- Learns and predicts as it goes along.
- Add new data immediately and automatically.
- Interactive to improve the relevance of results.
A dynamic AI is the one that learns and predicts as it goes along. An AI that adapts to a changing world without re-training. An AI to interact with in different and unpredictable ways. With all three parts working together in concert to discover (‘things’) such as images, video, audio, db records and documents — this is dynamic AI.
Conventional AI transforms raw data into feature vectors and feeds them into a machine learning algorithm. Working its magic on the feature vectors, the algorithm outputs a trained model also called a predictive model(*). Show the model unseen data and it will return a prediction (or score). This is a big deal. But, it is also limited.
This is odd
Predictions remain fresh until new data arrives at which time the model has to be re-trained to bring it up to date. Two problems arise: the current model predictions are out-of-date and the length of time to re-train is proportional to the volume of data. The industry solution is to re-train the model as fast as possible.
On many levels, this is an odd situation because modern software systems don’t work like this. A good counter-example is a database management system (dbms) to manage and manipulate data with simple crud and query statements. Add, update and delete data at any time while people continue to query the database. A dbms is a dynamic system whereas an AI service is a re-trained static model.
AI for the here and now
AI is a category term for different deep learning and machine learning methods for specific tasks including regression, classification, clustering and recommendations. Thingy is a dynamic AI for the discovery of items (“things”) such as images, video, audio, db records and documents. It is characterized by:
- Automatic generalization: Given one or more items as a query, Thingy learns in realtime to discover other relevant items. This means there is no machine learning ‘model training’ phase.
- Interactive: Engage with Thingy in different and unpredictable ways through various query options.
- Dynamic: Add new data to Thingy automatically at any time.
This is the first in a series of three articles to cover the three areas.
To kickoff, the first slidedeck shows a query with one image and Thingy discovers other relevant images in ranked order. So far so good. Next, add another image to the query. Well, this is not a typical machine learning scenario but Thingy amiably discovers other relevant images satisfying both query items. And, so on as more images are added to the query.
The concept is “given a handful of example images, discover others like them” with the learning taking place in realtime. Thingy learns and predicts as it goes along.
Items (“things”) can be images, audio, video, db records, documents, genes and in fact, all datatypes. But, not text per se though text documents are good to go.
Thingy is a dynamic recommendation engine and the next article in the series walks through interactive discovery with example results.
Bottom line: A dynamic AI learns and predicts as it goes along. Add new data at any time. Engage interactively in different ways to improve the relevance of results.
(*) In the image and audio space, trained deep learning models are employed without a machine learning component for object classification and recognition.
Articles in series
If you liked this, click the clapping hands below so other people will hear about this on Medium. Thank-you.