Question ChatGPT

[nandobadam]

What is the level of accuracy and truth in the answers provided by ChatGPT for general topics related to Windows, HDD/SSD/Flash memory, PC Hardware, consoles, electronics?

I usually use Windows Forum ChatGPT Costly AI Model

 

[Solaris17]

That is unfortunately almost entirety unquantifiable. I use it to write some code pretty frequently when I am needing new functions. I always double check it. It saves me a ton of time but there are times I read over it and fix something or just do it myself because it didn’t really make the mark.

I consider that complex. Then yesterday I asked it a question and it responded and gave a reference URL. The URL being totally made up.

Most models if you go to there sites will tell you what that model or what there service excels at. It won’t really answer any kind of accuracy question but asking a coding model about plant biology probably isn’t the smartest.

I think AI is helpful but you ALREADY need to fundamentally know what you are doing so you can contextually understand its answers.

If you don’t know what your talking about out. In any given subject asking chatGPT is a leap of blind faith.

 

[Macro Device]

If your question only implies knowledge then it's mostly sensible.
If your question also implies math and thinking process then AI will be erroneous most of the time and will only provide correct answers by pure accident. For example, ChatGPT 3.5 can't calculate PPI for non-16:9 resolutions. Spouted random bollocks like 83, 98, 79, and even 257 for the 3440x1440 on 40" PPI (actual PPI is 93.23 in this case).
No idea about this particular model but all models I've seen operate dated data, sometimes 2+ years old. Keep that in mind because AI might have negative clue what you're talking about.

 

[nandobadam]

more technical questions such as how SSD/HDD/Flash memory storage devices work
how Windows works and functions
how PC and console hardware works and technical questions
release dates of some hardware or implementation date of some Windows service

questions like what was the first version of Windows with support for ejecting drives, what year was autoparking inserted into the firmware of HDDs, do my HDD models and enclosure case models mentioned support autoparking? what are the names of the commands sent by Windows after ejecting? and more questions about PC hardware and consoles and electronic components

Is any ChatGPT free question reliable or not?

what is the level of accuracy and truth of ChatGPT?

 

[Solaris17]

Is any ChatGPT free question reliable or not?

what is the level of accuracy and truth of ChatGPT?

it’s not that simple.

If your asking these kinds of questions then maybe you need to understand more about how AI works; and for the time being it may be better for you to continue using search engines and asking your specific questions in forums.

 

[nandobadam]

In my questions I realize that some information is accurate but researching all the information he gave me is very complex because some of it I can't find on Google or it is datasheet or PDF information hidden on some website
I wonder if all these chatgpt have some source or base from which they collect the information and generate the answers?
I will give an example from the Windows Forum website, it uses GPT
I asked:
What year released 2.5" HDDs Seagate ST500LM030, WDC WD10JPVX-08JC3T5 and APPLE HDD HTS541010A99E662?
Answer:

The release years for the 2.5" HDDs you mentioned—Seagate ST500LM030, WDC WD10JPVX-08JC3T5, and APPLE HDD HTS541010A99E662 (Hitachi/Toshiba)—align with product cycles from their respective manufacturers. Here are the details and approximate release dates:

---

1. Seagate ST500LM030​

• HDD Series: This HDD belongs to the Seagate Barracuda Mobile series, designed for laptops and portable devices, combining low power consumption with good storage capacity.
• Release Year: 2016
  • Introduced as part of Seagate's transition to SMR (Shingled Magnetic Recording) technology in 2.5" drives for budget-friendly storage solutions.
• Notable Specifications:
  • Capacity: 500GB.
  • RPM: 5400.
  • Load/Unload technology for head parking.
• Source of Confirmation:Official Seagate datasheets and product specifications.
  • Seagate Barracuda Series HDD.

---

2. WDC WD10JPVX-08JC3T5​

• HDD Series: This HDD is part of the WD Blue series, which is targeted toward mainstream laptops and external enclosures for consumers looking for reliability and general performance.
• Release Year: 2013
  • Part of Western Digital's push to modernize its 2.5" HDD lineup with features like IntelliPark for automatic head parking and better support for portable systems.
• Notable Specifications:
  • Capacity: 1TB.
  • RPM: 5400.
  • IntelliPark and improved shock resistance.
• Source of Confirmation:Western Digital product guides and datasheets for the WD10JPVX series.
  • Western Digital Blue Series Specifications.

---

3. APPLE HDD HTS541010A99E662 (Hitachi/Toshiba)​

• HDD Series: This drive is manufactured by Hitachi (HGST) and is part of the Travelstar 5K1000 series but branded for use in Apple devices, such as MacBooks.
• Release Year: 2012
  • The original Hitachi Travelstar 5K1000 series launched in 2012, offering quiet operation via Load/Unload ramp technology and optimized for laptops and external use cases.
• Notable Specifications:
  • Capacity: 1TB.
  • RPM: 5400.
  • Shock tolerance and Load/Unload parking.
• Source of Confirmation:Archived Hitachi documentation and widely available product details for the 5K1000 series.
  • Hitachi (HGST) Travelstar Product Manuals via Western Digital.

