top | item 38185493

Data as investments in LLMs, art models, and AI startups?

6 points| anshyyy | 2 years ago

Been working on an idea where people can use their data to invest in AI startups.

A lot of the platforms (MTurk) out there are geared towards getting companies a buttload of low quality data from workers on the platform, who don't get paid what's worth for their data.

We want to try and flip this concept on its head - basically, people can not only rent and sell their data to AI startups, but also have the ability to invest them as a % of profits for future AI products - earn % of profits directly attributable to the company's model (all relevant AI products, services) for some (ex: 4-5 yr) period once the company earns revenue, with options to profitably contribute more. For most cases, it will probably be some combo of the three.

Similar idea goes for research ($$ on any end product it'll be used for, particularly nice if open source) -- y'all just have to add some data and you can have a stake in an AI company, while they can source the data without much upfront liability. All companies are vetted, and can add extensions to process / verify data, and automatically accept the ones they like.

Essentially, data is the bottleneck for a lot of AI companies, and we wanted to see if we can enable y'all to help 'em build out their products while having a voice in their future LLM, art model, AI products, etc.

We chatted with a couple companies, and we felt a bit interest.

What do y'all think? Would you be down to invest your data in AI (assuming there's extensions for anonymize data, clear privacy standards, etc.)?

11 comments

order

altdataseller|2 years ago

1. How will companies know your data is accurate and not fake?

2. Why would companies pay or give away equity for your data when they could get most of it for free if it’s on the web?

3. Why would anyone sign up for the hassle of giving companies data when it’s worth maybe a few dollars at most? Models rely on millions of data points and yours standalone by itself is worth very little

anshyyy|2 years ago

1. We're planning on providing an extension system to run preliminary validations on the data.

2. Cause nearly every who's-who AI company is facing lawsuits for doing just that.

3. You're referring to LLMs and other high data models (this doesn't really apply as much to things like healthcare). While that's true, think about it this way -- each piece of data might not be worth a lot in the beginning, it's tied to the model.

As the model brings in more revenue, your data goes up in worth. And so what you're really holding is an opportunity / contract for future profits. Each one might not be worth anything, but it turns into an investment opportunity for an outsider -- you can buy the rights to the contract from these individuals once you see the company is clearly rising.

tldr; Because it provides long term opportunities, and allows you to have a stake in a company with barely any effort.

izyda|2 years ago

Check out startups like the below:

- https://www.caden.io/ - https://www.joinklover.com/faqs

anshyyy|2 years ago

Caden's amazing! We're trying to enable users to monetize in a different manner, beyond "selling their data off as a one-time profit opportunity"; moreso, we're gearing ourselves more towards allowing people to profitably drop data into AI companies in a way that minimizes initial risk for the startups, and then reap the rewards as they grow.

anshyyy|2 years ago

Really love the idea of being able to automatically toggle your data from specific companies. We can add some support to allow you to port your data to invest, and recommend startups with some analytics.

vasili111|2 years ago

Does the current legislation ready for that kind of model to work?

anshyyy|2 years ago

We're seeing green lights from both American and EU legislation. We're also keeping ethicality at the forefront of design, and going to try and vet startups on the platform.

bulla|2 years ago

How much do you think a single individual's data worth?

anshyyy|2 years ago

Depends with use case -- a single piece of healthcare data would easily go for triple digits, while high volume data (blogs, text, images) would go for single or double.

The investment model also throws things a bit off, since the value of contract for profits also has a worth, and fluctuates with company performance.

freeredpill|2 years ago

The business model is the product. Would you like to Zoom to discuss?