Page 3 of 3 [ 41 posts ]  Go to page Previous  1, 2, 3

ToughDiamond
Veteran
Veteran

User avatar

Joined: 15 Sep 2008
Age: 71
Gender: Male
Posts: 11,709

29 Apr 2024, 1:28 pm

Is this a rival to the "real" Chat GPT?

https://chataigpt.org/

Usually I go here:

https://chat.openai.com/

but it wasn't working for a while (strange problem, just got a blank page) so I looked around for an alternative. Its answers are much shorter and it annoyingly types "Hello!" before the answer and "If you have any more questions or would like further information on this topic, feel free to ask!" after it, every time. Not recommended, though if all you want is a quick potted answer to a question, it might do the job.

Anyway the "proper" one is working again, so it's business as usual.



Fenn
Veteran
Veteran

User avatar

Joined: 1 Sep 2014
Gender: Male
Posts: 2,700
Location: Pennsylvania

01 May 2024, 11:56 am

I think https://chataigpt.org/ is bogus. Checked a few web sites that list these kind of things and they think it is bogus too.

Example of ChatGPT helping me with social stuff at work:

Me:
ChatGPT, I wrote the following JIRA comment. I have, perhaps not provided enough context. Please advise me on how to add missing context:
For discussion: code change
It makes sense to change the code that creates the high and low water marks at 10 percent of the most recent value (number of rows) for each map. This was probably designed as an early warning or fail-safe against data errors. For example, if a feed suddenly sent a change effectively removing an entire list of needed data by mistake of because of an upstream glitch. I propose to change the code to calculate high and low water marks at most recent value plus or minus 50 percent (in place of most recent value plus or minus 10 percent). n +/- (n * .50) in place of n +/- (n * .10). This should decrease false-negatives from this fail safe and increase stability.

Technical Details: The package is named "PCK_CHECK_AUDIT" in the "GPR_ADMIN" schema, the PROCEDURE is named "UPDATE_MAPPING_RUNTIME_NUMBERS".


ChatGPT:
Your JIRA comment is clear and concise, but adding a bit more context can help ensure that everyone understands the reasoning behind the proposed code change. Here's a suggestion for adding missing context:

Context:
In our current implementation within the "PCK_CHECK_AUDIT" package in the "GPR_ADMIN" schema, the PROCEDURE "UPDATE_MAPPING_RUNTIME_NUMBERS" calculates high and low watermarks based on the most recent value plus or minus 10 percent. This was originally designed as a safeguard to catch sudden data errors, ensuring stability in our processes. However, upon reviewing our system's performance and potential false alarms, I propose a modification to enhance this safeguard.

Proposed Change:
I suggest adjusting the code to calculate high and low watermarks based on the most recent value plus or minus 50 percent (n +/- (n * .50)). This adjustment from the current 10 percent margin to a 50 percent margin aims to reduce false-negative triggers while maintaining a robust fail-safe mechanism.

By implementing this change, we expect a decrease in false alarms triggered by abrupt data changes, thereby enhancing system stability and reliability.

This additional context provides a clearer picture of why the change is proposed, its potential impact on system behavior, and how it aligns with ensuring data integrity and system stability.

Editor's note:
Social Skills is not just about making friends.
"Social Skills" means "People Skills" and most jobs require some people skills and people interaction.


_________________
ADHD-I(diagnosed) ASD-HF(diagnosed)
RDOS scores - Aspie score 131/200 - neurotypical score 69/200 - very likely Aspie


ToughDiamond
Veteran
Veteran

User avatar

Joined: 15 Sep 2008
Age: 71
Gender: Male
Posts: 11,709

01 May 2024, 1:11 pm

^
I wonder what they mean by "bogus?" Not the official one, not as good, not AI, more dangerous with regard to privacy? Certainly its answers are much shorter and mostly less informative. Example:

Me: What is the meaning of the Beatles' song "She Said She Said"

Chat GPT: "She Said She Said" by The Beatles is a song that has been subject to various interpretations, but it's widely believed to have been inspired by an encounter between John Lennon and actor Peter Fonda during a LSD trip. The song's lyrics are somewhat cryptic and abstract, typical of Lennon's style at the time.

