ML is a discipline within artificial intelligence (AI) that allows computer systems to learn and improve automatically without being specifically programmed. ML will take structured (such as database records) and unstructured data (such as images and voice recordings) and then analyze it to look for hidden patterns, dependencies, and so on which can then be used to develop predictive or explanatory models.
These models can then be used for future decision-making. For example, historic traffic records can be assessed to develop models that can be used to manage future traffic flows more effectively.
While ML provides some real benefits, it (like nuclear power) needs to be treated with respect otherwise it will cause disasters
- Beware of handing over full control to a computer. Always keep humans and other governance controls in the loop.
- Make Sure the Model (or Models) Work – do not rely on ML blindly – check that the models work and produce meaningful outputs.
- Remember That ML Models Have Been Developed/Tested Using Historic Data which means they may not cope with ‘new’ situations.
- ML needs Good, Timely, and Accurate Data – Garbage In and Garbage Out.
- Organisations should always have a clear business reason to use and implement ML
- Do not rush ML implementations. Start small. Learn. And then built up.
- ML needs dedicated and often new skills. This will require recruitment and/or staff development and in turn costs.
THIS AREA IS DISCUSSED IN MY BOOK ON EMERGING TECHNOLOGIES – CLICK HERE TO BUY A COPY
Recent Comments