GR Research Speaker Upgrade Review (Sierra-2EX V2) [Video]

[G|force]

Bad comparison. Wikipedia is being constantly peer-reviewed by the public, which is why (as studies found) it isn't any less accurate than the average encyclopedia.

None of this happens with AI. You have to check it every time anew by yourself.

The co-founder of Wikipedia Larry Sanger says it has become far from accurate in some subjects, and is a captured entity. If you watch his interviews you can see why.

 

[Karmacoma]

The co-founder of Wikipedia Larry Sanger says it has become far from accurate in some subjects, and is a captured entity. If you watch his interviews you can see why.

Wikipedia is a battlefield of the information war, when treating of topics even remotely political.

 

[VQR]

No need for you to stand corrected. Almost everything on Wikipedia is 'not' peer-reviewed. Your old grade school teachers take is 100% correct.
Much depends upon the topic regards the validity of the information presented there.

Moreso I meant in terms of comparison. Unlike Wikipedia, LLM's are a black box where you can't readily know or check the veracity of what they were trained.

At least with Wikipedia you can see what the editor and author cited, then corroborate and even look through the changes made. And indeed, using Wikipedia as a primary source for serious information or controversies is not good practice.

Other than what sites it summarizes from a web search, LLM's don't have some form of transparent review why they output what they do.

 

[Karmacoma]

Moreso I meant in terms of comparison. Unlike Wikipedia, LLM's are a black box where you can't readily know or check the veracity of what they were trained.

At least with Wikipedia you can see what the editor and author cited, then corroborate and even look through the changes made. And indeed, using Wikipedia as a primary source for serious information or controversies is not good practice.

Other than what sites it summarizes from a web search, LLM's don't have some form of transparent review why they output what they do.

The problem is if LLM's are the ones modifying wikipedia. I've seen an LLM creating a whole paragraph that didn't exist in a well known book just to satisfy the request, they don't know what is true or not and can easily create their own sources if they're given a narrative. We're just at the beginning of the internet turning into a giant slop. Hopefully we'll get some counter-fires.

 

[RexrothPigeon]

The co-founder of Wikipedia Larry Sanger says it has become far from accurate in some subjects, and is a captured entity. If you watch his interviews you can see why.

This guy's a true loony toons who lucked into some involvement in early wikipedia but everything else he's touched has failed. he's also an antivaxxer fundamentalist paranoiac. I say this not to convince you, Gjforce (I'm sure you and I will never agree on anything, since you believe Sanger's firehose of bullshit), but for other readers who may not know about this clown.

 

[G|force]

Yeah very early. Lucked into? OK.
Sanger co-founded it with Jimmy Wales.
He coined the name.
Please stick to alternative facts on ASR.

 

[RexrothPigeon]

Yeah very early. Lucked into? OK.
Sanger co-founded it with Jimmy Wales.
He coined the name.
Please stick to alternative facts on ASR.

He did, yeah. I know because I was involved in Nupedia, the project that preceded and eventually folded into Wikipedia, and Sanger was an incredibly ineffectual and disorganized "editor-in-chief".
These are facts I know firsthand.

 

[G|force]

Fair enough, I withdraw. No hard feelings on my end.

 

[RexrothPigeon]

More broadly, while i understand skepticism about wikipedia, I think it's a bit misplaced, or at least a bit lazy. The best thing about wikipedia for me is that it's an incredibly deep resource to find the primary sources. Citing wikipedia articles is stupid and teachers/publishers are right to disallow it, but if you are doing research, wikipedia has links to the real sources, the real conversations, and it's immensely valuable to have all that information in such an accessible format.

 

[G|force]

Depending on the subject matter, I believe it is an excellent resource. There are some things known today as fact from evolutionary biologists that were not known during Project Warp Speed.

Not only is that data not reflected, it is outright rejected. I'm not emotional about that, but I am skeptical indeed.

 

[BobbyTimmons]

More broadly, while i understand skepticism about wikipedia, I think it's a bit misplaced, or at least a bit lazy. The best thing about wikipedia for me is that it's an incredibly deep resource to find the primary sources. Citing wikipedia articles is stupid and teachers/publishers are right to disallow it, but if you are doing research, wikipedia has links to the real sources, the real conversations, and it's immensely valuable to have all that information in such an accessible format.