In the song, the lyrics include phrases such as "She said, 'I know what it's like to be dead,'" and "She said, 'You don't understand what I said.'" These lines are thought to reflect Lennon's experiences and conversations during his drug-induced state, particularly the exchange he had with Fonda, who allegedly kept repeating "I know what it's like to be dead" during the LSD trip.

The song can be interpreted as exploring themes of existentialism, identity, and the nature of reality, all of which were common motifs in the psychedelic music of the 1960s. The repetition of "she said" in the lyrics suggests a conversation or dialogue, though the exact meaning of the exchange remains open to interpretation.

Overall, "She Said She Said" is one of The Beatles' more enigmatic songs, and its meaning may vary depending on the listener's perspective and interpretation.

Chataigpt: The Beatles' song "She Said She Said" was written by John Lennon and was inspired by a conversation he had with actor Peter Fonda during a party in Los Angeles. Fonda kept saying "I know what it's like to be dead" after having a near-death experience, which intrigued Lennon and became the basis for the song's lyrics. The song explores themes of confusion, introspection, and existentialism.

Chataigpt correctly mentions Fonda's near-death experience. Chatgpt does not.
Chataigpt's notion that Lennon was "intrigued" by Fonda's claim, is misleading. There's good evidence that Lennon found it an extreme pain in the butt, and the song's lyrics seem to reflect that.
Chataigpt presents its answer as brute fact. Chatgpt qualifies its answer with such phrases as "......various interpretations, but it's widely believed........"

As for your coding clarification, sadly I can't judge it because I don't know enough about the subject to make sense of even the clarified version, but it does kind of look clearer, and I believe you.



Fenn
Veteran
Veteran

User avatar

Joined: 1 Sep 2014
Gender: Male
Posts: 2,700
Location: Pennsylvania

03 May 2024, 1:14 pm

^ Bogus: masquerading as a well known web site. I don’t think it is affiliated with openai. Riding on coattails. If you go to the ChatGPT page on Wikipedia is has a link to chat.openai.com. That is the real deal.

The GPT code that ChatGPT uses is open source, but the training data is not, nor the human training tweaks. Also a pretender may not be able to throw as much cpu/cloud-computing horsepower at your query. So anyone can buy a stock car and enter a stock car race, but it is the tweaks that makes a winner. Google and Microsoft and everyone and his duck can do GPT these days, but the folks at openai are still winners. Some IBM PC knockoffs are actually better than the original, but most aren’t. I am particularly skeptical of someone who wants to make me think they are the well known “real thing” instead of openly branding as a legitimate competitor. YMMV

^ My brain can reconfigure itself for understanding and writing computer code. I can also talk to humans. Going back and forth is very hard. Like stripping gears on a stick. Also writing sometimes makes me work so hard on the mechanics of writing (spelling, grammar, punctuation, structure) i can actually loose track of what it is I wanted to say. It is like a bad stutter when Porky Pig just gives up and says an entirely different word or phrase or even a different thought. ChatGPT can take the gist and make more human English out of it.


_________________
ADHD-I(diagnosed) ASD-HF(diagnosed)
RDOS scores - Aspie score 131/200 - neurotypical score 69/200 - very likely Aspie


Fenn
Veteran
Veteran

User avatar

Joined: 1 Sep 2014
Gender: Male
Posts: 2,700
Location: Pennsylvania

03 May 2024, 1:37 pm

Correct: ChatGPT is not open source but close cousins are

https://qz.com/elon-musk-xai-grok-open- ... 1851343578

https://fortune.com/2024/04/08/open-sou ... nai-gpt-4/

https://github.com/nichtdax/awesome-tot ... en-chatgpt


_________________
ADHD-I(diagnosed) ASD-HF(diagnosed)
RDOS scores - Aspie score 131/200 - neurotypical score 69/200 - very likely Aspie


ToughDiamond
Veteran
Veteran

User avatar

Joined: 15 Sep 2008
Age: 71
Gender: Male
Posts: 11,709

03 May 2024, 2:03 pm

Fenn wrote:
^ Bogus: masquerading as a well known web site. I don’t think it is affiliated with openai. Riding on coattails. If you go to the ChatGPT page on Wikipedia is has a link to chat.openai.com. That is the real deal.

