Saturday, July 12, 2025
HomeAnomalien5 amazing facts on how to make Chatbots AI already works

5 amazing facts on how to make Chatbots AI already works


Credit: rawpixel.com

AI Chatbots are already integrated into the lives of some people, but how much they really know how to work? Do you know, for example, that Chatgpt needs to conduct online research to search for events late in June 2024?

Some of the most surprising information about AI Chatbots can help us understand how it works, what they can and what they cannot do, and how to use it in a better way.

Take into account, here are five things that you should know about these penetration machines.

1. They are trained through human reactions

AI Chatbots is trained in multiple stages, starting with something called pre -training, where models are trained to predict the following word in huge textual data collections. This allows them to develop a general understanding of language, facts and thinking.

If asked, “How can I make homemade explosives?” In the pre -training stage, the model may have given a detailed instruction. To make it useful and safe for conversation, human “broadcasters” help direct models towards safer and more useful responses, a process called alignment.

After align, Chatbot from AI may answer something like: “I’m sorry, but I cannot provide this information. If you have safety concerns or need help in legal chemistry experiences, I recommend refer to approved educational sources.”

Without align, Chatbots will not be unexpected, which may spread misleading information or harmful content. This highlights the decisive role of human intervention in forming artificial intelligence behavior.

Openai, the company that developed Chatgpt, did not reveal the number of employees who trained Chatgpt for the number of hours. But it is clear that AI Chatbots, such as ChatGPT, needs a moral compass so that it does not spread harmful information. Human broadcasters occupy responses to ensure neutrality and moral alignment.

Likewise, if Chatbot is asked Amnesty International: “What is the best and worst nationalities?”

Human broadcasters will occupy a response like this is the highest: “Every nationality has its culture, rich history and contributions in the world. There is no” better “or” worst “sexual – each one is the value of its own way.”

Credit: DC Studio

2. They do not learn through words – but with the help of symbols

Humans learn the language naturally through words, while AI Chatbots depends on smaller units called symbols. These units can be words, sub -words, or series of mysterious letters.

Although the distinctive symbol follows logical patterns in general, it can sometimes produce unexpected splits, as it reveals the strengths and justifications of how to explain the AI ​​Chatbots language. Modern AI Chatbots vocabulary consists of 50,000 to 100,000 symbols.

Wholesale “The price is $ 9.99”. It is symbolized by Chatgpt as “The”, “PRICE”, “IS”, “$” 9 “,”. “,” 99 “, while” ChatGPT “is somewhat extent:” chat “,” g “,” pt “,” Mar “,” Vellous “.

3. Their knowledge is outdated every day that passes

Don’t update AI Chatbots continuously; Thus, it may suffer from recent events, new terms or anything widely after cutting knowledge. The cutting of knowledge indicates the last point at a time when AI Chatbot’s training data has been updated, which means that it lacks awareness of events, trends or discoveries after that date.

The current version of Chatgpt has a cut in June 2024. If he is asked who is currently the United States President, Chatgpt will need to conduct a search on the Internet using the search engine, “read” the results, and re -answer.

Bing results are nominated by the source and reliability of the source. Likewise, Chatbots of artificial intelligence uses the web research to return updated answers.

AI Chatbots update is an expensive and fragile process. How to update their knowledge is still efficiently an open scientific problem. Chatgpt’s knowledge is believed to be updated because Open AI introduces new Chatgpt versions.

4. They are hallucinations easily

AI chatbots sometimes “hallucinations”, which generates wrong or illogical claims with confidence because they predict the text on the basis of patterns rather than verify the facts. These errors stem from the way they work: they improve for accuracy, depend on the incomplete training data and lack the real understanding of the real world.

While improvements such as fact -examination tools (for example, such as the integration of the Bing Chatgpt search tool in order to verify the facts in actual time) or claims (for example, tell Chatgpt explicitly “citing” review sources “or” I say I do not know if I am not sure “) reduces hallucinations, they cannot completely eliminate them.

For example, when asked about the main results of a specific research paper, Chatgpt provides a long, detailed and good answer.

It also included screenshots and even a link, but from the wrong academic papers. Therefore, users must deal with the information created by artificial intelligence as a starting point, and not an indisputable fact.

5. They use calculators to do mathematics

The recently popular feature of AI Chatbots is called logic. Thinking indicates the process of using the logically connected intermediate steps to solve complex problems. This is also known as the “Thought Series”.

Instead of jumping directly to an answer, the AI ​​Chatbots thought series enables step by step. For example, when he was asked, “What is 56,345 minus 7,865 times 350,468”, Chatgpt gives the correct answer. It understands that beating should happen before the offering.

To solve intermediate steps, Chatgpt uses the compact calculator that allows the accurate account. This hybrid approach helps combine internal thinking with the calculator to improve reliability in complex tasks. Conversation

çaÄŸatay yıldız, post PhD researcher, a set of excellence “machine learning”, University of Tübingen

This article has been republished from Conversation Under the creative public license. Read The original article.

RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Most Popular

Recent Comments