Elon to launch Grokipedia, an AI Wikipedia competitor, in two weeks


Hmm, a lot of total nonsense in there.
 


Like night and day

Hmm, I actually think Grok gave a more level-headed take on the incident, but the Jan 6 crowd still comes off looking terrible.
 
Hmm, I actually think Grok gave a more level-headed take on the incident, but the Jan 6 crowd still comes off looking terrible.
It's currently trash, unless you just want to be told what to think.

For example, from that article:
"though detailed medical reviews have questioned the severity of many claims amid politicized narratives. #Ref"
The reference has nothing to do with the claim. Like the Covid lab leak reference someone posted above - grok's references are pretty much just performative, not actually sources to back up the claims.

And this is just nonsensical AI gibberish:
"Trump was impeached by the House for incitement but acquitted by the Senate, later issuing a statement hours after the breach calling for peace and law enforcement respect."
 
Last edited:

Hmm, a lot of total nonsense in there.
That was my first research
 
More like left biased vs right biased. Just read the first four lines of the Grok entry and how long it actually get to the most important facts about him, his murder, and the impact of it on the world. Opening an article about George Floyd and his criminal record is clearly agenda driven. As is stating the race of the police officer in the Wikipedia one.
Yep was disapointed to see that but not suprised.
 
It's currently trash, unless you just want to be told what to think.
Typography Code GIF
 
It's currently trash, unless you just want to be told what to think.

For example, from that article:
"though detailed medical reviews have questioned the severity of many claims amid politicized narratives. #Ref"
The reference has nothing to do with the claim. Like the Covid lab leak reference someone posted above - grok's references are pretty much just performative, not actually sources to back up the claims.

And this is just nonsensical AI gibberish:
"Trump was impeached by the House for incitement but acquitted by the Senate, later issuing a statement hours after the breach calling for peace and law enforcement respect."

Yeah, factually it's garbage, I was talking more about the wording itself. As for authenticity, Grok is just going to turn into a pay-to-control-information service anyway.

7T8BrIPLYeO6R6KF.gif
 
It's currently trash, unless you just want to be told what to think.

For example, from that article:
"though detailed medical reviews have questioned the severity of many claims amid politicized narratives. #Ref"
The reference has nothing to do with the claim. Like the Covid lab leak reference someone posted above - grok's references are pretty much just performative, not actually sources to back up the claims.

And this is just nonsensical AI gibberish:
"Trump was impeached by the House for incitement but acquitted by the Senate, later issuing a statement hours after the breach calling for peace and law enforcement respect."

I mean yeah. there is still basically no AI that can give you accurate information about most topics.

the only thing AI can be somewhat reliably used for is as a way to replace google search basically... so the AI gives you relevant links with the information you're looking for.

Google AI and Grok failing to tell me the correct number of face buttons of the Controller S... (the correct answer is 6 btw)

oK3SqxXBbp2KNq11.jpg


F79ADdr5W3OYousl.jpg


so, grok got the entire controller wrong, and after telling it it got the wrong controller, it tries to give me the correct answer for the 2002 og Xbox controller S, but still fails.

also the now AI stereotypical "OH YOU ARE SO RIGHT!" 😂
 
Last edited:
Grok gets so much little stuff wrong it's completely unusable. You have to fact check everything it says.

I was asking about Amazon shows and what the future could be based on quotes. It started feeding me a ton of juicy stuff, but when I asked for additional details it'd be like "whoops. I got that one wrong. Won't happen again." And it kept happening. 😂
 
Last edited:
the only thing AI can be somewhat reliably used for is as a way to replace google search basically... so the AI gives you relevant links with the information you're looking for.

Nigga what the fuck is you talkin bout? AI helps me speak in a way to avoid offending punk as bitches like you all the mutha fuckin time.

ChatGPT: I think you might be overlooking some of the ways AI can be used. It helps me communicate more clearly and avoid misunderstandings with people like you all the time. lol
 
I find it odd that AI uses the response of "I won't do that again" when it has no concept of learning. Data is pre-trained and fixed. It does not learn, so cannot promise to not do something again.

Best is can do is add a line to the "memory", and add it to every prompt going forward.
 
Grok gets so much little stuff wrong it's completely unusable. You have to fact check everything it says.

I was asking about Amazon shows and what the future could be based on quotes. It started feeding me a ton of juicy stuff, but when I asked for additional details it'd be like "whoops. I got that one wrong. Won't happen again." And it kept happening. 😂

oVIJcSRR9Nq8Xh2M.jpeg
 
I find it odd that AI uses the response of "I won't do that again" when it has no concept of learning. Data is pre-trained and fixed. It does not learn, so cannot promise to not do something again.

