This looks interesting. But I was hoping the website would explain more about how it works. How do you hire qualified taskers? How do you ensure you have enough of them? Are you paying them well and are you complying to employment laws (I don't to participate in immoral/illegal setups)? What are good instructions and what are not? How much creativity can I expect from a tasker (some tasks, such as prosecting, require a bit more thinking and creativity if you want accurate results)? Some taks require a bit of domain knowledge (e.g. knowledge of my own products), how do I ensure good results? These are just some of the questions that immediately pop up to me. I was disappointed by the lack of an FAQ section.
I went through their recruitment process last fall for a summer internship, and I can confirm: Their developer-focused approach is definitely what make them stand out.
There is a reason Stripe succeeded where many (many) other payment companies tried and failed.
Combining great UI/UX and a developer-centric experience can go a long way into ensuring your business is future-proof, especially when your core service is an API :).
I noticed you show a list of top earners...humans doing the tasks. Just monthly earnings shown though. Are the leaderboard people working an abnormal amount of hours, or are they just doing well due to accuracy?
As far as I know, there isn't a consistent one. The Wall Street Journal recently increased the level of protection their paywall provides. Some have reported success using the "web" link in an private browsing window.
Edit to add: I had forgotten about the hack my sibling mentioned. Thanks, 'ayw!
Right now we're very focused on image, audio, and video applications. In the near future, it will include transcription, captions, object detection, and feature detection. We plan on adding features around NLP and language applications as well, since that's an area with a ton of interesting work.
In general, we hope to tackle anything that humans are better than machines at and would make sense as APIs.
I'm Alex, CEO of Scale. We definitely hope to solve many of the same problems as MTurk, and think it's a brilliant idea.
Our main focuses for our product are around high quality, ease-of-use, and reliability. In our experience, MTurk exhibits none of those.
It requires a lot of work to get truly high quality results, and virtually impossible above a certain quality threshold. MTurk promises no SLAs and forces the customer to do a lot of work such as price discovery, building UIs, and quality control, not to mention how difficult to use their API is. Our philosophy is handle all of that complexity for the developer.
We've been inspired by our own experiences using easy-to-use APIs like Stripe, Twilio, SendGrid, etc., and wanted to build a similar experience for human intelligence. Hope that helps!
We're approaching the problem differently than both of those companies, and over time I think the difference will be more and more apparent.
Scale is vertically focused, currently focusing on image annotations and audio transcription. This allows us to provide extremely high accuracy and an ever-improving developer experience by investing heavily in tooling, automation, and screening. It's very difficult to do that as a more generic crowdsourcing service.
We're also entirely developer-focused as a company. Our users are engineers, and we have a singular focus to provide the best experience in using Scale to build more complex products.
