Author: Josh Ho & Teng Yan, Chain of Thought; Translated by: 0xjs@Golden Finance
Sahara has been quiet for months, but just this week announced it had raised a massive $43 million in funding.
That’s a massive chunk of change, making it one of the largest funding rounds in the crypto AI space. It’s also a strong signal of investor interest.
They’ve secured some top-tier backers, with Pantera, Polychain, and Sequoia all involved. But what really caught our attention were the advisors, which include the co-founder of Nous Research and a founding member of Midjourney.
We have a soft spot for teams with big ambitions, so we did what we do best: dug into everything we could find on Sahara and distilled it into a quick 5-minute read.
Sahara CEO Sean Ren appears on Show Me the Crypto (Episode 134). We watch the entire hour so you don’t have to.
Collaborative AI Economy
Sahara AI is building the AI infrastructure layer to make decentralized AI a reality.
Their plans? Ambitious to say the least. They are working on:
AI-native blockchain
User-friendly development tools
Secure data storage, etc.
Users contribute their personal data to build domain-specific knowledge bases, developers use this data to train and deploy AI agents, infrastructure providers host Sahara nodes, and enterprises get tailored AI solutions.
But let’s face it – you can’t achieve a great vision by trying to do everything everywhere at once.
You have to start with a sharp product wedge and then grow from there.
Sahara’s first product: AI Marketplace
Sahara’s initial product is the AI Marketplace, which is a key pillar of its marketing strategy.
In discussions with their seed investors, it became clear to them that this marketplace was a preview of how they planned to achieve their ultimate vision. This was their first step, even before the Sahara Chain testnet was launched.
Source: Sahara
In the marketplace, users can complete tasks that help train AI. These tasks range from solving math problems to summarizing videos, which are essentially all about data labeling and annotation.
The submissions will then be used to train an AI model through reinforcement learning. In return, those who complete the task will receive points and, potentially, tokens in the future (although this is pure speculation at this point).
Enterprise clients such as MIT, Microsoft, USC, and Motherson Group have already started using the decentralized data marketplace to train their AI models. It will soon be available to the public.
The marketplace offers a robust set of tools including a data toolkit (for collection, tagging, QA, etc.), a distribution engine, and project management tools.
Why did Sahara do this?
Sahara concluded that current challenges facing AI require new levels of data labeling and annotation.
Sean Ren, CEO of Sahara, explained:
“We are trying to build our network to enable 200,000 data annotators to do these enterprise-level tasks. These data annotators come from all over the world, mainly from Africa and Asia Pacific, and they have different expertise, language and cultural backgrounds.
Some of them had never worked on a data labeling task before, but that was the challenge. How do you make a product that looks very user-friendly so that first-time data labelers can get involved and generate high-quality data for large enterprise customers like Microsoft, Amazon, Snapchat, etc.?”
Your knowledge, your AI
One of the biggest issues facing AI right now is copyright. Earlier this year, several US newspapers filed a copyright infringement lawsuit against OpenAI — and that’s just the tip of the iceberg.
When OpenAI trained ChatGPT, it did not acknowledge contributions from Reddit, Wikipedia, and GitHub users — the massive databases it used to train its models.
Sahara believes this is a fundamental flaw. They recognize that these user contributions are extremely valuable and it is only fair that users benefit from them. Sound familiar? We mentioned this before when we wrote about Vana, the Robin Hood of User Data.
Sahara argues that if you contribute value to the AI process — whether by sharing knowledge on an AI marketplace or uploading personal data for your own AI agent — you should own it.
This is where the word “Provenance” comes in.
In Sahara’s context, Provenance means adding a user watermark to your data. Messages, emails, content — everything will have your unique watermark. That way, anyone who uses your data to develop an AI application or train a model will have to pay for it. Provenance also extends to models and applications.
In this way, Sahara ensures that the original contributors are recognized and compensated accordingly.
Alpha
Sahara is still in its early stages, but momentum is building. Back in May, they opened up testnet registrations, which attracted 25,000 applicants.
The Sahara mainnet is expected to launch sometime in Q4 2024. You can join their waitlist here.
team
Sahara CEO and co-founder Sean Ren has a rich background in AI. He is an associate professor at the University of Southern California, specializing in natural language processing and artificial intelligence, and has won multiple awards for his research and innovation in the field of AI.
Sahara COO and co-founder Tyler Zhou served as Investment Director of Binance Labs prior to founding Sahara.
You can learn about the background of other members of the team here.
Our thoughts
Sahara has an ambitious roadmap for the rest of the year, which includes the delivery of the mainnet. If they can achieve this, it will be a strong testament to the team’s capabilities.
To be honest, there isn’t enough public information available to evaluate Sahara in depth yet (although Litepaper is coming soon). A lot of what we’ve heard sounds promising and aspirational, but the real test will be whether they can deliver on their goals.
