Briefly
Sahara AI launched a platform that pays customers in crypto to label AI coaching information.
It makes use of a bounty mannequin with $450K+ in token rewards at launch.
Sahara emphasizes high quality management and shared information possession.
On the lookout for a technique to earn crypto with out writing code or buying and selling tokens? Attempt labeling information for AI.
Los Angeles-based Sahara AI debuted its Knowledge Providers Platform (DSP) as we speak, a brand new system that pays customers in cryptocurrency for finishing small however important data-labeling duties—tagging photos, transcribing audio, or evaluating AI-generated textual content.
The platform goals to show the grueling, typically underpaid work of coaching AI techniques right into a decentralized gig financial system.
“We have been operating some closed beta model of it prior to now few months, and now we’re shifting into this open model, the place everybody can come to our platform to verify on the information duties which are printed by our ecosystem companions,” Sahara AI co-founder and CEO, Sean Ren, informed Decrypt.
The platform is modeled on bug bounties the place corporations, analysis labs, and crypto initiatives publish information duties that contributors full in change for rewards. As an alternative of discovering software program bugs, although, DSP customers receives a commission to annotate information. In keeping with Sahara, greater than $450,000 in crypto rewards can be found at launch.
Ren, who additionally teaches laptop science on the College of Southern California, mentioned the bug bounty mannequin helps guarantee higher information high quality whereas discouraging dangerous actors via strategies like automated checks, peer overview, contributor repute scores, and requiring customers to stake tokens as a protection towards assaults.
Sahara follows different crypto-native bounty fashions, comparable to Immunefi, which pays out for safety vulnerability disclosures, and Gitcoin, which helps open-source software program improvement via grants, bounties, and funding.
In keeping with Sahara, DSP duties fall into three classes:
Enterprise Duties, which pay in $SAHARA tokens for serving to main purchasers construction or label information.
Twin-Reward Duties, which pay in $SAHARA and ecosystem associate tokens.
Neighborhood Duties, which don’t supply quick payouts however give contributors possession stakes in ensuing datasets and recurring earnings when these datasets are used or bought.
Sahara additionally helps rewards in stablecoins, together with USD1, USDC, and USDT as different reward choices.
Whereas bug bounties aren’t new, Ren mentioned the mannequin focuses extra on offering high quality information, not simply fixing holes.
“Knowledge labeling is key for coaching correct AI fashions,” he mentioned. “Labeled information acts as a information to show fashions to acknowledge patterns and make predictions.”
To take care of information high quality, Sahara employs a number of safety and overview mechanisms.
“We monitor the efficiency of information labelers utilizing honeypot questions,” mentioned Ren. These are designed to detect customers trying Sybil assaults, the place one particular person has a number of faux identities to sport the system.
“Cheaters might be rapidly recognized with our high quality assurance mechanisms after which be briefly banned or long-term banned,” he mentioned.
Likewise, the corporate is utilizing AI to catch cheaters—satirically, utilizing AI to search out customers who’re themselves utilizing AI to create content material.
Usually Clever Publication
A weekly AI journey narrated by Gen, a generative AI mannequin.
