[cblconfederate]

Web content makers have to realize their days of monetizing are numbered. Free web content is free food for NLP algorithms , which have already become impressive. In one of the next Google updates, they will eliminate web results compeltely and just give you the predicted answers. We're bound to see original content trying to hide themselves from google in order to remain relevant.

[eloff]

Except nobody wants to click on something produced by an NLP algorithm.

I don't doubt people will do that as they have every other shady SEO technique, but Google will continue to fight it as spam.

At the end of the day I think quality content is still king and the only viable long-term SEO strategy.

[Reedx]

A possibility: Parent comment generated by GPT-3

[nabilhat]

See also (comment on same parent): https://news.ycombinator.com/item?id=24146991

Will the ELIZAs learn to know each other somehow and throttle input, or will we someday approach the critical mass necessary for endless GPT-3 comment chains? Is there a similar term for Kessler Syndrome, but applied to comments?

[rchaud]

Aren't Siri, Alexa, Cortana, Google Assistant etc all providing answers generated by NLP algorithms? People seem to be using them pretty frequently.

Google already provides no-click search answers by scraping data from a website and bolding a paragraph that may answer your question. The branding of the original site is removed on these snippets.

How long before Google decides not to link out the website altogether?

[eloff]

I don't use those products, they only work for very specific types of queries. While I expect they will make it more broadly useful, I don't see it replacing content sites altogether.

[cblconfederate]

As NLP models keep increasing in size, they 'll reach the scale of the google index, and they give better answers. It's basically a huge associative memory that can produce responses it has and hasn't seen -it s definitely the evolution of searching.

Google has strategically bought every browser address box. It has eliminated the referrer , so websites don't know what search keywords led them there. It owns the whole process, and at this point , web content is just fodder for the next iterations of NLP models. It's already become cheap enough and it will become unmonetizable (unless it's opinion content, which is fodder for endless social media rambling).

Currently the most monetizable content is video. This might take longer to commoditize as deepfakes are still too early.

[idoh]

The point is that SEO itself is not a long term strategy.

[eloff]

I disagree with that point. People are not going to stop searching for things online, and while Google does try to capture some of that traffic and keep people on Google, it doesn't work, even when they get it right, except for the simplest of queries. Anything more involved and I always click through to read more of the source material. I don't think that will change either.

[idoh]

A - Agree that people will continue to search for things online.

B - Google does and will continue to capture that traffic, and it will continue to get better. Case in point, I read earlier this week an article about a site devoted to celebrity net worth and how their business was eaten up by the change.

C - Agree that more involved information is safe or safer.

But you know, it is obvious that Google will continue to eat up more and more queries. I don't know what percentage they are currently able to eat up and keep to themselves, but they are surely devoted to upping that rate. So as a long term play, do you really want to put yourself in competition with all the Google data scientists?

[eloff]

To the question raised in C, what's the alternative?

I think it's the best strategy. If you have a business, like the celebrity net worth one, you're in trouble. It's worth keeping in mind these days if your business has long term staying power based on the trends we know are currently underway.

[idoh]

It’s never been better to be in the content creation business. But it is a business, and one has to find ways of getting your customers to pay you money. There is substack, patreon, and probably lots of other ways to do it. But the perspective has to revolve around getting $X0 from thosands of customers rather than getting $0.50 per thosand impressions.

[paulcole]

"SEO" has worked for 20+ years. What SEO is changes but it will always be a viable strategy.

[paulcole]

I agree to an extent but there's more varieties of search than ever before. There's product search, local search, organic search, image search, etc.

You can absolutely "win" at some of those without kingly content.

[paulcole]

Lol, love the people downvoting me. I work in SEO and content.

[jermier]

It's as if the author writes content to /please/ Google. There are other players in town that will spread my message wide, other than Google. Google is a single point of failure too. If most of your traffic relies on a black-box algo developed by Google, at some stage you are going to be butthurt by that algo. Others will celebrate their success at gaming Google's algo and getting good rankings consistently, but these people are mostly blackhat SEOs probably trying to peddle cialis with a cheap discount.

[manquer]

> I don't write to please Google.

You can afford to do that if the website is not your primary source of income , not everyone has that luxury.

Online small and medium biz have to pay for their expenses , it is not for ad driven models either , organic traffic is significant source for people to purchase of your site. if your content is invisible to google for most sites that is a killing blow .

[hinkley]

Given how often I get top results that are clearly machine-generated content, I anticipate a lot of broken overfitting will happen in the near term.

If you train NLP on text generated by NLP, you're gonna have a bad time.

[colejohnson66]

Now I’m curious: what would happen if one repeatedly trained GPT-3 on text generated by a “previous generation” GPT-3? (Similar time successive JPEG saves)

[AnimalMuppet]

I'm pretty sure it would drift further and further from something a human would recognize as intelligible.

[datameta]

It would be a kind of DeepDream for NLP