The GPT code that ChatGPT uses is open source, but the training data is not, nor the human training tweaks. Also a pretender may not be able to throw as much cpu/cloud-computing horsepower at your query. So anyone can buy a stock car and enter a stock car race, but it is the tweaks that makes a winner. Google and Microsoft and everyone and his duck can do GPT these days, but the folks at openai are still winners. Some IBM PC knockoffs are actually better than the original, but most aren’t. I am particularly skeptical of someone who wants to make me think they are the well known “real thing” instead of openly branding as a legitimate competitor. YMMV

^ My brain can reconfigure itself for understanding and writing computer code. I can also talk to humans. Going back and forth is very hard. Like stripping gears on a stick. Also writing sometimes makes me work so hard on the mechanics of writing (spelling, grammar, punctuation, structure) i can actually loose track of what it is I wanted to say. It is like a bad stutter when Porky Pig just gives up and says an entirely different word or phrase or even a different thought. ChatGPT can take the gist and make more human English out of it.

You're probably right about the imposter being a cheap knockoff job. It also has a dark "theme" which AFAIK can't be changed, and I find dark themes (white letters on black background) hard to read.

Thanks for the links - I'm wary of anything with Elon Musk in its name, but is he saying that ChatGPT is run by Microsoft? If so, which is the service that's least likely to marketise users' data? I see GitHub mentions a few variants.



ToughDiamond
Veteran
Veteran

User avatar

Joined: 15 Sep 2008
Age: 71
Gender: Male
Posts: 11,709

03 May 2024, 2:23 pm

Another wrong answer from Chat GPT on a "popular culture" question:

Me: In which episode of The Lone Ranger does the main protagonist say that nothing must stop the expansion of the West?

ChatGPT: In "The Lone Ranger" radio series, which aired from 1933 to 1954, the titular character, the Lone Ranger, often expressed sentiments about justice, fairness, and the expansion of the American West. While there isn't a specific episode where he explicitly says, "Nothing must stop the expansion of the West," the general theme of westward expansion and the advancement of civilization often permeated the show's narratives. The Lone Ranger himself embodied the values of frontier justice and the pioneering spirit of the West. If there's a particular episode you have in mind, I can help with more specific details from it!

I know for a fact that the Lone Ranger does say words to the effect of "Nothing must stop the expansion of the West" in an episode of that show. On further questioning, Chat GPT admitted that it couldn't find it, and then that "as an AI" it doesn't have direct access to specific databases or transcripts. I'm disappointed.



ToughDiamond
Veteran
Veteran

User avatar

Joined: 15 Sep 2008
Age: 71
Gender: Male
Posts: 11,709

12 Jun 2024, 12:57 pm

Now they seem to have removed the option to use it "anonymously," i.e. you now have to give them an email address and log in.



Fenn
Veteran
Veteran

User avatar

Joined: 1 Sep 2014
Gender: Male
Posts: 2,700
Location: Pennsylvania

20 Jun 2024, 12:54 pm

ChatGPT has been giving free trials of GPT4 a few hours a day. This may be to create more user data for new and experimental features or in hopes of selling more non-free accounts or both. New feature: ChatGPT will sometimes reply to a question with a real-time web search. This addresses the issue of a limited database/dataset (ChatGPT 3 has been trained on a dataset many times the size of a full set of encyclopedias but many times smaller than the dataset Google uses) and address ChatGPT being unable to answer questions about current events and even its own UI (which didn’t exist when the training was occurring). It also makes ChatGPT more competitive with Google AI (Majel/Gemini) and Microsoft AI (Bing/Sydney/Co-Pilot). And makes one wonder about who is really doing the innovating.


_________________
ADHD-I(diagnosed) ASD-HF(diagnosed)
RDOS scores - Aspie score 131/200 - neurotypical score 69/200 - very likely Aspie