What is the future of the button similar to the era of artificial intelligence? Max Levchin—The co -founder of Paypal and CEO Affirm – consists of a new role extremely precious to love data to form AI to arrive at conclusions more in accordance with those that a human decision -maker would do.
It is a well -known dilemma in automatic learning that a computer presented with a clear reward function will engage in incessant strengthening learning to improve its performance and maximize this reward, but that this optimization path often leads AI systems to very different results from those which would result from humans exercising a human judgment.
To introduce corrective force, AI developers frequently use what is called learning to strengthen human feedback (RLHF). Essentially, they put a human thumb in scale as the computer arrives at its model by forming it on data reflecting the real preferences of real people. But where do these data come from a human preference and what part is necessary for the entry to be valid? So far, this has been the problem with RLHF: it is an expensive method if it requires hiring supervisors and human annotators to grasp comments.
And this is the problem which, according to Levchin, could be resolved by the similar button. He considers the accumulated resource in which today is today FacebookFarewell hands to any developer wishing to form an intelligent agent on human preferably data. And what is the size of an agreement? “I would say that one of the most precious things possessed by Facebook is that the mountain of data loves,” said Levchin. Indeed, at this inflection point in the development of artificial intelligence, having access to “what content is loved by humans, to be used for the formation of AI models, is probably one of the most precious things on the Internet.”
While Levchin envisages AI learning human preferences through the similar button, AI already changes the way in which these preferences are shaped first. In fact, social media platforms actively use AI not only to analyze the likes, but to predict them – potentially making the button itself obsolete.
It was a striking observation for us because, as we have spoken to most people, the predictions came mainly from another angle, not describing how the similar button would affect the performance of the AI, but how the AI would change the world of the similar button. Already, we have heard that AI is applied to improve social media algorithms. At the beginning of 2024, for example, Facebook experienced Use of AI To rethink the algorithm which recommends wrapped videos to users. Could this find a better weighting of variables to predict which video would a user like to watch the most then? The result of this early test has shown that he could: the application of the AI to the task paid in the longer watch moments – the Facebook performance metric hoped to increase.
When we asked YouTube The co-founder Steve Chen What the future contains for the similar button, he said: “I sometimes wonder if the similar button will be necessary when the AI is sufficiently sophisticated to say to the algorithm with 100% precision what you want to look at the simplest way for the content of the content platform, but the end of the objective is to make the easiest and the excitement of the content platform.
He then pointed out, however, that one of the reasons why the similar button may always be necessary is to manage net or temporary changes in vision needs due to life events or situations. “There are days when I want to look at content that is a little more relevant to, say, my children,” he said. Chen also explained that the similar button can have a longevity because of its role in the attraction of advertisers – the other key group alongside viewers and creators, because the same are the simplest hinge as possible to connect these three groups. With a tap, a spectator simultaneously transmits the assessment and comments directly to the content provider and evidence of commitment and preferably to the advertiser.
