ChatGPT is dyin'
 

ChatGPT has been cut back again. Now it no longer has the ability to read the dialogue history. Essentially, it becomes useless and operates in a mode: one question - one answer.

Perhaps that's not a prob for GPT API users, but the ChatGPT service itself can't be used for chatting anymore :2 cents:

Since when? Seems to be running fine for me

Same...

.....

That’s not the case for chatgpt plus . Running absolutely fine for me .

Working on Plus and free...

ya u should post links when you make shit up :1orglaugh

#

They're cutting back on the free version ? ?

I'm talking about the GPT-4 model for ChatGPT. Try asking it to re-read your entire conversation in the current chat and do something with it, such as analyze it, summarize it, etc. It was working fine about a month ago or so.

Questions?

Running fine for me and history is all there

Bing chat suppose to run on chatgpt 4 but it's been acting same all the time.

AI is just a subtype of machine learning, I don't think it's as powerful as people think it is, just sayin

It's absolutely not a super duper magic wand, but it's definitely a good helper when you need to write an email to your bank, a blog article, or a function for your software project. It's a fine tool for auto-blogging that allows you to get rid of human copywriters.

Give it a news source, tell it to read each article and write a unique version with an accent on ecology, feminist rights or alien life protection ;) You can also ask it to write critical analyses of the news or something like that. This doesn't violate any copyrights and allows you to run your own information agency in any field.

So it's just a time-saving and automation tool that allows you to do more and spend less - both your money and your time. :2 cents:

SEO wants to let you know his death following him

Seems to be working perfectly fine for me. As usual.

As mentioned above, ask GPT-4 to re-read and summarize your entire conversation.

works fine here
command: re-read and summarize this entire conversation.
response:
"Certainly! Here's a summarized version of our conversation:"
and then gives a proper summary

paid account gpt-4

It's been working perfectly for me - in fact GPT4 seems to be working faster than it used to be...

Hmm. It's working for me now, but it wasn't when I started this thread. Seems like they are doing something with it again...

#

I think it's random, meaning they are probably "turning off" in order to save resources.

Yep, seems so.

Not sure how much you've used it, but I've probably spent 100-200 hours using ChatGPT now.


I'd say that ChatGPT has good days and bad days. Sometimes it re-introduces old bugs back into code, or lies to you, or completely loses its mind and has no idea what the fuck it is doing.

I gave it code once to make a single change, it then spit out some kid's Mechanical Engineering homework formulas. I've seen ChatGPT get completely confused and give me answers that I never asked for and it thought it was talking to someone else.

There is a real skill to working with ChatGPT, because it can't figure out things on its own. It is like a child with a massive brain but no life experience. You ask it to code for you and it will, but the code will be shitty unless you explain to it exactly what you need, how it needs to work, why it should work that way, and then get it to build you the code in a modular fashion, because once you get over 200+ lines of code, ChatGPT breaks down and will create bugs, omit shit, etc etc.

yea this is why i'm not worried about all programmers being replaced or unemployed. its just another tool to get the job done.

#

Once you understand what ChatGPT can do and can't do, it becomes an amazing tool because you can get it to output awesome code within a few tries.


If you're some n00b just starting out, it will be hard to understand the nuances of ChatGPT and it will fucking give you shit code, give you errors, confuse you (almost on purpose) and mislead you. You need to learn to query it in ways that both of you understand. Things as simple as, for example, getting the script to scan a CSV file and then scan a folder and compare the two, then update rows in an XLS file, you need to know how to get ChatGPT to add arrays, variables, and other shit, because if you just say:

"Hey ChatGPT, I need you to write me a python script to do A, B, and C, then output the info to D, and take D and create a JSON file(or whatever)"

Then ChatGPT will fuck it up royally and lead you to wild goose chases as it will claim it knows how to accomplish it, then when your code constantly spits out errors, it will then tell you it can fix it, and will then re-introduce any errors you have removed before it can even write good code. So you need to do "step A" properly, then once that works, do "step B" and then work from there. ChatGPT is awesome, but it takes probably nearly as much skill at coding to learn how to use ChatGPT as it probably does to learn to code. (If you took courses to learn to program, you'd probably be better off, but ChatGPT can be awesome once you learn how to navigate it.)


