Artificial intelligence controls your search engine. This is a look under the cover

For decades, the way we find online information has changed only in small ways. The traditional Google research procedure today does not feel all that is different from when, in the nineties, it will ask Jeeves. Many certainly changed under the cover, the results are likely to be much more relevant and the interface has some new features, but you are still writing in the keywords and getting a list of web sites that may keep the answer.

This method of research appears to have started on the Altavista road, may rest in peace.

In May, Google announced its new launch Artificial intelligence mode For research, which uses a Obstetric artificial intelligence model (Based on the company twin A large language model) to give you a conversation answers that you feel more like a chat and less like combing a set of links. Other companies, such as Perplexity and Openai, have published search tools based on Gen AI. These tools, which combine Chatbot functions and a traditional search engine, are gaining steam quickly.

You cannot also escape One in five searches This type of summary now appears, according to the Pew Research Center report. I am surprised that it is nothing more.

these New search tools Feel a lot like model Chatbot, like ChatgptBut they do things a little different. These differences share a lot of DNA with the ancestors of the search engine. Here’s a look under the hood on how these new tools work, and how you can use it effectively.

Search engines for searching for artificial intelligence: What is the difference?

The basic technology of the search engine is similar to the old library card catalog. The engine uses robots to crawl the vast areas of the Internet to find an endless number of endless web pages. After that, when you search to ask about who played the role of Dr. Angela Hicks on ER, because you are trying to remember what I saw in it, he will return pages to things like ER crew or actor’s biography, Cch Pounder. From there, you can click on these pages, whether on Wikipedia, IMDB or elsewhere, and you know that you know Pounder Cch from the appearance of its Emmy Award guest in an X.

“When customers have a specific question, they can write this question in Google, and then Google runs their classification algorithms to find the best content for a specific inquiry,” told me Eugene Levin, head of the marketing tools company and SEO, Semrush.

In general, with traditional search, you should click on other websites to get the answer you are looking for. When I was trying to find out where I got to know Cch Pounder, I click at least half of a scale of different sites to follow. This included the use of a video search from Google- which releases videos index via various hosting platforms-to find clips of their appearance on X.

These multiple searches should not necessarily occur. If you just want to know the ER team, I can write “Cast of ER” and click Wikipedia page On top.

You will usually find Wikipedia or another confident site in or near the highest search results page. This is because the main method of search algorithms today is to track sites and pages that get most links from anywhere on the web. Sofik Das, associate professor at the Institute of Interaction between Human and Revenge at the University of Carnegie Mellon, said this model, which “changed the game for research” when it was launched Google in the 1990s, was more reliable than indexing systems that relied on things like the number of times the keyword appeared on the page, said Sofik Das, Associate Professor at the Institute of Interaction between Human and Revenge at Carnegie Mellon University.

“There are many recipes for cookies on the web, but how do you know which of them will appear first?” Das said. “Well, if a group of other websites link it to this site for the main words of” cookies recipe “, it is extremely difficult for the game.”

Self -powered search engines work slightly differently, but they work on the same basic infrastructure. In my endeavor to find out where I got to know Cch Pounder, I asked AI’s position from Google, literally, “Where do I get to know the actress that Dr. Angie Hicks plays on what?” In a conversation I felt chatting with a robot more than searches, which were narrowed. The first result gave me a list of shows and films that I did not see, so I asked for a broader list, which appeared on the appearance of her guest in other shows. Then I can order more details about the appearance of X files, and this is narrow.

Although the way I interacted with Google was different, the research mechanisms were basically the same. Ruby Stein, Vice President of Google Search and its treatment, told me that artificial intelligence mode has only used the Gemini to develop and process dozens of searches on the various web to collect the necessary information. “The user could inquire about each of these inquiries themselves.”

Basically, the act of artificial intelligence did the same thing it did, much faster.

A lot of searches, little time

The approach here is called the “query fan”. The artificial intelligence model takes your request and divides it into a series of questions, then searches to answer these demand components. Then he takes the information you collect from all searches and websites and put them together in an answer to you. In the heartbeat.

