I used to think that the value in AI was simple to find.
A model gets smarter a product gets better. The person who built it gets the benefits.
But the more I looked at how AI gets better the more I thought about one thing:
When an AI learns from data, where does that training value go?
Because training is not one thing that happens.
It is a process.
Someone collects data.
Someone makes sure the data is good.
Someone labels the data.
Someone updates the data.
Someone checks if the data is still good month not just today.
Then the model uses all of that work to get better. Until it gives better answers, better suggestions and better predictions.
You can see the results.
You cannot see what goes into making it happen.
That is where the problem starts.
In systems the people who help the model get better do not get recognized or rewarded.
The value goes to the model and the product. It does not go back to the people who helped make it happen in a fair way.
When you have a lot of people working on it it gets even harder to make sure everything is consistent.
It is like running a company with many stores.
One store can be great. When you have many stores it is hard to keep everything the same.
Small things can be different like the food, the standards and the service.
Nothing big goes wrong. People can start to trust you less over time.
Data systems can have the problem.
If you do not track who contributes to the data if you do not check the quality of the data and if you do not give credit to the people the system can still work. But it is harder to trust.
When people do not trust it the AI model does not get better.
This is the problem that @OpenLedger (OPEN) is trying to solve.
It is not a saying like "data is important”. It is a real question:
How do you build a system where people can work together on data check its quality and give credit to the right people. So the value of the training data does not get lost?
Because over time the advantage may not just come from making an AI model once.
It may come from the systems that keep getting better the feedback loops, the data flows and the discipline to keep everything as more people join.
So when I ask "where does AIs training value go" I am not asking a question.
I am asking a question, about how to make systems work better.
#OpenLedger makes me think that this question is important and cannot be ignored.$OPEN
