Chat GPT

[thetiger2015]

A lot of this is solving problems that don't exist. People can write the code faster than having the AI write it and then go back through looking for errors.

I've seen lots of people being amazed by it but having tried some of the copywriting things out, it was waste of time. The 'machine' just nabs other peoples content and re-writes it in to a mess of paragraphs that need sorting out. It doesn't understand anything about the industry, it just pulls wiki pages and other bits, it makes no sense to a human that understands the industry we're in.

I've also seen people going on about how its great at coding or building spreadsheets. It isn't. You have to go over everything manually because it breaks things and you can't see where it's broken. Because it's not a real brain, it's just pulling from a data source and doesn't know the context.

AI/Robots only work with the data inputted. Hence, they can easily be tripped up by human error within the data they're using. It could be used for health screening? Please god no. It's useless. It will drag error filled data in and make huge mistakes, which will get worse over time and nobody will be able to undo it because they won't know where it's come from.

The amazing thing about the human brain is the complexity and speed at which it processes vast amounts of data. Sure, a machine may be able to play a game of chess but that's one task, it can't play chess, take the dog for a walk and then bake a cake.

All I see is tech people getting excited about something that will make life even more complicated. There will be humans trying to fix AI software that just breaks itself, because it doesn't understand humour or lies or how to filter good quality data from poor.

 

[DontAsk]

I've seen lots of people being amazed by it but having tried some of the copywriting things out, it was waste of time. The 'machine' just nabs other peoples content and re-writes it in to a mess of paragraphs that need sorting out.

Garbage in, garbage out.

 

[DontAsk]

Not ChatGPT, but Bing https://arstechnica.com/information...loses-its-mind-when-fed-ars-technica-article/

 

[fisicx]

There have been a number of technical reports on ChatGPT and the underlying engine.

The main problem is that it's a closed system. It can only learn form what it already knows.

The energy costs in training a new dataset is enormous. Which means there is going to be ongoing costs for any business who rely on any AI type service

 

[AndyBowness]

I've used it recently to help with a product listing of a new product, and then with the caption for that product's Instagram announcement post.

I had to guide it then do slight edits to the copy, but it was very impressive. I can see how it will save me time once I get used to doing prompts, and the potential for these kinds of tools is pretty amazing.

 

[DazRave]

I had to guide it then do slight edits to the copy, but it was very impressive. I can see how it will save me time once I get used to doing prompts, and the potential for these kinds of tools is pretty amazing.

Nail on head.

People seem to keep thinking it's just tech people getting excited and that it's incorrect half the time etc but so far it's biggest USP has been just how quickly it allowed people to do things.

I use it to research for me, I ask it questions and follow-up questions and within a minute or two I've got a very good idea of the subject I'm researching. Would have painfully took me ages to digest the same information going in and out of web links.

It's not perfect but it's also not claiming to be.

The people who are dismissing this now as a fad or just an over hyped tech gadget, I feel, are the sorts of people who'd have dismissed the idea that absolutely everyone will want to have a computer/phone connected to the internet.

AI, depending how it's eventually used, will become a disrupter. Right now a lot of people can't fathom how it will ever be needed just like a lot of big impact tech over the years.

 

[PashatGavim]

I like that it can be used as google and it will find you answers to those questions that you could manually search for much longer

 

[Maria Tanasie]

You made a great point there about having to manually tweak the output of ChatGPT. I think that's the essence of the tool and also the differentiator between a good user and bad user of ChatGPT.

From my personal experience, this tool was very useful when I was extremely tired and didn't have any creative juice left for content creation. With a few prompts, ChatGPT gave me an alright copy for some social media posts which I could adjust for my client's brand identity e voila! Saved me a ton of time and frustration.

 

[InternetMarketingHQltd]

None of them are very good. And there are all sorts of copyright complaints as the AI uses existing scripts (from GitHub) without acknowledgement.

I suspect things will improve but right now a lot of AI generated content isn’t that good. Consider if the energy industry all used AI for writing about pricing, because they could all use the same generator you will end up with very similar content.

A business I write technical manuals for have tried using an AI tool. It’s useless. It doesn’t even understand SI units. And because the manual has to explain the software settings using screenshots the AI just gives up.

Let's give it some time, but your points are very valid.

