“” Every artificial intelligence for everyone “is a kind of our slogan,” says Gupta. “We organized all AI models that we can find today.” The YUPP website encourages programmers to reach if they want to add their language or image model. Currently, there are no contracts with the creation of AI models and provides these answers to create API calls.
Every time someone uses YUPP, they participate in comparison with two Chatbot models, and sometimes receives a reward for transferring opinions and winning the winning answer. Basically, this is a study of users hidden for a entertaining game. (The site has nave emoji.)
He sees the data in this situation to users as more pronounced than previous consumer applications, such as Twitter – which quickly tells me that he was the 27th employee, and now he has one of the co -founders of this company, Biz Stone, as one of his sponsors. “It’s a bit of a departure from previous consumer companies,” he says. “You provide feedback that the data will be used in an anonymous manner and sent to builder models.”
Which leads us to the place where real money is located: selling human feedback to AI, which desperately want more data to customize their models.
“Human assessments are what we do here,” says Gupta. He estimates that the amount of cash that users can earn can be added enough for several cups of coffee per month. Although this type of data label, often called Learning to strengthen with human feedback In the artificial intelligence industry, it is extremely valuable for companies that release iterative models and adapt the results. In all of San, Francisco is worth much more than the most significant cup of coffee.
The main competitor of YUpp is the website called Lmanwhich is quite popular among people from AI to obtain feedback on up-to-date models and bragging if the up-to-date version increases to the top of the package. Whenever a powerful model is added to Lmarena, it often fuels rumors about which main company is trying to test its up-to-date version in Stealth.
“It is a two -sided product with consumers’ network effects that help model builders,” says Gupta. “We hope that models designers improve models and fold them back to consumers.” It shows me the version of the YUPP leaders table, which is live today and includes a general model ranking with more detailed data. Rankings can be filtered according to how well the model works with specific demographic information, which users provide during the registration process, such as their age or in a specific brisk category, such as questions related to healthcare.
At the end of our conversation, Gupta is raising artificial general intelligence-superintelligent-like algorithms similar to man-as inevitable technology. “These models are built for people at the end of the day, at least in the near future,” he says. This is a fairly common belief and marketing point among people working in AI companies, although many researchers are still questioning whether the technology underlying enormous language models will be able to produce Agi.
Gupta wants YUPP users who can worry about the future of humanity to imagine that they actively shape these algorithms and improve their quality. “It’s better than free because you do great things in the future AI,” he says. “Now some people would like to know it, and others just want the best answers.”
And even more users can simply want additional cash and they willingly spent a few hours, providing feedback during Chatbot’s talks. I mean that 50 USD is $ 50.