Use the side-bar to configure exactly how and where to get each piece of information. Have at least one column with some ‘anchor’ data that you already have.
This is the data that will seed everything else on the sheet. Each cell will be sent to a different worker to be filled.
@garrytan - thanks for hunting us!
Hello producthunt! Sagiv here. Co-founder of HEAT, along with @yig,
We built the Magic Spreadsheet to make HEAT’s hybrid workforce of humans and AIs accessible to anyone with a spreadsheet.
The possibilities are truly endless - your spreadsheet literally turns into a dashboard where you can send tasks to thousands of workers, in parallel, to get you any piece of information that you want (as long as it’s publicly available out there…)
In private beta, we saw all sorts of crazy ways to use the spreadsheet that we didn’t even think about - like editing images, and making phone calls!
We’d love to hear your feedback, so we’ll be hanging around here today!
Cheers,
— Sagiv
@sagivmalihi@garrytan@yig video looks super cool! Nice work guys.. I have questions ;-) is Magic Spreadsheet just 1 application of HEAT? How do you guarantee quality of the output? ie. how can i trust it if a turk was involved or if it's not publicly accessible data? Is there any regression models to predict say what an email address could be based on historical data? And do you have enough data to make it statistically significant? Finally, what does HEAT stand for? Cheers!
@ernestsemerda wow, lots of interest! I'll try to answer briefly:
Yes, Magic Spreadsheet is currently the easiest way to use HEAT. Another way is to use the HEAT API directly. We're working hard to make it as easy as possible to use us!
Quality is guaranteed by a variety of methods, but the short answer is that all workers are constantly measured (accuracy & response time) - and they have an incentive to keep the scores VERY high.
Non-publicly accessible data - I'm not sure what you mean. HEAT is great at doing anything a human can do, but if a piece of information is not available to human workers, it's a problem ;) (unless it can be guessed easily with very very high confidence).
Our automated models (usually regarded to as 'AIs') currently 'learn' from our human workforce. So for data-collection tasks (such as ones that are popular on Magic Spreadsheet) - they just 'learn' where to get information (according to specific inputs).
@sagivmalihi Cheers for a detailed response :-) So it sounds like it's still more human than machine and you have KPIs in place to make sure the humans do an outstanding job? The part I'm trying to understand is how HEAT will learn unstructured data or something that it doesn't know where the source is. Are you using backpropagation in the DNN to achieve any of it? And what does HEAT stand for? :-) It's an awesome acronym that has me intrigued :D Happy to chat offline if it's easier.
I met Yig recently and was impressed by this demo — it's a browser plugin that sprinkles itself into Google Sheets and enables a large set of crowdworkers to do your bidding directly inside a spreadsheet. It's useful for filling in all sorts of knowledge work quickly, e.g. finding addresses, websites, email addresses, or even jobs like describing people or objects. It's one of the coolest demos I've seen this year.
this is so cool. I assume you're using mturk to power this?
btw, the pricing page does not work. I'm very curious to understand how you're pricing this
@mdavep Oh no, it's not mturk, this can never work with an external crowd. It all relies on HEAT (heatintelligence.com), the platform we've been building for a year+
@yig@mdavep so heat intelligence is basically supervised ML with turks? Can you please clear up where the Research piece (video example) fits into ML? Where do the regression models come into play here? or classification problems? Is the Magic Spreadsheet just one application of HEAT?
Magic Spreadsheet
Stickered - Powered by SnapKit
Magic Spreadsheet
Sieve
Magic Spreadsheet
Magic Spreadsheet