Claude AI Safety & Sweeps Offers

Hey guys. I've been having a lot of trouble with Anthropic's Claude 3.5 Sonnet for affiliate work. I really like the artefacts feature and find that Claude 3.5 is a lot better than ChatGPT for routine use/editing/translation etc, but every time I try something affiliate/sweeps Claude just clams up and says it won't help me because of AI Safety (it think I'm being unethical/scamming people).

Perplexity is excellent for search & research, ChatGPT seems pretty average now & I get average-tier work out of it. But have any of you tried anything else for coming up with strategies/copy which is as effective as Claude 3.5? Or do you have some other preferred AI tool for marketing?

But aren't sweeps offers scammy?

Depends on the offer? Don't think they're blanket all scammy. Many are though so I can understand why Anthropic would block them.

Hehe, prompting it differently always works, but that requires understanding of how that works.

I never tell anyone to do anything by editing.

I tell them to do xyz, provide corrections, provide corrections, provide corrections, provide corrections within the same chat window.

Example 4 messages (with 3 corrections in chat window after testing code):

1 message (worked, but not in fashion wanted): "i need to know how to make a script to change all pixels in an image, so they're unique"
2 message (broke and didn't work due to syntax errors): "i need to make a python script to change all pixels in an image, so they're all unique, but the image stays the same"
3 message (nothing worked): "write from scratch; nothing works"
4 message (all worked, but did too complex manipulation): "it has changed everything into a rainbow like image; need something normal"

And then after this set I got a working script that someone on the market tries to push for $250. I think that's a scam offer, because it's a simple algorithm, but that's what it is. People in sales / marketing tend to not respect anything these days.

Back then it was different. There were people who had 20 likes under a post.
https://www.blackhatworld.com/members/popzzz.42969/
The difference was that people were sharing and they were getting stuff back. Today they're not sharing and keeping for themselves expecting things to work living nothing but a secret. Secrets are not useful. Privacy is not real. The reality sees everything. You can't hide.

AI safety doesn't exist for me.

A nuclear something for example... you would ask about nature of atoms and so on. Not about the nuclear thing itself.

I mean, if you go deep into your own mind, you will ask better questions to AI and that's how it works.

It's an extension of you. Just like any tech gadget. Improve yourself, not AI.

Walsingham said:

Hey guys. I've been having a lot of trouble with Anthropic's Claude 3.5 Sonnet for affiliate work. I really like the artefacts feature and find that Claude 3.5 is a lot better than ChatGPT for routine use/editing/translation etc, but every time I try something affiliate/sweeps Claude just clams up and says it won't help me because of AI Safety (it think I'm being unethical/scamming people).

Perplexity is excellent for search & research, ChatGPT seems pretty average now & I get average-tier work out of it. But have any of you tried anything else for coming up with strategies/copy which is as effective as Claude 3.5? Or do you have some other preferred AI tool for marketing?

Click to expand...

Hi! I've heard that many people face similar issues with Claude 3.5, especially when it comes to affiliate marketing. It might be worth trying Jasper AI or Copy.ai for creating strategies and copies — they are quite effective for marketing tasks and may be more accommodating to such requests.

OccultKingdoms said:

Hehe, prompting it differently always works, but that requires understanding of how that works.

I never tell anyone to do anything by editing.

I tell them to do xyz, provide corrections, provide corrections, provide corrections, provide corrections within the same chat window.

Example 4 messages (with 3 corrections in chat window after testing code):

1 message (worked, but not in fashion wanted): "i need to know how to make a script to change all pixels in an image, so they're unique"
2 message (broke and didn't work due to syntax errors): "i need to make a python script to change all pixels in an image, so they're all unique, but the image stays the same"
3 message (nothing worked): "write from scratch; nothing works"
4 message (all worked, but did too complex manipulation): "it has changed everything into a rainbow like image; need something normal"

And then after this set I got a working script that someone on the market tries to push for $250. I think that's a scam offer, because it's a simple algorithm, but that's what it is. People in sales / marketing tend to not respect anything these days.

Back then it was different. There were people who had 20 likes under a post.
https://www.blackhatworld.com/members/popzzz.42969/
The difference was that people were sharing and they were getting stuff back. Today they're not sharing and keeping for themselves expecting things to work living nothing but a secret. Secrets are not useful. Privacy is not real. The reality sees everything. You can't hide.

AI safety doesn't exist for me.

A nuclear something for example... you would ask about nature of atoms and so on. Not about the nuclear thing itself.

I mean, if you go deep into your own mind, you will ask better questions to AI and that's how it works.

It's an extension of you. Just like any tech gadget. Improve yourself, not AI.

Click to expand...

"AI safety doesn't exist for me."

Based. Love this comment.

MiSparrow said:

Hi! I've heard that many people face similar issues with Claude 3.5, especially when it comes to affiliate marketing. It might be worth trying Jasper AI or Copy.ai for creating strategies and copies — they are quite effective for marketing tasks and may be more accommodating to such requests.

Click to expand...

Thanks! Surprised Jasper is still around. I used it a fair bit back in 2022 and assumed it got killed off by OpenAI - will look into Copy AI as well.

Are you using the API or the frontend? Frontend is a lot more censored