“The entire idea of captchas was that humans are better at solving these puzzles than computers. We’re learning that’s not true.”
This is one of the biggest misunderstandings of Google's Captcha system. Yes, it's a bot detection system for consumers, but for Google it's a bot training system.
The first generation of Captcha's trained OCR systems. The current version trains image recognition and self-driving car systems (that's why you're often asked about traffic lights, motorbikes, bridges, signs etc. It's all data from Google's vehicles and partners, and you're tagging it).
Google takes aggregate data, and once they've got statistically meaningful answers for each frame of video they finalise that dataset and it goes off to be used by bots online and offline for a variety of image recognition purposes. Meaning it is difficult to submit false answers and ruin their data, since they can wait for a 99% consensus per image across millions of captcha submissions. This is also why you can think you've nailed a Captcha and you still get a 2nd one, Google has no idea what the right answer is, it's just comparing your answer to other people's answers until you reach a threshold of majority matching answers.
We're the product/service, not ReCaptcha v2. It's not going anywhere until we figure out that Google should be paying for our labour in training their models.
This is one of the biggest misunderstandings of Google's Captcha system. Yes, it's a bot detection system for consumers, but for Google it's a bot training system.
The first generation of Captcha's trained OCR systems. The current version trains image recognition and self-driving car systems (that's why you're often asked about traffic lights, motorbikes, bridges, signs etc. It's all data from Google's vehicles and partners, and you're tagging it).
Google takes aggregate data, and once they've got statistically meaningful answers for each frame of video they finalise that dataset and it goes off to be used by bots online and offline for a variety of image recognition purposes. Meaning it is difficult to submit false answers and ruin their data, since they can wait for a 99% consensus per image across millions of captcha submissions. This is also why you can think you've nailed a Captcha and you still get a 2nd one, Google has no idea what the right answer is, it's just comparing your answer to other people's answers until you reach a threshold of majority matching answers.
We're the product/service, not ReCaptcha v2. It's not going anywhere until we figure out that Google should be paying for our labour in training their models.