 

[thetiger2015]

The people who are dismissing this now as a fad or just an over hyped tech gadget, I feel, are the sorts of people who'd have dismissed the idea that absolutely everyone will want to have a computer/phone connected to the internet.

Slightly different. Phones/internet were new fangled and confusing for older people. These AI things are just rubbish. It's the disappointing part, not the new technology part.

New technology should be innovative and clever, inspiring and something that makes you think 'oh wow, I couldn't do that'. This AI stuff isn't AI at all, it's just bots and scrapers, programmed by a human, with mistakes included. There's no intelligence in it, just lines of code doing what they've been told to do.

Might be good for pulling together research bits, if the data already exists and you just want to group things together from multiple sources but that software has existed for over a decade. Scrapers. It might just be a clever scraping tool I guess.

I'm just disappointed with it. Nothing to do with 'fad' or 'gadgets'. I just don't see this as AI. It's not learning, it's just pulling in more and more irrelevant fluff and one small problems becomes a huge problem, because there's no intelligence in it. If it was intelligent, it would be able to tell the different between good and poor quality content and design, it would ask itself questions, it would be a 'brain'.

 

[InternetMarketingHQltd]

Let's give it some time, but your points are very valid.

The tool is being refined.

 

[DazRave]

Slightly different. Phones/internet were new fangled and confusing for older people. These AI things are just rubbish. It's the disappointing part, not the new technology part.

New technology should be innovative and clever, inspiring and something that makes you think 'oh wow, I couldn't do that'. This AI stuff isn't AI at all, it's just bots and scrapers, programmed by a human, with mistakes included. There's no intelligence in it, just lines of code doing what they've been told to do.

Might be good for pulling together research bits, if the data already exists and you just want to group things together from multiple sources but that software has existed for over a decade. Scrapers. It might just be a clever scraping tool I guess.

I'm just disappointed with it. Nothing to do with 'fad' or 'gadgets'. I just don't see this as AI. It's not learning, it's just pulling in more and more irrelevant fluff and one small problems becomes a huge problem, because there's no intelligence in it. If it was intelligent, it would be able to tell the different between good and poor quality content and design, it would ask itself questions, it would be a 'brain'.

Have you used it, properly?

 

[DanH]

I was pretty impressed with it to be honest, it writes better than some alleged copywriters, and certainly useful as a starting point for certain marketing projects that involve writing e.g. social posts/short blogs etc. I can see a lot of Fiverr and PeoplePerHour gigs disappearing...

As to whether Google will be able to tell and therefore penalise, it may be able to now but I bet it will reach a point, and rather quickly, where it can't. And, to be fair, if you look at some of the listicles that appear at the very top of Google when you search for 'best x product', most of them are very poorly written or constructed using a scraper or similar so I don't think Google really cares from what I see on a regular basis!

 

[InternetMarketingHQltd]

I was pretty impressed with it to be honest, it writes better than some alleged copywriters, and certainly useful as a starting point for certain marketing projects that involve writing e.g. social posts/short blogs etc. I can see a lot of Fiverr and PeoplePerHour gigs disappearing...

As to whether Google will be able to tell and therefore penalise, it may be able to now but I bet it will reach a point, and rather quickly, where it can't. And, to be fair, if you look at some of the listicles that appear at the very top of Google when you search for 'best x product', most of them are very poorly written or constructed using a scraper or similar so I don't think Google really cares from what I see on a regular basis!

Whatever the case, that's where we are headed. They will make it work, they can't bring it this far for it just to be mediocre software, of course, there are listening to well this feedback? We might as well embrace it instead of criticising and rubbishing the invention? If we may call it that? Happens all the time anyway when there is a disrupter.

 

[vivente]

I think I was being overly strict too! It'll be interesting to see how student examination boards respond to it - they have enough problems already with third parties writing papers for a fee.

However, I'm sure there'll be a burgeoning software industry analysing whether or not a piece was written by a bot.