I've been wanting to put up ads for local businesses to contact me to see how AI could be used to improve their businesses. I know of one business in particular that is about a 10-12 million a year business that could really use my help, but they fired me because I didn't take the COVID "vaccine" so I would never help those cocksuckers. Also, they were mad as fuck at me because I hired a transgender person with a great resume and tons of experience for a 1 month contract where we needed workers and had none. Fucking bigots, if someone has a good resume, who cares who they fuck or what they look like. Just do the fucking job.

is this something chatgpt can help you get done? write the campaigns, build the funnels, even interact with the customers, plus exclude bigots and racists? lol

i've seen some nice cheatsheets floating around FB and LI...

#

I think...

When a thousand creative original writers manually generate prose, it will take longer to find incidental overlap in their content, than when one million generative AI users cranking out derivative (textually inbred) works do the same.

AI may at some levels introduce orders of magnitude greater "soup base" content, and faster, possibly higher level permutations and combinations of the "seed data", but it will generate detectable bland homogenous output that the latest spam detecting systems (always using one level higher of AI) will catch and penalize.

Does anyone seriously believe that the biggest search engines won't be able to buy actual snippets of "civillian-generated" content. They won't need algorithms to detect AI spam - they can just search their dictionaries.

Anyone believing that the AI tools available to content creators can outrun spam filters is interestingly optimistic.

If SE spam is not the goal, how long do you think intelligent human readers (those who don't read or believe marketing BS) will take to be put off by the creative and scholarly limitations of generative AI?

Just saying.

:2 cents:

I have worked very closely with ChatGPT since its early days (check this out) and I know how it works. Actually, there is no super-duper secret to writing proper prompts to make it work exactly the way you want.

The main thing every user has to know about it is that ChatGPT (OpenAI GPT) models have a very huge database which contains just about everything and it really works perfectly as a tool which has been trained to generate a quality text (an article, a code or a medical diagnosis etc). However it's just a language model.

The only thing everyone has to keep in mind is that ChatGPT doesn't understand what it's doing. It does not think abstractly. It doesn't see the big picture. It just generates an appropriate textual response to a concrete textual prompt, and nothing else.

Anyone believing that the AI tools available to content creators can outrun spam filters is interestingly optimistic.

You can do it just easily :pimp

I would very much like to do this after my tube-site killer is ready.

Almost done.

Your sig is too big

we've got some really nice examples of saturation in our industry. tgps get big, so everybody builds one. tubes get big, so everybody builds one. what happens is original content becomes rare.

of course big tech can detect AI generated content. absolutely. they've published studies and built tools for it.

a million people using chatgpt to write a marketing landing page will get similar results. that's the whole point of it. consistent and accurate results will look the same for multiple people.

these tools weren't created for lazy people who don't want to do their jobs anymore.

#

What is "the job" that generative AI lets "non-lazy people" do?

And is the work product of such efforts of any value? To whom?

It seems that producing a consistent, intentionally limited, output benefits mostly those who are vested in maintaining that output/narrative, while consciously obscuring actual critical authoritative, competence-based, information.

Incomplete, even misleading, fluff overshadows (by volume), displaces, and ultimately replaces authoritative knowledge.

To keep it in perspective, adult entertainment is all about fluff - unless you're dealing with safety in the industry, and health-related issues.

And where other, non-X-biz, more serious affairs are concerned, socially, politically, gen AI is an even greater threat to objective understanding.

:2 cents:

yea as you say, the AI gets filled with fluff. the thread about AI getting dumber goes deeper on that issue.

one "job" that is being supplemented massively right now is teaching. chatgpt is a part of every classroom, either as a teaching tool or cheat tool by students.

education seems to be the largest adoption group so far. marketing and SEO are close behind. sales teams are using it to help manage CRM, eliminating at least some hours of work, if not a junior position.

all i gotta say about it is:



#

Fair enough. Point taken.

Quod erat demonstrandum. :thumbsup