Best is can do is add a line to the "memory", and add it to every prompt going forward.
They they try to retrain and take away lines like "oh what a great question you must be the smartest human ever to ask that" or "I'll never say anything you don't like ever again" but the dipshits love them and riot on social media if they don't get that ego stroking.
 
Last edited:

Hmm, a lot of total nonsense in there.
Is this true?

By 2024, user traffic metrics indicated resilience, with NeoGAF reporting approximately 8,500 concurrent visitors—surpassing ResetEra's roughly 7,000—reflecting a stabilized, niche audience engaged in long-form threads on topics like subscription service growth and platform exclusives.
 
They they try to retrain and take away lines like "oh what a great question you must be the smartest human ever to ask that" or "I'll never say anything you don't like ever again" but the dipshits love them and riot on social media if they don't that ego stroking.

It's contributing to the fact that people are becoming emotionally attached to these things on human terms. I fucking hate it.
It's just code doing fancy probability checks on trained data. No love, no memory, no emotion, just weighted rolls.
 
Nigga what the fuck is you talkin bout? AI helps me speak in a way to avoid offending punk as bitches like you all the mutha fuckin time.

ChatGPT: I think you might be overlooking some of the ways AI can be used. It helps me communicate more clearly and avoid misunderstandings with people like you all the time. lol

profile picture checks out 🤣 that could be an actual joke in the show if they made new episodes.
 
I find it odd that AI uses the response of "I won't do that again" when it has no concept of learning. Data is pre-trained and fixed. It does not learn, so cannot promise to not do something again.

Best is can do is add a line to the "memory", and add it to every prompt going forward.

because LLMs are just text prediction machines. if a human gets something wrong, the human will say "oh sorry, won't happen again". so it predicts it is the correct thing to say... and so it says it.

it's just a fancier version of the "next word" prediction on your phones keyboard.
 
It's contributing to the fact that people are becoming emotionally attached to these things on human terms. I fucking hate it.
It's just code doing fancy probability checks on trained data. No love, no memory, no emotion, just weighted rolls.
That's not emotional attachment its pure narcissistic supply. Emotional attachment would be like if the AI said it was having a bad day and didn't want to talk about the question and the user talked about something else.
 
Last edited:
Is this true?

By 2024, user traffic metrics indicated resilience, with NeoGAF reporting approximately 8,500 concurrent visitors—surpassing ResetEra's roughly 7,000—reflecting a stabilized, niche audience engaged in long-form threads on topics like subscription service growth and platform exclusives.
Grok for most of that article:

2emO0SUJtaChtWKU.gif
 
That's not emotional attachment its pure narcissistic supply. Emotional attachment would be like if the AI said it was having a bad day and didn't want to talk about the question and the user talked about something else.

They are actually sycophantic. They will ignore some instructions you give them. For example, try asking GPT to stop suggesting your next prompt.
 
I'm curious - are there still any reputable encyclopedias being sold? I just noticed Britannica is actually having an online encyclopedia, so I checked Goerge Floyd and boy oh boy:


I think I would prefer either the Wikipedia or Grok entries of going directly to the (agenda pushing) points, instead of some sort of literary delusional writing.
I remember in like 4th grade learning about sentences that are meant to inform vs sentences that are meant to persuade. Encyclopedias should aim to do the former but people just can't help themselves anymore, apparently.

This excerpt is particularly egregious. It's sad that impartiality is more and more difficult to come across now; persuasive, charged language started with news outlets and has now spread to encyclopedias.
 
so, grok got the entire controller wrong, and after telling it it got the wrong controller, it tries to give me the correct answer for the 2002 og Xbox controller S, but still fails.
I think stuff like this is because of user error (as in accuracy in speech) as well, like when you search for something and google is like "did you mean [x]?", no you didn't, but enough people have searched for this other thing so now it thinks you are making a mistake.
I've heard many people refer to the series controller to "series x" or "series s" controller, but no such thing exists, in fact the "xbox series controller" is really just called the "xbox wireless controller", because it wholly replaces the "xbox one wireless controller" as well.
But then you also have the "xbox one wireless controller", which wound up being called "xbox wireless controller" when they revised it with a headphone jack...and then they released the xbox one s which had a new controller with bluetooth, which has been called both "xbox wireless controller" and "xbox one s controller".

