Question Here is the essence of the stupidity of AI

Toggle sidebar Toggle sidebar
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,611
14,591
136
I asked chatgpt a question. How stupid is this response. And people think AI is something they want ?????? It does not know what Zen 5 is ??? And it conjectures "what if" and does not give any useful response. So first It does not exist, then a what if...Google provides more useful information, the same as the last 20 years or so.

1712624396913.png
 
DAPUNISHER

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
28,610
21,004
146
Here's Copilot's answer to the question -

Certainly! AMD Zen 5 is the next-generation Ryzen CPU architecture, and it’s slated for a launch sometime in 2024. Here’s what we know about Zen 5 so far:

  1. Release Date and Availability:
    • AMD confirmed that Zen 5 is on track for a release in the second half of 2024.
    • Considering the launch of Zen 4 was in September 2022, we would expect to see Zen 5 desktop processors debut around the same timeframe, possibly with an announcement in the summer at Computex.
    • Zen 5 CPUs with 3D V-Cache (a mid-generation refresh) are expected to show up in Spring 2025, following a similar pattern as Zen 4’s 3D V-Cache chips.
  2. Specs and Improvements:
    • Zen 5 is rumored to have a maximum of 16 cores, similar to Zen 4.
    • While clock speed improvements may be limited, architectural improvements are expected to boost performance significantly.
    • Leaks suggest up to 30% IPC improvements, making it the largest inter-generational IPC increase since the original Zen generation in 2017.
    • Zen 5 will be built on TSMC’s 3nm process and may feature an older process for the I/O die.
    • Enhanced AI accelerators will improve AI-driven tasks locally.
    • Zen 5 processors will feature the new RDNA 3.5 graphics architecture, potentially offering better onboard graphics than Ryzen 7000 CPUs.
In summary, Zen 5 promises exciting advancements, especially in IPC performance. Stay tuned for official announcements
 
  • Like
Reactions: SolidQ, Tlh97 and Elfear
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,611
14,591
136
DAPUNISHER said:
Here's Copilot's answer to the question -

Certainly! AMD Zen 5 is the next-generation Ryzen CPU architecture, and it’s slated for a launch sometime in 2024. Here’s what we know about Zen 5 so far:

  1. Release Date and Availability:
    • AMD confirmed that Zen 5 is on track for a release in the second half of 2024.
    • Considering the launch of Zen 4 was in September 2022, we would expect to see Zen 5 desktop processors debut around the same timeframe, possibly with an announcement in the summer at Computex.
    • Zen 5 CPUs with 3D V-Cache (a mid-generation refresh) are expected to show up in Spring 2025, following a similar pattern as Zen 4’s 3D V-Cache chips.
  2. Specs and Improvements:
    • Zen 5 is rumored to have a maximum of 16 cores, similar to Zen 4.
    • While clock speed improvements may be limited, architectural improvements are expected to boost performance significantly.
    • Leaks suggest up to 30% IPC improvements, making it the largest inter-generational IPC increase since the original Zen generation in 2017.
    • Zen 5 will be built on TSMC’s 3nm process and may feature an older process for the I/O die.
    • Enhanced AI accelerators will improve AI-driven tasks locally.
    • Zen 5 processors will feature the new RDNA 3.5 graphics architecture, potentially offering better onboard graphics than Ryzen 7000 CPUs.
In summary, Zen 5 promises exciting advancements, especially in IPC performance. Stay tuned for official announcements
Click to expand...
So, what you are saying is thats its all about the implementation of AI. I assume that copilot is all AI with no intervening human intervention ? If so, how is chatgpt so stupid ?
 
