Last week, the robotics company UR and Scale AI reached a cooperation agreement and released the UR AI Trainer. This also marks that Scale AI, a data annotation giant, is determined to enter the field of robotics (embodied intelligence). The reason why Scale AI is keen on expanding new businesses is well known, as the former giant's position is now somewhat subtle. Since the mid year of last year, when Scale AI was acquired by Meta with a 49% stake, Alexander Wang from Lianchuang has also joined Meta directly from Scale AI. This has caused Scale AI to lose a certain degree of neutrality and also led to some customers moving away. For example, Google, but since Scale AI became a semi subsidiary of Meta, Google stopped working with it and instead began collaborating with institutions such as Surge AI. For example, in the past, OpenAI and Scale AI had a deep collaboration, and OpenAI had to buy a large amount of annotated data from Scale AI. However, after Meta's semi acquisition, OpenAI also significantly reduced the weight of Scale AI as a data annotation supplier. According to market speculation, the revenue of Surge AI may even be no less than that of Scale AI. Surge AI is led by Edwin Chen, but Chen is much more low-key than Wang. Actually, data annotation is a very profitable business, and it is labor-intensive. Those data annotators from India, the Philippines, or Kenya annotate images and text, earning money for each task, giving off the feeling of delivering takeout or driving ride hailing services. This is particularly reminiscent of Crypto, as they are all about doing one job and getting one incentive. In fact, after all these years of Crypto, I dare not say anything else, at least I have made the incentive issue quite clear 。 In fact, Crypto also has data annotation projects such as Sahara. Recently, there has been another project that has been quite popular. The coin has not been released yet, but it has already been included in Coinbase's listing roadmap. This project is called Perle @ PerleLabs. Coincidentally, the founder was the former head of growth at Scale AI. His name is Ahmed Rashad, which sounds like a Middle Eastern. After all, Scale's current valuation has reached an astonishing $29 billion. So, as the former executive of Scale AI, founder Rashad naturally gained the favor of investors for Perle, which raised $17.5 million in the cryptocurrency industry. Of course, to be honest, despite raising so much money, it is still difficult for Perle to directly challenge its former owner Scale AI at this stage. Scale AI is taking a big and comprehensive data route, with whatever data is needed, and even robot data is now being prepared. What is Perle's idea? Rashad had probably seen too many data annotators from India and the Philippines before when he was the Director of Scale AI Growth. He thinks that this click farm like working state may be a bit problematic, so he now wants to take the "high-precision and cutting-edge route". Rashad, on the other hand, hopes that Perle can focus on data annotation tracks with relatively high unit prices, such as law, medicine, finance, and so on. Then, combined with the incentive effect of Crypto, we hope to attract more expert level annotators, rather than labor level annotators like data annotators, to join. Of course, making high-precision and cutting-edge products is not a market gap. For example, Surge AI has a lot of high-precision and cutting-edge data annotation work. Meanwhile, due to size limitations, top tier companies like Google may not be able to collaborate with small factories like Perle at this stage due to their very strict precision requirements. So what is Perle targeting more? It's the kind of second or third tier AI company. They urgently need to differentiate and be more cost sensitive. On the other hand, there are many projects that involve fine-tuning large models and LoRA, and they also covet these high-precision data.