Artificial Intelligence is changing SEO faster than you think.
Today everyone has already heard about Google's RankBrain - a new system of artificial intelligence (AI) based on machine learning, which is considered to be the most up-to-date and most outstanding development from Mountain View (California). It’s not so obvious how fast the SEO industry is changing because of this. In this article, I will use simple examples to show that some of the old SEO rules no longer apply, and what steps you can take to avoid flying a bend and continue to conduct successful SEO campaigns for your business.
Before we start predicting, let's see how the SEO market has already changed. My friend, Carnegie Mellon University (USA) graduate Scott Stouffer, technical director and co-founder of Market Brew, dedicated to search engines for the Fortune 500 SEO team, helped me with this. Here are some of his tips for the SEO industry in the context of Google’s artificial intelligence.
Inside Google, there are a number of algorithms. RankBrain’s job is to decide which complex of these algorithms is best applied to each type of search result. For example, in some search results, RankBrain may decide that the most important is META Title. But in another search result, the same signal can have a terrible correlation with a good search result.
From this it becomes obvious that it no longer makes sense to do a regression analysis of the site, without having the context of the search result.
For these reasons, today a regression analysis has to be done for each specific search result. At the same time, Google’s algorithmic shifts can be measured. First, you can record the current situation, as well as each calibration by keyword. Then re-calibrate each time after detecting a shift in the rating, identify the delta between the two parameters. Using this approach, you can discover which particular algorithm prioritizes RankBrain for each individual context, raising or lowering its value in each search results output. By understanding this, we can focus on improving these specific parts of SEO for sites for these unique search results.
In addition, Google realized that they could teach RankBrain to distinguish between “good” and “bad” sites. Just as it determines the weights of existing search algorithms differently for each search result, they also realized that in each such vertical there are different examples of “good” and “bad” sites. This is undoubtedly because different verticals have different CRM, different templates and different data structures in general. Evaluation of sites takes place by comparing them with the best and worst sites. A simple example: if you work in health care and your site is similar to the authoritative WebMD.com, then you will be recognized as a “good” site, if he notices you have something in common with spamming sites - you will get a “bad” stigma.
Now think what your chances of success are, if your task is to explain RankBrain that a site that, for example, is dedicated to many different industries at the same time is “good”?
So what happens to sites that have many different categories? A good example of these types of sites are How-To sites. In these cases, the algorithm breaks down. And what does Google do? Answer: Selects randomly one or another category from all submitted for ranking. For well-known sites like Wikipedia, Google makes exceptions to ensure that the new algorithm does not destroy the existing search experience (but this is only for those who are “too big to collapse”).
But what will happen for less known sites? Who knows? Apparently, the algorithm has an automated way to classify each site before trying to compare it with other sites in the same category.
The conclusion is simple: for sites like How-To, it makes sense to create separate domains for each direction and this will help him improve his position in each category from this industry. Focus on one niche.
Let's see how the new algorithm affects the backlinks. Now more than ever, it is important to engage in relink. This is due to the fact that RankBrain to determine the result in the issue will compare the backlinks of your site with the backlinks of “good” and “bad” sites - that is, it is one of the comparison parameters along with the structure, etc. Perelinkovka with friendly but irrelevant resources threatens you have the same problem as an excessive number of categories on the site - by randomly choosing a ranking algorithm, comparing it with other categories of sites.
Future SEO and AI
No one knows exactly how the AI ​​will improve and where it will lead us, but something can still be argued for sure today:
On the one hand, the method of depth learning has simplified the work of optimizers. Now, knowing that there is RankBrain and similar technologies that work almost on the same level with a person, one thing is clear: there are no loopholes anymore.
On the other hand, things are a little more complicated. The SEO field will continue to become highly technical. Analytics and big data are on the agenda, and any SEO specialists who are not familiar with this will have to catch up. But, those of you who have these skills can count on large salaries.
So what is AI?
There are at least three types of AI:
1. Artificial Narrow Intelligence (ANI) is an AI created for one specific task (for example, defeating a world chess champion)
2. Artificial General Intelligence (AGI) is an AI capable of performing many tasks (comparable to human capabilities)
3. Artificial Superintelligence (ASI) is an AI capable of performing many tasks at the highest level (exceeding human capabilities)
When we talk about Google's RankBrain, and other machine learning algorithms that currently work at Google, we are obviously talking about ANI.
Similarly, RankBrain Google, ANI are those spam filters in our mail, my favorite Google Translate and IBM's Watson, and the Amazon feature that recommends us buying, and unmanned vehicles.
')
Inside ANI, there are a lot of different approaches. Pedro Domingos, in his book The High Algorithm, divided all scientists involved in the problem of improving AI into several groups:
RankBrain made in the camp of communicators. They believe that all our knowledge is encoded in the connections of the neurons of our brain.