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,611
14,591
136
And I asked copilot the same question and got even a better response

Copilot
The Zen 5 architecture, AMD’s next-generation Ryzen CPU, is eagerly anticipated by enthusiasts. Here’s what we know so far:
  1. Release Date and Availability:
  2. Specs:
    • While raw specifications for Zen 5 haven’t been officially released, there are some confirmed details and rumors:
      • Cores: Zen 5 is expected to have a maximum of 16 cores, similar to Zen 4.
      • Clock Speeds: Rumors suggest limited or non-existent clock speed improvements, but architectural enhancements may compensate.
      • IPC Improvements: Strong rumors point to up to 30% improvements in instructions per clock (IPC), a significant leap compared to previous generations.
      • Memory Support: Zen 5 may offer native support for slightly faster memory.
      • CPU Cache: Expect increased cache across L1, L2, and L3 levels.
      • Process Technology: Zen 5 will be built on TSMC’s 3nm process.
      • AI Accelerators: Enhanced AI accelerators for improved AI-driven tasks.
      • Graphics: Zen 5 processors will feature the new RDNA 3.5 graphics architecture, enhancing onboard graphics capabilities2.
In summary, Zen 5 promises exciting performance improvements, especially in IPC, making it a highly anticipated addition to AMD’s lineup! 🚀🔥

From Microsoft Start Partners​

Learn more​

2digitaltrends.com3tweaktown.com5makeuseof.com6en.wikipedia.org
 
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,611
14,591
136
poke01 said:
It’s all in implementation. Copilot is good.
Click to expand...
Well, if copilot is 100% AI and no human intervention, then I agree. It just seems there is such a disparity that it seems copilot has human intervention.
 
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,611
14,591
136
OK, so I found a question, that while answered OK< was terribly outdated. 14nm ????

You
why is intel so stupid


Copilot
Intel, a prominent semiconductor company, has faced criticism and challenges in recent years. Let’s explore some reasons why people might express frustration or disappointment with Intel:
  1. Outdated Technology:
  2. Price-Performance Ratio:
  3. Energy Consumption:
  4. Socket Compatibility:
  5. Lack of PCIe 4.0 Support:
  6. Market Dynamics:
In summary, it’s not that Intel is inherently “stupid,” but rather a combination of technological choices, pricing, and market dynamics that have led to criticism. As consumers, we benefit from healthy competition between Intel and AMD, driving innovation and better products for all of us.
 
mikeymikec

mikeymikec

Lifer
May 19, 2011
17,767
9,724
136
Markfw said:
I asked chatgpt a question. How stupid is this response. And people think AI is something they want ?????? It does not know what Zen 5 is ??? And it conjectures "what if" and does not give any useful response. So first It does not exist, then a what if...Google provides more useful information, the same as the last 20 years or so.

View attachment 96613
Click to expand...
You asked the Internet a question and got "just google it" for an answer? Seems like this AI business is shaping up to be like humanity pretty well!
 
  • Haha
  • Like
Reactions: MadRat, Tlh97, IEC and 3 others
M

mikegg

Golden Member
Jan 30, 2010
1,784
413
136
Here's the Premium version of ChatGPT (using GPT4):

1712898982712.png

Answer seems correct.

I don't understand the OP's rage.
 
poke01

poke01

Senior member
Mar 8, 2022
782
783
106
Markfw said:
I asked chatgpt a question. How stupid is this response. And people think AI is something they want ?????? It does not know what Zen 5 is ??? And it conjectures "what if" and does not give any useful response. So first It does not exist, then a what if...Google provides more useful information, the same as the last 20 years or so.

View attachment 96613
Click to expand...
I see OP you used GPT3.5.

The latest one GPT4.0.
 
I

igor_kavinski

Lifer
Jul 27, 2020
16,613
10,601
106
GPT4 is paid or requires signing in? I hate doing that.

This is what I got when trying to see if ChatGPT knows enough or if it can sift through online articles to tell me how I can accomplish a 7800X3D overclock:

1712899762035.png
Brilliant, isn't it?

I refuse to login to any AI service to use it because I don't need to do that for search.

But the fact that AI is getting something so simple wrong means one would need to tread very, very carefully when doing sensitive work, like writing code for any important business critical system. Use AI and you need to expend double the effort checking your work just to make sure that the AI didn't hallucinate and give you something that in rare circumstances will do the exact opposite of what you want to accomplish. This is very easy to do with something like C/C++.
 
