TL;DR
-
Masa, a decentralized AI network, is launching a ‘contribute-data-to-earn’ model — where you download an app, share your social data, and earn token rewards in return.
Full Story
So, you’re probably familiar with the concept of decentralized compute power…
You let folks access your computer’s processing capabilities → you earn crypto in return.
The concept has gained a lot momentum ever since the AI boom started — cause Large Language Models (LLMs) like ChatGPT require a whooole bunch of compute power to run.
(Enough to make crowd sourced compute power make sense).
The other side of the AI equation — the side that’s talked about most, but rarely paid for — is the data used to train these models. Right now, most LLMs are being trained for free on the internet’s collective content archive.
Masa, a decentralized AI compute network, is launching a ‘contribute-data-to-earn’ model — where you:
Which is very cool! But also brings two big questions to mind:
-
If AI companies can already get this data for free, why would they all of a sudden start paying for them?
-
What data are folks incentivized to contribute (and what makes it so valuable)?
Well, the Masa app isn’t live yet — but we do have a press release that sheds some light on the concept, and helps answer those questions.
First, what we’ve stitched together from the press release:
There’s a huge amount of online data that is walled off, and those walls are getting higher every day…
Now that these LLMs (and the training data that powers them) are becoming so darn valuable, platforms like Facebook, Instagram, X, etc. are incentivized to block AI companies from scraping their data, and instead use it to build their own LLMs (e.g. Facebook’s Llama, and X’s Grok).
Masa finds a side door into these walled gardens, by incentivizing users to grant access to their social accounts (X, Discord, YouTube, TikTok, etc.), and — if they choose to — profit by selling this access to third parties.
Second, a wild guess at the future:
The internet is a mess of information, which contributes to AI ‘hallucinations’ (aka making stuff up), like this.
At some point, someone’s going to make a marketplace for vetted data on a range of topics, so that reliable use-case-specific AI models can be spun up.
This could be a first run at such a concept.
Our takeaway:
Being able to profit from our personal data, instead of just giving it away for free?
That’s a nice option to have.
