“” Each AI for everyone “is somehow our slogan,” said Gupta. “We have organized all the AI models that we can find today.” Yupp’s website encourages developers to reach out if they want their language or image model to add to options. He does not currently have transactions with AI model manufacturers and provides these responses by making API calls.
Each time someone uses Yupp, he participates in a tête-à-t-head comparison of two chatbot models and sometimes gets a reward to provide his comments and choose a winning response. Basically, this is a survey on disguised users in a fun game. (The website has a lot emoji.)
He considers the compromise of data in this situation for users as more explicit than previous consumption applications, such as Twitter – which he does not tell me that he was the 27th employee and now has one of the co -founders of this company, Biz Stone, as one of its donors. “It’s a bit of a gap compared to previous consumer companies,” he says. “You provide feedback data, which data will be used anonymized and sent to model manufacturers.”
Which brings us to the place where real money is: selling human comments to AI companies that desperately want more data to refine their models.
“Externalized human assessments are what we do here,” says Gupta. He believes that the amount of species that users can make will be added enough for a few cups of coffee per month. However, this type of data labeling, often called learning to strengthen human feedback in the AI industry, is extremely precious for businesses because they publish iterative models and refine the results. This is much more than the most candle cup of coffee of all San Francisco.
The main competitor of Yupp is a website called Lmarena, which is very popular with IA insiders to obtain comments on new models and brass rights if a new version rises to the top of the pack. Whenever a powerful model is added to LMARENA, it often stimulates rumors on the big company that tries to test its new version in Stealth.
“This is a bilateral product with consumer network effects helping manufacturer models,” said GUPTA. “And model manufacturers, hopefully, improve models and submit them to consumers.” It shows me a beta version of the Yupp ranking, which is put online today and includes a global classification of the models alongside more granular data. The classification can be filtered by the way a model works with specific demographic information sharing users during the registration process, such as their age, or on a rapid category, such as questions related to health care.
Towards the end of our conversation, Gupta raises artificial general intelligence – the theory of superintelligent algorithms and as man – as an imminent technology. “These models are under construction for human users at the end of the day, at least in the near future,” he says. It is a fairly common belief and a point of marketing, among people working in AI societies, despite many researchers always wondering if the underlying technology behind large language models can ever produce Act.
GUPTA wants YUPP users, who could be worried about the future of humanity, to imagine by actively shaping these algorithms and improving their quality. “It’s better than free, because you do this great thing for the future of AI,” he says. “Now, some people would like to know, and others just want the best answers.”
And even more users could simply want additional money and be ready to spend a few hours giving comments during their chatbot conversations. I mean, $ 50 costs $ 50.
