GPT is it threat or opportunity for Elsmar cove

[v9991]

the down side of GPT/LLM models have their pros and cons; i it is DEFINITELY NOT 100% reliable;
be it, hallucination, or just plain 'taking for granted approach i dont know...the reason but its fact that GPT is not always reliable on simple as well as complex queries.
* bing used to have three models accurate - standard -creative modes; now its gone; ( and it was not always useful when accurate mode is selected because of the restrictive nature; and while other modes were not reliable, because of the rationale etc )
* we heard of 1 such experiment by chatgpt, where it tried to automate the selection of model which best fits the query etc., which was also not right.
* most of it could be also due to insufficient prompts; if not incomplete prompts.

the upside for integration of GTP/LLM models with the queries with elsmar could be that., the fundamental and basic can be cleared with reasonable details and diversity; however, when it comes to decisions and creative ways to resolve scenarios, etc., human touch, ingenuity, and innovative solutions custom to the environment.

the point is.,
simpler use case could be elsmar could have an GPT INTERFACE APPENDED OR SUMMARY APPENDED to each thread;

 

[Hendi]

At least its absolutely annoying, posting the prompts would be sufficient for real intelligence. (Besides shitting on copyright, devaluing human work, burning up the planet and making individuals with remarkable behaviour quite rich.)

 

[Miner]

From what I have seen testing Chat GPT against Copilot, the answers are:

• Incomplete - I will get answers from each that are different with partial overlap. Each are "correct but incomplete"
• Theoretical (book knowledge) - answers you would get from a textbook and not tempered by experience

At this time, I see AI in Quality as "Entry Level", smart but incapable of functioning at the level of an experienced quality professional.

 

[ChrisM]

The old expression/acronym comes to mind.... "GIGO"
Garbage In, Garbage Out.
AI and Chat GPT etc responses are only as good/accurate as the information that they have been fed with, or seek out. An interaction of some sort (yes, even on an internet forum with "experts" responding, should always produce a more accurate and interesting outcome than an automated "intelligent" response (IMHO)

 

[Sidney Vianna]

the point is.,
simpler use case could be elsmar could have an GPT INTERFACE APPENDED OR SUMMARY APPENDED to each thread;

The Cove is in the business of knowledge sharing. Any AI delivery platform promotes a similar function and, myself have used a specific AI tool to research on a Cove related enquiry and report back here on the specific thread, making sure I disclosed the source of the material.

The exponential growth of AI based solutions mean that, without a question, in a “short” time, people will be able to seek and reach answers for many of the issues discussed here, elsewhere in the AI ecosystem. In those platforms, users won’t have to deal with grumpy contributors, misleading marketeers, false prophets and misguided egos. Until that happens this community will serve a valuable purpose.

Based on my interaction with him, I believe Marc Smith would be looking into some form of solution to integrate AI into this forum because he was always receptive to improvement and efficiency gains. Unfortunately we are not in a position to make any significant change to the Cove framework at this time.

So, while your suggestion has merits, it cannot be effected. At least, not at this time.

 

[Golfman25]

From my experience, sources like Chatgpt are scanning forums like this for some answers.

 

[Ed Panek]

AI is helpful but in the broad stroke department only. If you want it to outline a 20 page PPT presentation giving you the content description of each slide, it will be great to use. If you try to literally make it say what you need to say, youre going to have a bad time.

 

[Hendi]

In those platforms, users won’t have to deal with grumpy contributors, misleading marketeers, false prophets and misguided egos.

But how exactly shall it separate the good from the bad texts? Its a language model.
How will you pass every audit?
Actually, contrary what is often tols the truth is every auditor will report one less NC if you serve them acceptable gravy at the beginning.

 

[Miner]

It can't. It'll also have problems on sites like this where there are differences of opinion and different approaches. People can make judgments based on who's opinion they trust more. AI cannot and would probably go with the majority opinion even if that is wrong.

 

[Bev D]

Nor can it discern the truth in the haze of misinformation…