Unfortunately I think the students of the future will be forced into learning and researching in the virtual environment. They will do all their research in that environment and it will be monitored. Everything they write will be recorded. Every single word. The sources they use. The notes they take. All monitored. If they suddenly start writing things that is not related to anything they've researched, taken notes on, etc then it will be detected as probably not their work. They'll probably not even make it to submission. It will flagged immediately and if a human reviewer is to intervene it will be almost immediate. If they can't prove how they came to x or y under questioning when it is fresh in their head then they will not be credited and repeated failure to explain conclusions will end in failure. Can you imagine anything worse

 

[fisicx]

Unfortunately I think the students of the future will be forced into learning and researching in the virtual environment. They will do all their research in that environment and it will be monitored. Everything they write will be recorded. Every single word. The sources they use. The notes they take. All monitored. If they suddenly start writing things that is not related to anything they've researched, taken notes on, etc then it will be detected as probably not their work. They'll probably not even make it to submission. It will flagged immediately and if a human reviewer is to intervene it will be almost immediate. If they can't prove how they came to x or y under questioning when it is fresh in their head then they will not be credited and repeated failure to explain conclusions will end in failure. Can you imagine anything worse

That makes no sense at all. Not everything will be monitored or recorded.

 

[DazRave]

Unfortunately I think the students of the future will be forced into learning and researching in the virtual environment. They will do all their research in that environment and it will be monitored. Everything they write will be recorded. Every single word. The sources they use. The notes they take. All monitored. If they suddenly start writing things that is not related to anything they've researched, taken notes on, etc then it will be detected as probably not their work. They'll probably not even make it to submission. It will flagged immediately and if a human reviewer is to intervene it will be almost immediate. If they can't prove how they came to x or y under questioning when it is fresh in their head then they will not be credited and repeated failure to explain conclusions will end in failure. Can you imagine anything worse

Bloody hell, I'll start making a tinfoil hat shall I.

It's the same story when the internet became a problem helping people find answers quickly, they will just adapt.

 

[MBE2017]

For anyone interested, the next version of chatGPT4 is being released next week and will be a lot more powerful, and supposedly draw from not just text but video etc.

I suspect this will make this AI even more in demand, the speed of change is incredible.

 

[fisicx]

I’m now getting loads of emails from people offering me AI generated content. If enough people use these sort of services the net will be even more awash with junk. It’s already bad enough but soon you will be researching something and see inaccurate and misleading articles, posts, product descriptions and so on.

Google dumped article sites, spun content and guest posts. It won’t take long before they develop tools to ignore ML generated content.

It’s not AI. True AI is years away. What you have right now is a closed dataset with some clever programming.

I do agree that the content can be good but it’s also often wrong.

 

[tzach]

I've created a WooCommerce extension that auto-generates eye-catching product descriptions using GPT 3.5 Turbo.

 

[fisicx]

I've created a WooCommerce extension that auto-generates eye-catching product descriptions using GPT 3.5 Turbo.

Have you tested how well your descriptions improve conversions?

Does it do both short and full description? Does it rework the meta description to align with the generated content?

 

[tzach]

Have you tested how well your descriptions improve conversions?

Does it do both short and full description? Does it rework the meta description to align with the generated content?

It does both short and full. I'm beginning to try it out these days, I'll make sure to update on how it goes

 

[DontAsk]

I’m now getting loads of emails from people offering me AI generated content.

It would be fun to use AI to generate replies

 

[Gussie]

Has anything like chatgpt been available to the public before? I wouldn't expect a Tony Stark Jarvis A.I at first run, therefore what they have given is adequate for the present for me. With advanced prompt engineering I managed to get some insights, which would have made my research longer. I am also glad that it has disrupted certain things especially for google and other serach engines, an industry shouldn't be ran on monopolistic tendecies. Alot of problems arise and its a repeat of the Carnegie, Rockerfeller era. Google arleady has fluff content with very good H1s anywhere

 

[japancool]

I asked it to write me a short product descriptoion, to see what it would come up with.

It got some basic facts wrong, like the item's colour scheme.

 

[fisicx]

Has anything like chatgpt been available to the public before? I wouldn't expect a Tony Stark Jarvis A.I at first run, therefore what they have given is adequate for the present for me. With advanced prompt engineering I managed to get some insights, which would have made my research longer. I am also glad that it has disrupted certain things especially for google and other serach engines, an industry shouldn't be ran on monopolistic tendecies. Alot of problems arise and its a repeat of the Carnegie, Rockerfeller era. Google arleady has fluff content with very good H1s anywhere

Was that generated by ChatGPT? If so it’s not very good

 

[DazRave]

It’s not AI. True AI is years away. What you have right now is a closed dataset with some clever programming.

I don't really understand this point. I think you are confusing/expecting artificial consciousness with artificial intelligence.

