Tell HN: Seeing an HN ad for YC company that portends AI-powered blogspam
I just saw the following ad on HN.
> Contentedge (YC W17) is hiring to build AI content writing software for SEO
From its landing page there's the following sentence:
> Instantly create briefs or text that’s modeled after high-ranking content and sounds human-made.
This kind of service seems specifically designed to drown the Internet in blogspam, even more so than it already is, and hence worries me.
I know it was inevitable given the rise of convincing AI text generation and sure if one company doesn't do it another will (and no doubt there are already those that do), but please Contentedge, reconsider if this is truly the product you want to create for this world.
26 comments
[ 3.6 ms ] story [ 60.0 ms ] threadIt seems like you equate SEO with blogspam, which is just false!
We help businesses create more useful content and get it to the right people. Google is the gateway to the internet. I don’t view it as gaming the system; we’re trying to empower smaller businesses and writers to get found.
Give Contentedge a try and see if we're just building a spam machine like you contend! (no credit card required) https://secure.contentedge.com/users/signup/free
This comment does nothing to convince me that it isn't the subset you're targeting, because that last sentence sounds a lot like you're describing blogspam.
(parent comment was edited after I wrote this, read "last sentence" as referring to the paragraph beginning with "We help businesses create more useful content and get it to the right people.")
What's the intended usecase for creating AI content other than blogspam-style bait and switch? Shouldn't these businesses already have content of their own?
It would make more sense to me if your product was aimed at helping websites SEO-optimize the content they already have, as I can't see how a human would ever find your AI-written content useful.
I’m frankly shocked you’re YC funded…. I’m guessing a pivot from what they originally funded you for
If you were to read some content that's useful to you, does it matter the source? If the content generated by this platform is just "spam", then wouldn't it fail on its own without the need for an HN brigade against them?
I get that google is increasingly being degraded by SEO content, but that's a problem with google, not this site allowing users to use AI generated content as a writing tool.
Nonetheless there are degrees to this and pushing a set of features designed to automatically generate text that closely mimics what other high-ranking content looks like has an outsize impact on the quality of the "commons": in this case Google or any other index of internet content. As other commentators have pointed out, the current state of AI means that while the style of the text is likely to be in line with other content, the actual content of the text may be either useless or actively wrong.
Given the prevalence of such articles already in search results, it is unlikely that the normal mechanism of human perusal of search results will be an effective quality filter.
It is true that, as I allude to, such articles already exist and in this sense Contentedge is not enabling a novel new harm per se. However, it drastically changes the ease of doing so, and doing the same thing, just at a larger scale and higher efficiency, given sufficiently big values of "larger" and "higher" can have a qualitative impact on the harm rendered.
What does success mean? Let's assume you mean "Profit from visits generated by SEO"
> If you were to read some content that's useful to you, does it matter the source?
I am doubtful about there being any intersection in a venn diagram between
* Content that is useful to me
* Successful at extracting money through SEO
* AI generated
> I get that google is increasingly being degraded by SEO content, but that's a problem with google, not this site allowing users to use AI generated content as a writing tool.
This is disingenuous. There is no way we can believe that AI generate content will be as accurate and useful as human generated content. We want the quality human generated content. Until we get to AIs that are as intelligent as human beings, and share our context.
Using AI as a tool where the human edits it is probably OK. But blasting out AI-generated articles with no review is going to be horrible. And using Google as justification - well Googles weakness is what makes this idea profitable!
yes, very much. the source of the content is half of its value.
there is a big difference if i read a physics paper from richard feynman vs one from a high school student or one generated with AI.
i expect the feynman paper to be thoroughly researched. i expect the high schoolers paper to be a collection of interesting facts that the student learned and is excited about and i expect the AI paper to be a random collection of stuff found on the internet.
the problem is that it has already been proven that it is possible to write good sounding nonsense and have that published as scientific research. therefore the only thing that makes me trust any text is the reputation of the author.
an AI generated text can't possibly be useful because it is inherently untrustworthy. at best it can function as entertainment, but i believe we already have enough mediocre entertainment, we don't need AI to generate any more and drown out the less common better quality work out there.
This seems to be an issue with how science publishing works, how things are 'peer reviewed', etc. If things were _truly_ peer reviewed, would the peers not catch nonsense in papers? And if it was not caught by a true review, then maybe it's not nonsense after all.
> an AI generated text can't possibly be useful because it is inherently untrustworthy
Probably 25% of the code I 'write' these days has been written by AI, through Copilot. Copilot isn't perfect but it can create a basis to start from that saves me lots of time. This is how I view these content writing tools, something to get the ball rolling, but not something you would use to generate all the content on your site without editing.
The problem I have with this is that it presumes knowledge on the part of the reader.
The AI will frequently produce wrong content that I incorrectly believe was valuable to me.
The AI will frequently omit important things an actual human expert would not have.
It will become increasingly difficult to tell the difference between trustworthy and non-trustworthy content.
Most those sites trained with GPT-3 model.
"reconsider if this is truly the product you want to create for this world" If they were self-serving and crass enough to start the idea, they'll only 'reconsider' if it fails an A-round.
I guess I shouldn't be surprised that YC actually gives money to a company like that.
But hey, potential future returns for me are better than making everyone else's experience on the internet worse, right?