RankBrain's particular strategy is that experts call the back-propagation method of the error, and after rebranding they began to call “deep learning”
Communicators claim this strategy is able to learn everything from raw data, and, ultimately, automate the search for new knowledge. Google, apparently, also believes: On January 26, 2014, Google announced the acquisition of DeepMind Technologies, which was essentially based on the backpropagation method.
Summarizing, we can say that RankBrain is an artificial intelligence created for a narrow task based on the “depth learning” technology. Having dealt with this, let's see how far they have come and, more importantly, how will this change the SEO business?
We tend to underestimate the future - all because in our assessments we wrestle with past historical experience. However, the reality is that progress accelerates over time. Ray Kurzweil calls this the Law of faster returns. The scientific rationale for his theory is that more developed societies have the opportunity to progress more rapidly than less developed ones because they are more advanced. Of course, the same conclusions can be applied to the field of artificial intelligence.
Here is a visualization of the Law of faster returns:
As you can see, and indeed feel intuitively, the improvement of the methods of deep processing and computers follows from the Law of faster returns. Here is another shocking revelation: at some point, the processor power will surpass not only one person, but all the people together.
In fact, it now seems that we can create multitasking artificial intelligence (AGI) by 2025. Technologies are clearly growing at a rapid pace, and, according to many testimonies, most of us will be caught off guard.
As I explained above, RankBrain is just one of the forms of ANI, which means that it can perform tasks better than humans in one particular area, which means it is a relatively weak form of artificial intelligence.
But soon we may be stunned by how quickly this “weak” artificial intelligence can easily turn into something with which we have no idea how to deal.
Here you can clearly see that Google's RankBrain, being super smart in one particular task, is still in the general context of things, rather weak in the general sense of intelligence.
But what happens when we apply the same Law of faster return to the field of artificial intelligence? Here are the conclusions of such a thought experiment by Tim Urban:
“... So, watching the development of AI, we will soon see how easy it is for him to become smarter than an animal. Then, when he reaches the minimum level of human intelligence - let's say a 'village idiot' - we will exclaim: “Oh wow, this is so silly. Pretty little. ” Only here the range of lacking intelligence between the fool and Einstein is not that great - so very soon after the “fool” we will hear about the creation of AGI, which is even smarter than Einstein “.
1. Artificial Narrow Intelligence (ANI) is an AI created for one specific task (for example, defeating a world chess champion)
2. Artificial General Intelligence (AGI) is an AI capable of performing many tasks (comparable to human capabilities)
3. Artificial Superintelligence (ASI) is an AI capable of performing many tasks at the highest level (exceeding human capabilities)
When we talk about Google's RankBrain, and other machine learning algorithms that currently work at Google, we are obviously talking about ANI.
Similarly, RankBrain Google, ANI are those spam filters in our mail, my favorite Google Translate and IBM's Watson, and the Amazon feature that recommends us buying, and unmanned vehicles.
')
Inside ANI, there are a lot of different approaches. Pedro Domingos, in his book The High Algorithm, divided all scientists involved in the problem of improving AI into several groups:
- Symbolists
- communicators (Connectionists)
- Evolutionaries
- Baesians
- analyzers (Analogizers)
RankBrain made in the camp of communicators. They believe that all our knowledge is encoded in the connections of the neurons of our brain.
RankBrain's particular strategy is that experts call the back-propagation method of the error, and after rebranding they began to call “deep learning”
Communicators claim this strategy is able to learn everything from raw data, and, ultimately, automate the search for new knowledge. Google, apparently, also believes: On January 26, 2014, Google announced the acquisition of DeepMind Technologies, which was essentially based on the backpropagation method.
Summarizing, we can say that RankBrain is an artificial intelligence created for a narrow task based on the “depth learning” technology. Having dealt with this, let's see how far they have come and, more importantly, how will this change the SEO business?
The exponential growth of technology (and AI)
We tend to underestimate the future - all because in our assessments we wrestle with past historical experience. However, the reality is that progress accelerates over time. Ray Kurzweil calls this the Law of faster returns. The scientific rationale for his theory is that more developed societies have the opportunity to progress more rapidly than less developed ones because they are more advanced. Of course, the same conclusions can be applied to the field of artificial intelligence.
Here is a visualization of the Law of faster returns:
As you can see, and indeed feel intuitively, the improvement of the methods of deep processing and computers follows from the Law of faster returns. Here is another shocking revelation: at some point, the processor power will surpass not only one person, but all the people together.
In fact, it now seems that we can create multitasking artificial intelligence (AGI) by 2025. Technologies are clearly growing at a rapid pace, and, according to many testimonies, most of us will be caught off guard.