 

[ctrlbrk]

@Ozzy not a ChatGPT post per se, but I'm genuinely interested as to whether your mod team here looks out for AI-generated posts and are able to spot them?

I see this phenomenon in another forum I frequent and, for a couple of years now, bots have been posting and have become increasingly smarter, which means people aren't always able to detect them as non-humans.

 

[fisicx]

We had one member here posted a bunch of AI generated posts. They were really easy to spot. Mainly because they took 200 words to say something that could be done in 10.

The other big clue was the writing style. It was inconsistent across the posts.

 

[ctrlbrk]

We had one member here posted a bunch of AI generated posts. They were really easy to spot. Mainly because they took 200 words to say something that could be done in 10.

The other big clue was the writing style. It was inconsistent across the posts.

If you're talking about manual copy and paste, not talking about someone who copies and pastes AI-generated text.

Talking about bots that register on forums, pick a topic and reply autonomously.


Or did I misunderstand you?

 

[fisicx]

That was copy and paste. We had one bot but the initial post was held for moderation and got nuked.

 

[tzach]

I asked it to write me a short product descriptoion, to see what it would come up with.

It got some basic facts wrong, like the item's colour scheme.

What was your prompt? It's crucial for generating quality descriptions.

 

[japancool]

What was your prompt? It's crucial for generating quality descriptions.

Write me a short description of "Product X".

The product in question is a specific, branded product. It's like asking it to write a description of a 330ml Coke can, and it telling you the can was blue.

 

[tzach]

Write me a short description of "Product X".

The product in question is a specific, branded product. It's like asking it to write a description of a 330ml Coke can, and it telling you the can was blue.

ChatGPT works best when you define its role and task.

Try something like: "Act like an expert in the sales and marketing industry and create a captivating eCommerce product description for "Popover Heavyweight Hooded Sweatshirt in Red". The current product description is: "With an oversized fit and a fluffy inside, this hooded sweatshirt is ideal for informal settings, lounging and post-workout.".

You can expect significantly improved outcomes

 

[japancool]

Waste of time. I might as well just use the standard product description.

 

[amyjones]

I think you have to see it for what it is, a beginning of the AI journey. I’m not an expert but I think you will find ChapGPT will be very different in two years time. Imagine the first mobile phones with the suitcase full of batteries when they first launched, ten to fifteen years later they became the size of a matchbox, now they are becoming much larger again.

Everything evolves, but personally I have been impressed by it. In five years I think I will be amazed by it.

Yeah Maybe in the future ChaptGPT will be the new standard for AI development, who knows? I guess only time will tell. For now, it is a great start to what could potentially become something much bigger and better in the future. Thanks for bringing this up!

I think one thing we can all agree on is that no matter what happens with ChapGPT and AI development, it is going to be an exciting future. We've already seen the beginnings of things like self-driving cars and voice-controlled virtual assistants. If these continue to develop and become more widely available then I think our lives will become much easier in many ways. Who knows what else could come with it? The possibilities are seemingly endless!

 

[lesliedocherty]

ChatGPT works best when you define its role and task.

Try something like: "Act like an expert in the sales and marketing industry and create a captivating eCommerce product description for "Popover Heavyweight Hooded Sweatshirt in Red". The current product description is: "With an oversized fit and a fluffy inside, this hooded sweatshirt is ideal for informal settings, lounging and post-workout.".

You can expect significantly improved outcomes

We have been using it for product and brand descriptions for a few months now, big improvements (with a bit of tweaking), time saved at least 50% on product descriptions and brand descriptions take less than 10% of the time

 

[tzach]

We have been using it for product and brand descriptions for a few months now, big improvements (with a bit of tweaking), time saved at least 50% on product descriptions and brand descriptions take less than 10% of the time

Great to hear! I believe the potential is huge, especially for eCom businesses.

 

[fisicx]

If these continue to develop and become more widely available then I think our lives will become much easier in many ways. Who knows what else could come with it? The possibilities are seemingly endless!

Or it could result in a lot of people out of work. For example anyone providing business service would no longer be required. Your company for example would be replaced with AI tools.

 

[estwig]

Or it could result in a lot of people out of work. For example anyone providing business service would no longer be required. Your company for example would be replaced with AI tools.

Hopefully it'll work out how to design a loft conversion, then I can sit on my arse and let an app do the work for me!