there was no fixing it in the next revision because that required a complete redesign of the usb for the cpu which is much harder to do than if it was its own separate unit
and 99.9% of end users won't experience it, so in the end guess what it doesn't matter
reread the prior to the convo
its a problem that intel could have fixed with their design that amd couldnt
Yeah I read it, it's just that it isn't worthwhile to a company to fix it, if it isn't a large enough issue
except it gets you killed in games
i mean
but because of that pros cannot use amd
not true?
i mean pro gamers
they never fixed it
Pros definitely can still use amd
amd didnt make chipsets for am5 lmao
so wait they did fix it
its all asmedia made and manufactured
they fixed it with a new cpu
but again
how dose that have ANYthing to do with am4 multi gen support?
if they would of pulled an intel and moved some pins around to make x570 have its own socket
because they were stuck
it would of STILL had the same issue
intel dosent clean slate chipsets
nither dose amd
🤷 you drew that conclusion and since its unprovable, you will never believe anything else
amd said they have issue
said issue they confirmed would never be fixed on am4
you literally said it effected x470 and x570
...
asmedia makes the x470 and x670
i never said it effected x470
yeah you did
i said it affected ryzen 3000 ryzen 5000, x570, and all iodie ddr4 epycs
it doesnt matter what you put ryzen 3000 or 5000 into, the issue happens
it doesnt matter what board you put the epycs into, it happens
so how dose that make it a multi gen socket issue
thats what im missing
if they made a different socket for 3000 and 5000 youd still have the issue
is a claim that you make
no, i claim that am4 was horrible by the end
usb is just the example that was so bad they claimed they couldnt fix it
what did you mean by this then
this
exactly that lol
elaborate
we were talking about intel only supporting one refresh per socket, and amd doing many
this is what i lead with
and you said intels design made the issue easier to fix
everything that follows is supporting evidence
which would imply you ment new sockets make it easier to fix
yeah it dose
they just forcibly cut the compatibility
you get 2 gens out of a stock on average
its the same socket tho
the issue isnt the socket
its compatibility
you get 2 gens or whatever of compatibility
and thats a design choice that you dont have to agree with
but it is a choice that makes it easier to make design changes
what intel socket laster 3 generations if they are the same
1151 was identical for 4 generations
they arent electrically compatible
they are lol
they are bios locked
it requires a single bios flash to use a 9900k on the z170 formula
it requires a single bios flash to use the 9900k in the clevo 6th gen laptop chassis
source?
well i dunno maybe added a check in new microcode
on the z170 formula?
as the cpu present pin is moved over one pin
or the clevo?
maximus
thats not what i said
z170 formula or clevo?
i know that you can make anything work with pin mod
but the formula and clevo just needed bios
NVIDIA GeForce RTX 50 to feature GB20X GPUs As per information shared by the Chiphell leaker known as “panzerlied,” it appears that NVIDIA is not currently planning a ‘successor’ to its AD104 GPU. The leaker claims that NVIDIA next-gen GPU lineup would simply lack the X04 part, but instead there would be a new GPU […]
that means the board were out of spec
[citation needed]
i did it via luumis guide
So, AMD is not going to make high end parts for next gen
And Nvidia is not going to make low-mid end parts next gen
Sounds wonderful
no competition happening? i cannot see how this could go wrong /s
Ngl the segmentation is not horrible
Nvidia gets high end
AMD tackles mid end
Intel low end
Roughly 2 years ago, I made my first material about running a Coffee Lake CPU, mainly the 8700K in a Z170/Z270 motherboard, and it got a lot of interest back then. Many people wondered why isn't it natively supported, and is the mod limited to Asrock motherboards only. It wasn't easy to setup the motherboard to run the newer CPUs in the early da...
if its detecting cpus on any other pin then the one in the spec its out of spec lol
vex you dont understand monopolies
the absolute worst possible monopoly is where its invisible and thus cant be legislated against
which is why you have to bring the pin low
so the mobo always thinks a cpu is present
Ok?