---

Summary of Years​

HDD Model	Series/Manufacturer	Release Year	Linked Sources
Seagate ST500LM030	Barracuda Mobile (Seagate)	2016	Seagate
WDC WD10JPVX-08JC3T5	WD Blue (Western Digital)	2013	WD
APPLE HDD HTS541010A99E662	Travelstar 5K1000 (Hitachi)	2012	HGST

 

[Veseleil]

more technical questions such as how SSD/HDD/Flash memory storage devices work
how Windows works and functions
how PC and console hardware works and technical questions
release dates of some hardware or implementation date of some Windows service

questions like what was the first version of Windows with support for ejecting drives, what year was autoparking inserted into the firmware of HDDs, do my HDD models and enclosure case models mentioned support autoparking? what are the names of the commands sent by Windows after ejecting? and more questions about PC hardware and consoles and electronic components

Is any ChatGPT free question reliable or not?

what is the level of accuracy and truth of ChatGPT?

What @Solaris17 said...

Or just ask the AI those questions and then compare the results to the info from the reliable sources? Or do you want the forum members do that job for you?

 

[Count von Schwalbe]

Is any ChatGPT free question reliable or not?

It is inherently unreliable. Ask it a question, if you must, but use its answer as a starting point for research.

It will tell you a falsehood with the same confidence as a truth. Verify everything.

 

[Veseleil]

And @nandobadam, you've missed the opportunity to have a highly populated thread. If you did, let's say 100 PC hardware related questions for the chatGPT, compared the results with reliable sources, and then make a thread titled something like "chatGPT highly inaccurate, only X% of 100 answers ended up correct", you'd have couple of reply pages already...

 

[nandobadam]

I don't know where the reliable sources are located to compare with chatgpt

 

[notoperable]

It's a prediction engine sold as something intelligent. It's does this on such an incredible scale that the underlying hammer approach is washed out and LLMs can give you the impression of sophisticated reasoning - in terms of how a human views intelligent behavior.

LLMs are based on linguistics and probability chains and very far away from reasoning in terms of cognitive processes as we intuitively do.

There might be such mechanism put on top of LLMs in the futur, and that what Altman is selling everyone and tells it's around the corner, so it can enhance how this simulacrum performs, but ask it enough often and it will predict an answer, or, invent bullshit in simpler words.

Once it lands on a low confidence area, it's start to bullshit users better then any slimy used cars seller you ever met, if it hits the prediction correctly the while spiel goes on, if not, it apologize and recalculates the probabilities.

 

[ir_cow]

ChatGPT or any of those cannot be trusted at face value. Very helpful tool, but often wrong in the details. After all it's just scrubbing reddit and replaying back to you lol.

 

[notoperable]

ChatGPT or any of those cannot be trusted at face value. Very helpful tool, but often wrong in the details. After all it's just scrubbing reddit and replaying back to you lol.

It's full of shit 9 times of 10 if you put it under scrutiny

 

[nandobadam]

Is there a reliable research tool that covers the answers to my questions using reliable sources of information?

 

[Veseleil]

Is there a reliable research tool that covers the answers to my questions using reliable sources of information?

I use DuckDuckGo, and start from there.

 

[notoperable]

The whole trick about LLMs is the speed it does what it does: hammers down prediction at very very very high speed - and this way most users have the impression it's something very reasonable and cognitive capable - and this is not the case. I wonder what happens if they cant put any more lipstick on it and hit the wall with LLMs. Those won't and can't as per their design be some secret emergent way to intelligence when you feed it ever bigger amounts of data. It's a parrot that will repeat a confirmed result but it needs to be thought that a result is correct first, it predicts it's answer based on previous data without any reasoning of the causes that are producing the result. That's a bit crappy worded but I can't put I better into words now

I wouldn't rely upon it, it's very good at a summary or writing answers to government inquires, it can mimic a lawyer in writing quite well and come over as very sophisticated

 

[ir_cow]

Is there a reliable research tool that covers the answers to my questions using reliable sources of information?

ask the question into ChatGPT. Take that answer and google it. I certainly wouldn't use it as a source for term papers.

 

[notoperable]

ask the question into ChatGPT. Take that answer and google it. I certainly wouldn't use it as a source for term papers.

You forgot to add that by doing so you basically train OpenAI model for free

 

[nandobadam]

Doesn't GPT chat have the function of searching for questions on various websites, articles, collecting information and generating some answers based on what is spread across the internet?

 

[ir_cow]

They all do in some way. Co-pilot, meta, ChatGPT, Gemini, etc

 

[nandobadam]

He ChatGPT doesn't need to answer me directly but he needs to look for reliable sources and formulate an answer based on the information found on websites and documents to answer my question truthfully based on the sources.

 

[Solaris17]

He ChatGPT doesn't need to answer me directly but he needs to look for reliable sources and formulate an answer based on the information found on websites and documents to answer my question truthfully based on the sources.

When you figure it out, hire me so we can make billions.

 

[R-T-B]

I wonder if all these chatgpt have some source or base from which they collect the information and generate the answers?

That's exactly what happens. As for the souces, the answer is generally "the internet" and yes bad data can and does slip in.

Is there a reliable research tool that covers the answers to my questions using reliable sources of information?

There are curated LLM models that are generally subscription based. Perplexity is one I got a free year long subscription to from my ISP. It's not horrible.

 

[nandobadam]

That's exactly what happens. As for the souces, the answer is generally "the internet" and yes bad data can and does slip in.


There are curated LLM models that are generally subscription based. Perplexity is one I got a free year long subscription to from my ISP. It's not horrible.

What the best free LLM? or all same equal?

 

[Dr. Dro]

What the best free LLM? or all same equal?

In my experience it's easily X's Grok AI