What is AI?
Artificial intelligence, or “AI”, is the ability for a
computer to think and learn. With AI,
computers can perform tasks that are typically done by people, including
processing language, problem-solving and learning.
What is the Human-In-Loop-Process?
It is the integration of a human workforce in the AI
pipeline in order to train and validate models in a continuous way.
In other words, this is the process of combining machine and
human intelligence to obtain the best results in the long-term.
The Streamline NDIS Model requires human intervention so
that humans can verify the predictions of the AI model and send that feedback
to either replace the AI-generated prediction or to be used for future
re-training and fine-tuning of the model.
How does the Streamline Model Retrain
- For every prediction, the machine learning model
emits a confidence level.
- The machine learning approach enables the model
to learn from where it is least confident by routing low confidence predictions
to the human-in-the-loop process.
- We can then take the annotations from the
human-in-the-loop process to input into the model training process.
- For our Streamline NDIS Model, we are uploading
a sample of NDIS invoices and pushing them through the Human-In-the-loop
process and using our annotation team to do the annotation.
What is the Streamline Model Training Process
To improve data
accuracy in Streamline, we capture changes made in Streamline before the
Invoice is released to the Invoice Entry feature and send an updated data file
back to Sypht.
As an example,
if Streamline was unable to initially read the Invoice Number, we capture the
Invoice Number once it is manually changed and the Invoice is created, and send
an updated data file back to Sypht with these changes.
Sypht then run
a monthly process, using the updated information to improve the accuracy of
their AI as demonstrated in the process flow, below.