  • Haha
Reactions: Tlh97, Thunder 57 and DaaQ
M

mikegg

Golden Member
Jan 30, 2010
1,784
413
136
igor_kavinski said:
GPT4 is paid or requires signing in? I hate doing that.

This is what I got when trying to see if ChatGPT knows enough or if it can sift through online articles to tell me how I can accomplish a 7800X3D overclock:

View attachment 96825
Brilliant, isn't it?

I refuse to login to any AI service to use it because I don't need to do that for search.
Click to expand...
I don't use Ryzen Master so you tell me if this answer is good.

Again, this is from GPT4, premium version of ChatGPT:

1712899994439.png
 
M

mikegg

Golden Member
Jan 30, 2010
1,784
413
136
Markfw said:
I typed in chatgpt, and thats what I got. I guess you need AI to invoke the correct version of AI.
Click to expand...
You're using the free version of ChatGPT which uses GPT3.5. I'm using the $20/month version which is the using GPT4, a much better and much newer model.

GPT3.5 is pretty good but GPT4 is substantially better.
poke01 said:
I see OP you used GPT3.5.

The latest one GPT4.0.
Click to expand...
Correct. You can only use GPT4 through a premium subscription or the paid API.
 
I

igor_kavinski

Lifer
Jul 27, 2020
16,613
10,601
106
mikegg said:
I don't use Ryzen Master so you tell me if this answer is good.
Click to expand...
It's again a generic answer. Doesn't seem like it has "learned" anything from the loads of overclocking experience of users on forums or even guides on enthusiast overclocking sites. I'm not sure what you are gaining from it but $20 sounds like a lot for generic advice.
 
M

mikegg

Golden Member
Jan 30, 2010
1,784
413
136
igor_kavinski said:
It's again a generic answer. Doesn't seem like it has "learned" anything from the loads of overclocking experience of users on forums or even guides on enthusiast overclocking sites. I'm not sure what you are gaining from it but $20 sounds like a lot for generic advice.
Click to expand...
How about this? Anything specific you want to know?

1712900417186.png
 
M

mikegg

Golden Member
Jan 30, 2010
1,784
413
136
I like how this thread started as a way to -removed- on ChatGPT. But now we can see that ChatGPT is more correct than some humans.

:p

As you're already aware, profanity is not allowed in the tech forums.
Daveybrat
AT Moderator
 
Last edited by a moderator:
H

Hitman928

Diamond Member
Apr 15, 2012
5,372
8,189
136
mikegg said:
Is @Thunder 57 correct or is ChatGPT correct? Can someone be the judge?

View attachment 96832
Click to expand...

ChatGPT is (mostly) wrong. The x3d cores can’t be overlocked as it suggests. The only thing you can do is mess with PBO to increase limits and basically hope the chip actually pushes a tiny bit higher clocks, but there’s no direct overlocking or messing with the clock frequencies. There is also bclk overclocking on some motherboards, but that messes with the whole system clocks and is likely to cause issues.

I’d say @Thunder 57 is more accurate in his answer but both are incomplete.
 
  • Like
Reactions: Tlh97 and igor_kavinski
I

igor_kavinski

Lifer
Jul 27, 2020
16,613
10,601
106
We are still far, far away from a future where we search something and the AI sifts through the results, understands which information is pertinent and valuable and provides a great useful summary of it with direct links to the source. I suppose something like this could be done locally with an LLM that parses Google results but that's a bit out of reach for the common user. Installing, configuring and tweaking the LLM to output results in the desired format is likely going to need some Python experience. AI should've let us just "talk" to a computer and make it do things. We are still not there yet, at least not for the vast majority of the cases. Most of the really cool stuff is behind paywalls or even not exposed to the masses, mostly because it's too expensive to run and just give away to everyone. Even Google can't do that with its fleet of thousands or who knows, maybe millions of servers.
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom