Artificial intelligence uses like you to enter your head

What is it The future of the button is like in an era artificial intelligence? Max Livishin– The founder of Paypal and CEO of confirmation – a very new and very valuable role in admiring the data by training artificial intelligence on reaching more integral conclusions with these human decision -makers.

It is a well-known dilemma in machine learning that the computer that is presented with a clear reward function will participate in learning to enhance unnecessarily to improve its performance and increase this reward-but this course of improvement often leads to very different results, which may result from people who practice human rule.

To introduce a corrective force, artificial intelligence developers frequently use the so -called reinforcement learning from human comments (RLHF). Basically, they put a human thumb on the scale with the computer reaching its model by training it on data that reflects the actual preferences of real people. But where does this human preference data come from, and what is the necessary amount for the inputs to be valid? To date, this was the problem in RLHF: It is an expensive way if it requires employing human supervisors and explaining to enter the comments.

This is the problem that Levishen believes can be solved by the button like. Looking at the accumulated resource that is sitting today FacebookYour hands as a deity for any developer who wants to train a smart agent in human preference data. What is the size of the deal? Levishene told us: “I would like to claim that one of the most valuable things has a Facebook possessing is that the data mountain is likes.” In fact, at this turning point in developing artificial intelligence, and reaching “what content that humans love, to use it in training artificial intelligence models, it is probably one of the most valuable things on the Internet.”

While Livchin imagines Amnesty International’s learning of human preferences through a button like, artificial intelligence already changes the way these preferences are first formed. In fact, social media platforms are used with artificial intelligence activity not only to analyze likes, but to predict them – which makes the button itself old.

This was a fabricated note for us because, as we talked to most people, predictions often came from another angle, do not describe how the button that will affect the performance of artificial intelligence, but how Amnesty International will change the world of the button like. Indeed, we heard, Amnesty International is applied to improve social media algorithms. In early 2024, for example, try Facebook with him Using artificial intelligence To redesign the algorithm that recommends videos for users. Can it reach a better delivery of the variables to predict the video that the user wants to watch after that? The result of this early test showed that: artificial intelligence is applied to the task that was pushed at longer viewing times – the Facebook performance scale was hoping to increase.

When we asked YouTube He said: “I sometimes wonder if the button” that “there will be a need when artificial intelligence is sophisticated enough to tell the algorithm with 100 percent accurately, what you want to see after that, but it is possible that there will be more than any content platforms.

However, he continued to point out that one of the reasons that the button may always be needed is to deal with sharp or temporary changes in presenting the show needs due to life events or situations. He said: “There are days I want to see a little more relevant content, for example, my children.” Chen also explained that the button that suffers from age may have an longevity because of its role in attracting advertisers – the other main group alongside viewers and creators – because the like is the simplest possible hinges to connect these three groups. With one click, the viewer simultaneously transmits the appreciation and comments directly to the content provider and evidence of the participation and preference for the advertiser.