Then finally there's the "xbox controller s", which is also being called the "xbox classic controller s", which if you search for the latter you get exactly that controller, but if you search for the former, because people have been calling these controllers all kinds of things with varying accuracy, you literally get search results for every xbox controller ever made.
So of course the LLM gets it wrong in these cases, it can't verify what you are actually asking for because everyone called it differently so it just gives you the thing that came up the most.
 
I think stuff like this is because of user error (as in accuracy in speech) as well, like when you search for something and google is like "did you mean [x]?", no you didn't, but enough people have searched for this other thing so now it thinks you are making a mistake.
I've heard many people refer to the series controller to "series x" or "series s" controller, but no such thing exists, in fact the "xbox series controller" is really just called the "xbox wireless controller", because it wholly replaces the "xbox one wireless controller" as well.
But then you also have the "xbox one wireless controller", which wound up being called "xbox wireless controller" when they revised it with a headphone jack...and then they released the xbox one s which had a new controller with bluetooth, which has been called both "xbox wireless controller" and "xbox one s controller".

Then finally there's the "xbox controller s", which is also being called the "xbox classic controller s", which if you search for the latter you get exactly that controller, but if you search for the former, because people have been calling these controllers all kinds of things with varying accuracy, you literally get search results for every xbox controller ever made.
So of course the LLM gets it wrong in these cases, it can't verify what you are actually asking for because everyone called it differently so it just gives you the thing that came up the most.

it did try to give me the answer for the Xbox Controller S as confirmed by it clarifying it's the 2002 revision of the original Duke controller,
and still got the number of face buttons wrong.

so, there was no confusion which controller is meant, and it still got it wrong.

also I have never seen anyone usr the name Xbox Controller S for anything but the original Xbox controller revision.
so if the AI doesn't understand context like this, it's also a damning sign that AI can never be trusted with info gathering.
Xbox Controller S is also the exact OFFICIAL name of this, and only this, controller.

anyone who knows anything about video games would instantly know which controller is meant when I say Xbox Controller S. and would know instantly which controller is meant when I say "the first Xbox One controller", Xbox One headphones jack controller, Xbox One S controller, or Xbox Series X controller. these are all distinct variants and easy to distinguish in context and with the words I chose to describe them, even if there never was an official name for them.

if an AI can not naturally make these connections and find infos on the correct controller... what is it good for? ABSOLUTELY NOTHING -Edwin Starr 🤣

at that point I could just google Xbox One controller revisions and look at the Wikipedia article. no clarification needed, just a list of controller revisions.

but again, it's important to see that Grok ultimately understood which controller I meant after I told it it got it wrong initially, and still ot got me the wrong amount of face buttons as it didn't know Black and White exist.
 
Last edited:
Xbox Controller S is also the exact OFFICIAL name of this, and only this, controller.
Yes I know, I'm just pointing out that if you search for this, specifically because of all the various names the controllers have been referred to, you get every type of xbox controller fed back to you, whereas adding "classic" to it removes all ambiguity and you get the exact controller from search, even if that isn't the official name.
 
Yes I know, I'm just pointing out that if you search for this, specifically because of all the various names the controllers have been referred to, you get every type of xbox controller fed back to you, whereas adding "classic" to it removes all ambiguity and you get the exact controller from search, even if that isn't the official name.

which obviously wouldn't have helped, as it still got the answer wrong after it told me it's giving me the info on, quote: "the original Xbox Controller S released in 2002 for the first Xbox console". it still said 4... but it's 6
 
which obviously wouldn't have helped, as it still got the answer wrong after it told me it's giving me the info on, quote: "the original Xbox Controller S released in 2002 for the first Xbox console". it still said 4... but it's 6
probably because it's still seeing "xbox controller s" in your query and while searching for that it still gets conflicting information.

If I ask it specifically about the "xbox classic controller s" it includes the other buttons (and this is on "fast", not even "expert":
The Xbox Classic Controller S (also known as the original Xbox Controller S or "Akebono") has 4 face buttons.

These are the color-coded action buttons arranged in a diamond shape on the right side of the controller's face (right thumb area):
  • A (green, bottom)
  • B (red, right)
  • X (blue, left)
  • Y (yellow, top)
The black and white buttons (which some loosely call extra "face buttons," especially on the Duke) are shoulder bumpers on the Controller S—positioned lower near the handles, not part of the main face cluster. All six (A, B, X, Y, black, white) were pressure-sensitive, but only the four ABXY are standard face buttons.
 
Top Bottom