GPT4 update

gabasiq

Junior Member
Joined
Aug 19, 2014
Messages
148
Reaction score
76
Hey guys, I just wanted to share this latest chatGPT update impact it had on my own work, maybe it's useful for someone else.
1 - It can browse the internet, I gave it a google dork and asked for it to explain it to me and go search it himself. It translated the Dork from Google to Bing while Binging.
Imagem do WhatsApp de 2023-11-15 à(s) 20.53.07_16e28e81.jpg
2 - It makes some pretty ok logos. I know it always had Dall-e as a plugin, but it never made decent logos when I asked. This time I made a couple that came out pretty good, first try:
Screenshot_7.pngScreenshot_3.png
3 - I don't know if you ever got stuck with GPT while coding, but the damn thing would forget what we were talking about and repeat itself. Looping the same mistakes over and over. Now it remembers a lot more of the conversation.

Also, be careful when making logos or images with Dall-e because that will burn up your tokens real fast.

Other than that there's nothing I've noticed. I've asked it what are the new updates and it does describe generalities. But those 3 are the ones that made the most impact to me personally.
 
So where do you use this thing? What's the URL?
 
For logos and more photorealistic images, use Stable diffusion. I found it more advanced and detailed in creating real stuff than Dall-E 3.

And yes I agree with you Dall-E can burn your token credits really fast. Its around $0.040 per image generated if I remembered correctly.
 
For logos and more photorealistic images, use Stable diffusion. I found it more advanced and detailed in creating real stuff than Dall-E 3.

And yes I agree with you Dall-E can burn your token credits really fast. Its around $0.040 per image generated if I remembered correctly.
Agreed, I prefer stable diffusion for realistic images. I'm still using both stability and playgroundai.

For coding I'll test Grok against GPT once it's out, that should be interesting.
 
Agreed, I prefer stable diffusion for realistic images. I'm still using both stability and playgroundai.

For coding I'll test Grok against GPT once it's out, that should be interesting.

Wiw never heard about Grok before.. intresting
 
Agreed, I prefer stable diffusion for realistic images. I'm still using both stability and playgroundai.
Yeah good one. I like to play around with SD, just having some fun.

Prompt: A handsome American investor in his office.
OUTPUT: not far from real. It looks like the most real ones I get are those with no pre-defined 'Artist' and 'Style'


For coding I'll test Grok against GPT once it's out, that should be interesting.
Never heard about Grok tho, I like to see new competition in this kind of market especially what I would like to use. I dislike monopoly.
 

Attachments

  • 1.png
    1.png
    932.6 KB · Views: 11
  • 1.png
    1.png
    968 KB · Views: 10
Last edited:
3 - I don't know if you ever got stuck with GPT while coding, but the damn thing would forget what we were talking about and repeat itself. Looping the same mistakes over and over. Now it remembers a lot more of the conversation.
I've also noticed a distinct improvement in GPT-4 1106's API Retrieval Augmented Generation.

Yet, I can't speak for the ChatGPT side--haven't played around with ChatGPT's "GPTs".
 
For logos , I had a really great experience with LEONARDO AI , at first it can be hard before choosing the right model and do the right prompting. But once you figure that out it gives some solid results
 
I don't know if you ever got stuck with GPT while coding, but the damn thing would forget what we were talking about and repeat itself. Looping the same mistakes over and over. Now it remembers a lot more of the conversation.
GPT usually repeats itself during a communication bridge. More reasons why a surface reading is preferred.
 
Yeah good one. I like to play around with SD, just having some fun.

Prompt: A handsome American investor in his office.
OUTPUT: not far from real. It looks like the most real ones I get are those with no pre-defined 'Artist' and 'Style'



Never heard about Grok tho, I like to see new competition in this kind of market especially what I would like to use. I dislike monopoly.
Grok should be good competition for coding, I've been looking for coding AI's for a while and this one could be good.
There's also google dream generator which creates some impressive fun images. Other users upvote it so you can gauge what images people like for content. The prompts are just very uncertain all the time.
 
I've also noticed a distinct improvement in GPT-4 1106's API Retrieval Augmented Generation.

Yet, I can't speak for the ChatGPT side--haven't played around with ChatGPT's "GPTs".
Also now GPT can browse the internet, so I suspect it can read github for longer code to be sent. So it doesn't say the message was too big, but I still need to test that.
 
For logos , I had a really great experience with LEONARDO AI , at first it can be hard before choosing the right model and do the right prompting. But once you figure that out it gives some solid results
We use that at the gaming company I contract for, personally I never tried but the devs really enjoy it. Definitely one of their favorites!
 
GPT usually repeats itself during a communication bridge. More reasons why a surface reading is preferred.
Yeah eventually I got the hang of it, always reminding it and making it reread specific messages. It was a pain though, every time it gets more memory my coding life becomes a little more comfortable.
 
Yeah eventually I got the hang of it, always reminding it and making it reread specific messages. It was a pain though, every time it gets more memory my coding life becomes a little more comfortable.
When it hangs, sometimes all you have to do is quote where you want it to continue from and write continue.
 
now gpt4 is fun to use and more usefful too, with only data from 2021 was so limited now with a access to internet this get seriously good.
 
Back
Top