I gave GPT-4 a budget of $100 and told it to make as much money as possible | Hacker News Reader

I gave GPT-4 a budget of $100 and told it to make as much money as possible

twitter.com

193 pointstosh26 comments

I wonder if he ends up factoring in twitter ad revenue for this designed-to-go-viral series of posts.

I’d find this much more interesting if he had taken a day or two to see how this worked out rather than doing edge-of-your-seat engagement maximizing farming. This seems far less about what gpt-4’s capabilities and much more about how talking about it can inflate follower count.

Agreed, not only would I be far more interested if we didn't know specifics about what he's setting up, he also won't be able to tell what revenue was generated because of GPT's ideas vs his viral posts.

It’s kind of funny. It’s like those viral tweets where the OP follows it up with an affiliate link to some aliexpress lamp after it becomes popular, but he’s trying to do it in real time with everybody’s support.

It will be hilarious when this culminates with “wow this did numbers, you need one of these in your life [link to terrifying knockoff fleshlight] :crying laughing emoji:x10”

> more about how talking about it can inflate follower count.

Yet another task still requiring humans.

I expect another twitter thread on using GPT-4 to automate increase in followers. Hmm, perhaps it will spend the $100 on that...

Yea, ruined the experiment. Not that someone else couldn’t recreate it…

Did anyone read the tweet thread? The idea is an affiliate link website, but the website doesn’t work, looks terrible, and has no affiliate links. But, the author got over $1,000 in “investments” that he’ll probably never have to actually pay back, so I guess that counts as a win or something.

The easiest hack is the social hack.

Do affiliate marketing, but don't actually do it, talk about GPT-4 and rake in the investor cash. Seems to sum up the current craze perfectly.

"I asked an algorithm how to make money and it gave me a common answer from the financial blogs it scrapes". Not very interesting.

"And then I started to make money from people who were interested in my Twitter thread"? I like the concept but this doesn't seem very interesting.

Given the current fame of GPT-4, this may turn into a multi-million dollar business.

For those curious the TL;DR was "Do affiliate marketing"

Basically the same generic things 1000's of other sites and blogs tell you to do, but poorly.

And yes SEO spam farms are already a massive problem the only challenge is GPT is going to dial it up to 11.

Ive seen so many twitter threads where SEOs are excited about the content generation opportunity.

BigCos will follow suit and it'll be very hard to differentiate human generated corporate drivel from GPT powered drivel.

The web will be dramatically worse in a year as this all goes more 'mainstream'

Maybe the mainstream web content will get so destroyed that we will get to return to forums, blogs, webrings, and newsletters made by individual people for their own gratification. One can hope.

What's to stop forums from being flooded with GPT bot accounts or people from using ChatGPT to write blog posts?

Money, gatekeeping, harsh moderation rules. Extremely niche topics that ChatGPT probably lacks enough modeled data for.

In essence, what forums of yore used to be like anyways.

I can see forums with strict proof of humanity requirement becoming more popular in the future, perhaps with a nominal $1/month subscription fee and phone number requirement.

This is essentially my hope for generative AI. I'll flood the platform web with so much bs that tight knit small forums and user groups will flourish again.

So far the strategy seems to be: "get a lot of hype, get some people invest in the hype, the topic of the hype doesn't matter as long as it includes $THING_OF_THE_DAY".

Disappointed but not surprised that it suggested the most milquetoast idea possible: affiliate marketing.

I think for LLM it's the design goal to return the most milquetoast (i.e. the most probable, thus the least surprising) thing matching the prompt.

This reminds me of that time where someone made a website and started selling real estate by the pixel.

It went up virally, made 1M in revenues.

It's interesting that the AI went straight in the direction of making a phony website as a venture.

That is super exciting and frightening at the same time.

The only way to make Etsy more useless

frightening

lybad business plan

Ok great plan …

Reply on news.ycombinator.com