Wikipedia is terribly unreliable and biased on any controversial topic. But it can still be a useful website in its way.

 

[Alan O]

The problem is if LLM's are the ones modifying wikipedia.

There's a big fight against people using LLMs to write Wikipedia content at the moment, but I'm not aware of any LLMs being able to write it directly (I used to be an admin there... but I'm OK now )

Wikipedia is terribly unreliable and biased on any controversial topic. But it can still be a useful website in its way.

Exactly. It says itself that it's not a reliable source. But it seems pretty good at general topics, and for anything unsure/controversial you really need to check the sources. That's how I use it - as a starting point from which I investigate sources (pretty much the way I use LLMs too, in fact).

 

[Randolf]

Since Danny continues to publish false claims on the effect of time gating and smooting on the measurement resultion and quite a lot people replied to my comment in his video, I decided to become creative and cretaed my very firts short Youtube video on this topic:

The goal is to fight against Danny`s disinformation on this topic in a format many people with and without deep knowledge on this topic can understand. I hope it makes the world a tiny little bit better and less disinformed. Sorry to bother you with advertisment of my lastet block buster

I polished up the text description of my little youtube video quite a bit, with instructions how to reproduce it, links to useful videos on FFT, Klippel NFS, how Stereophile, Erin, Amir, KEF and Dynaudio measure and numbers of typically applied gating and smoothing degrees.

 

[RickS]

I polished up the text description of my little youtube video quite a bit, with instructions how to reproduce it, links to useful videos on FFT, Klippel NFS, how Stereophile, Erin, Amir, KEF and Dynaudio measure and numbers of typically applied gating and smoothing degrees.

How about Danny’s combo of gating, smoothing and scaling vs. Klippel?

I may have posted here but not on YT, but probably also should show comparison between standard 50 dB scaling and Danny’s use of 25 dB.

 

[dfuller]

I polished up the text description of my little youtube video quite a bit, with instructions how to reproduce it, links to useful videos on FFT, Klippel NFS, how Stereophile, Erin, Amir, KEF and Dynaudio measure and numbers of typically applied gating and smoothing degrees.

I know this is a rather big ask, but I'd rather hear a person than an AI generated voice.

 

[PristineSound]

I know this is a rather big ask, but I'd rather hear a person than an AI generated voice.

Com'on Bruh, really?

You see Mr Randolf, the moment you do something for people, you shall receive punishment for doing a good deed by more requests.

 

[a4eaudio]

I know this is a rather big ask, but I'd rather hear a person than an AI generated voice.

He just needs to really take advantage of AI... And have Samuel L. Jackson or James Earl Jones narrate the video!!

 

[BlackTalon]

Perfect! "Snakes in this motherf**ing GR Research!"

 

[Randolf]

I know this is a rather big ask, but I'd rather hear a person than an AI generated voice.

I could understand that, maybe at a later point of my very new YouTube career I may do that . But currently even at the age of almost 61 and after 30 years working in the Software industry I think my English is not polished enough with a typical german accent. I just started my early retirement phase and hopefully can further improve my English during traveling in the next years. Meanwhile I will stick to the faceless video format with AI text to speech. I personally also hate most AI generated video content, but here I just used text to speech AI, the content is 100% human intelligence. BTW the voice you hear is Mr. Mark Ellis from spektor.

 

[Randolf]

How about Danny’s combo of gating, smoothing and scaling vs. Klippel?

I may have posted here but not on YT, but probably also should show comparison between standard 50 dB scaling and Danny’s use of 25 dB.

Ok, you asked for it so here it is:

Still a faceless video but without any speech at all. I think it can stand for itself. Funny detail Danny suggested to reduce the tweeter resistor of the KEF R3 Meta a bit, while Erin said it can be a little bit to hot on the top end and recommended to toe it out 10 degree (he even did measurements with 2 different reference axis due to that). Shockingly GR Research sells an upgrade kit for 400-500$ (https://gr-research.com/product/kef-r3-meta/) with the usual parts quality and tube connector story.