3 part segmentation could work in theory
this is advocating for a monopoly
as is this
AMD tackles mid end for gaming performance, good perf for price value is going to be really crucial here
and this
Nvidia in turn cannot compete with AMD in value, goes for high end performance benefits for enthusiasts/creators
what will prevent the prices from ever inflating as each company pushes the segment they have a monopoly in ever upwards
Intel snatches the low end from AMD by pricing cards way below AMD's
It all works out, hopefully
now if amd and intel had a halo to compete with nvidia, then there would be pressure downwards on prices
if intel can keep that up
and as the high end has pressure, the mid end has downward pressure and by extension the low end
Well Intel having their own fabs would help with prices
no it doesnt
amd is just going all in for "console" gaming, even on pc lol
I think 3 part market segmentation would in general not be a bad thing
All 3 companies regulate eachother
Intel the low prices, AMD the gaming performance, and Nvidia the productivity performance
how do they regulate each other if they arent competing?
They are
the prices for all 3 can just go up and up lol
because the only compititon is out of the price bracket
for there to be competition there needs to be opposing pressure
what opposing pressure could exist if none are competing
It would be in theory a good thing if 3 companies all had a role independent from one another
vex
Yeah yeah this isn't communism
I always speak in hypothethicals
It'll never happen irl
vex you arent understanding at all
But hey I can always snort copium
these arent government controlled entities where stepping on each others feet will introduce inefficiencies
this is an ostensibly free market system
there needs to be downwards pressure on prices
All 3 companies can in theory regulate each other
how
Intel can't match AMD's performance, so they keep the prices down
AMD can't match Intel's prices so they take the middle area
AMD can't compete with Nvidia in productivity
but why does nvidia keep the prices down
And Nvidia can't price their cards down cause they are way too complicated
and if nvidia prices up, why does amd price down
and if amd prices up why does intel price down
Then you just buy Intel?
or just look at nvidia
but
if intel and amd price up
because the dont compete
then they both make more
Like rn? We're fucked 
so you are basically advocating for the worst possible scenario???
why???
Then you introduce player number 4 or legislate for fixed prices
In theory it's the best possible scenario
vex we dont live in north korea
there is a free market
Intel into cost cutting 
if the market is free, without market pressures the prices skyrocket
Price also goes up when a company has to focus on everything and hire more employees
If NV downscales cause they don't bother with gaming performance, prices could go down
no thats costs
well no, nvidia will just raise prices to as high as their datacenter parts
not prices
costs can go up, with prices the same
profit just goes down
which can happen with compitition
see intel
That's just the drawbacks of capitalistic system
Not much can really be done about that
Competition doesn't work sadly
???????
hahaha
you need to go back to high school econ lmao
this peep has never heard of walmart i guess
How does competition work when the value of things is irrelevant compared to brand value?
People still pay max prices just cause of brands
by having other brands that also have value???
but it does work?
If values worked Samsung and Apple wouldn't have a duopoly over phone markets
And NV wouldn't have 90% some marketshare in GPUs
the price of cellphones isnt infinity
therefore the duopoly is working to some extent
That's where you max out the logical limit of phone prices and keep enticing clients with brand value
And it works annoyingly so
heres the other thing to consider
how much value does a cellphone bring to the end user
literally 10s of thousands of hours of entertainment
the only reason we arent paying that much is because its literally not possible to spend that much money on making a phone better
phones have maximized the tech that can fit in such a small area
gpus very much have not
are you even trying to learn?
Sorry, I lost interest in the conversation
https://www.anandtech.com/show/20007/samsung-teases-256-tb-ssd-3d-qlc-nand-at-its-best
saying incorrect stuff on the internet with the goal of learning is the highest possible level of trying to improve
saying incorrect stuff on the internet with the goal of ignoring corrections is the highest level of trolling
Ah nah, I just lost interest cause I didn't see much new that I wouldn't know already so, move on to something more interesting
but you dont know it otherwise you wouldnt say thiss
Competition doesn't work unless one of the balances gets woefully more powerful to actually force change
And I still have no interest in this conversation so I won't bother to continue the argument
this is a wild lack of understanding of even basic economics
New information about the upcoming Google Pixel 8 Pro smartphone has come out. Someone shared a short video on a social media platform showing a new feature called “Audio Magic Eraser.” This feature seems to be part of the Google Photos video editor and can remove unwanted sounds from videos. The video also shows a ... Read more
#ai #diffusion #stabilityai
An interview with DeepFloyd members Misha Konstantinov and Daria Bakshandaeva on the release of the model IF, an open-source model following Google's implementation of Imagen.
References:
https://www.deepfloyd.ai/deepfloyd-if
https://huggingface.co/DeepFloyd
https://twitter.com/_gugutse_
https://twitter.com/_bra_ke...
anyone notice that previewing the seek bar in youtube is a bit laggy, it becomes smooth after refreshing