Superinteller Growth
As I explained above, RankBrain is just one of the forms of ANI, which means that it can perform tasks better than humans in one particular area, which means it is a relatively weak form of artificial intelligence.
But soon we may be stunned by how quickly this “weak” artificial intelligence can easily turn into something with which we have no idea how to deal.
Here you can clearly see that Google's RankBrain, being super smart in one particular task, is still in the general context of things, rather weak in the general sense of intelligence.
But what happens when we apply the same Law of faster return to the field of artificial intelligence? Here are the conclusions of such a thought experiment by Tim Urban:
“... So, watching the development of AI, we will soon see how easy it is for him to become smarter than an animal. Then, when he reaches the minimum level of human intelligence - let's say a 'village idiot' - we will exclaim: “Oh wow, this is so silly. Pretty little. ” Only here the range of lacking intelligence between the fool and Einstein is not that great - so very soon after the “fool” we will hear about the creation of AGI, which is even smarter than Einstein “.
SEO has changed forever
Before we start predicting, let's see how the SEO market has already changed. My friend, Carnegie Mellon University (USA) graduate Scott Stouffer, technical director and co-founder of Market Brew, dedicated to search engines for the Fortune 500 SEO team, helped me with this. Here are some of his tips for the SEO industry in the context of Google’s artificial intelligence.
Today's regression analysis has serious drawbacks.
Inside Google, there are a number of algorithms. RankBrain’s job is to decide which complex of these algorithms is best applied to each type of search result. For example, in some search results, RankBrain may decide that the most important is META Title. But in another search result, the same signal can have a terrible correlation with a good search result.
From this it becomes obvious that it no longer makes sense to do a regression analysis of the site, without having the context of the search result.
For these reasons, today a regression analysis has to be done for each specific search result. At the same time, Google’s algorithmic shifts can be measured. First, you can record the current situation, as well as each calibration by keyword. Then re-calibrate each time after detecting a shift in the rating, identify the delta between the two parameters. Using this approach, you can discover which particular algorithm prioritizes RankBrain for each individual context, raising or lowering its value in each search results output. By understanding this, we can focus on improving these specific parts of SEO for sites for these unique search results.
Focus on your niche clearly to avoid misclassification.
In addition, Google realized that they could teach RankBrain to distinguish between “good” and “bad” sites. Just as it determines the weights of existing search algorithms differently for each search result, they also realized that in each such vertical there are different examples of “good” and “bad” sites. This is undoubtedly because different verticals have different CRM, different templates and different data structures in general. Evaluation of sites takes place by comparing them with the best and worst sites. A simple example: if you work in health care and your site is similar to the authoritative WebMD.com, then you will be recognized as a “good” site, if he notices you have something in common with spamming sites - you will get a “bad” stigma.
Now think what your chances of success are, if your task is to explain RankBrain that a site that, for example, is dedicated to many different industries at the same time is “good”?
So what happens to sites that have many different categories? A good example of these types of sites are How-To sites. In these cases, the algorithm breaks down. And what does Google do? Answer: Selects randomly one or another category from all submitted for ranking. For well-known sites like Wikipedia, Google makes exceptions to ensure that the new algorithm does not destroy the existing search experience (but this is only for those who are “too big to collapse”).
But what will happen for less known sites? Who knows? Apparently, the algorithm has an automated way to classify each site before trying to compare it with other sites in the same category.
The conclusion is simple: for sites like How-To, it makes sense to create separate domains for each direction and this will help him improve his position in each category from this industry. Focus on one niche.
Backlinks smelt foul
Let's see how the new algorithm affects the backlinks. Now more than ever, it is important to engage in relink. This is due to the fact that RankBrain to determine the result in the issue will compare the backlinks of your site with the backlinks of “good” and “bad” sites - that is, it is one of the comparison parameters along with the structure, etc. Perelinkovka with friendly but irrelevant resources threatens you have the same problem as an excessive number of categories on the site - by randomly choosing a ranking algorithm, comparing it with other categories of sites.
Future SEO and AI
No one knows exactly how the AI ​​will improve and where it will lead us, but something can still be argued for sure today:
- The competitive environment of each keyword must be studied separately;
- Most sites will need to be niche to avoid misclassification;
- Each site should mimic the structure and content of search results leaders in this niche.
On the one hand, the method of depth learning has simplified the work of optimizers. Now, knowing that there is RankBrain and similar technologies that work almost on the same level with a person, one thing is clear: there are no loopholes anymore.
On the other hand, things are a little more complicated. The SEO field will continue to become highly technical. Analytics and big data are on the agenda, and any SEO specialists who are not familiar with this will have to catch up. But, those of you who have these skills can count on large salaries.
Source: https://habr.com/ru/post/302626/
All Articles