These searches use the same index that traditional research will do. “They are working on the same basis,” Levin said. “What changes are how to withdraw information from this basis.”

This exit process allows this fan to search artificial intelligence in relevant information from sites that may not appear on the first page of traditional search results, or withdraw a paragraph of good information from a page that contains more unrelated information. Instead of descending to a rabbit hole to find one small piece of the answer you want, artificial intelligence decreases a wide range of rabbits holes in a few seconds.

“You will expect, if you are looking for this, what is the next thing you might care about?” Levin said.

Read more: Amnesty International: 29 ways you can make the Gen AI operating for you, according to our experts

The number of searches that the artificial intelligence model will do is depends on the tool you use and the complexity of your question. Stein said that the artificial intelligence situation that uses the deep search of Google would spend more time and make more searches. Stein said: “If you are increasingly, if you ask a really difficult question, it will use our strongest models to respond,” Stein said.

Big language models that operate these search engines also have current training data to withdraw or use them to direct their searches. Although a lot of information comes from the updated content that he finds through searching in the web, some of which may come from these training data, which can include a set of information ranging from web sites such full libraries of books. Extensive training data to the extent that the lawsuits are about whether artificial intelligence companies are already The right to use this information It rapidly multiplies. (Disclosure: Zif Davis, the parent company CNET, filed a lawsuit against Openai, claimed that it had violated the copyright of ZifF Davis in training and operating artificial intelligence systems.)

Artificial Intelligence Research is not just chatbot

Not relying on training data is one of the things that puts an Amnesty International search engine regardless of the traditional Chatbot, although the basic language model may be as it is largely. Although the ChatgPT search will search for the Internet for relevant sites and answers, the regular Chatgpt may rely on its training data to answer your question.

“The correct answer may be there,” Das said. “It may also control a possible answer that is not anywhere in the pre -training data.”

Artificial intelligence research uses a concept called a generation to retrieve to integrate what he finds on the Internet in his answer. He collects information from the source of its direction to (in this case, the search engine index) and tells it to look there instead of making something if he cannot find it in his training data. “You tell artificial intelligence that the answer here, I just want you to find where,” Das said. “You get the best 10 Google results, and you tell artificial intelligence that the answer may be here.”

Can you really trust the results of artificial intelligence?

AI’s AI’s search tools may be more reliable than just using Chatbot itself, as they withdraw from the current relevant information and give you links, but you still have that Think critically about it. Here are some advice from experts:

Bring human doubts

Consider how bad people tell you when you are online. Then think about how bad the great language model is. This is how Amnesty International came from Google to the idea Put the glue on the pizza – By withdrawing information from the humorous Reddit post as if it were a real advice in cooking. “Artificial intelligence does not know what is authentic and what is the spirit of humor,” Das said. “All this information will be treated as it is.”

Remember to use your own judgment and search for sources of information. It may not be accurate as you think LLM, and do not want to make important decisions based on a person’s joke at an internet forum believed to be a real robot.

Amnesty International can still make things

Although they are supposed to withdraw from the search results, these tools can still make things in the absence of good information. This is how an artificial intelligence overview began to create Fake definitions of illogical sayings.

The generation of recovery may reduce an explicit danger Hallucinogenic But not eliminating it, according to DAS. Remember that LLM has no feeling what the correct answer is to the question. “It is just expecting the next English word that will come after this previous stream of other English words or other language words,” said Das. “She does not really have a concept of truth in this sense.”

Check your sources

Traditional search engines are very hands. They will give you a list of webs that seem related to search and allow you to decide if you want to trust them. Since the artificial intelligence research is uniformous and re -written by this information itself, it may not be clear when using an inaccurate source.

“These systems will not be completely mistaken, but I think the challenge is that over time you will lose the ability to capture them,” Levin said. “They will be very convincing and you will not know how to really go and check, or you think you don’t need to go and check.”

But you can check every source. But this is exactly the type of work that you hoped to avoid using this new system designed to save time and effort.

“The problem is if you are going to perform this analysis for every query you do in Chatgpt, what is the purpose of Chatgpt?” Levin said.