Machine Learning & Artificial Intelligence (AI)

Machine learning and artificial intelligence (AI) has gained prominance over the last few years with the release of tools including Dall-E 3, ChatGPT 3.5 and Microsoft Copilot.

These tools are the culmination of decades of work in developing algorithms and models that iteratively learn from data without being explicitly programmed. It has also coincided with the development of super computers using GPUs that can efficiently train complex models using a massive corpus of data.

While AI tools are being increasingly used, not all of them are ready for use in the enterprise. Individuals and organisations have used the output of models that have a tendency to hallucinate - fictious output not based on training data - and this has been very damaging for reputations and credibility.

There are a number of machine learning algorithms available:

• Logistic & linear regression
• Decision trees
• Random forests
• Support vector machines
• Gradient boosting & XGBoost
• Principal component analysis
• Dimensionality reduction
• Neural networks
• Deep netural networks, which include:
• Convolutional neural networks
• Recurrent neural networks
• Long short-term memory networks (LSTMs)
• Encoder decoder networks
• Transformer networks
• Generative adversarial networks
• Kolmogorov-Arnold networks

Many of these algorithms can be used for prediction or classification. The choice of algorithm will depend on the use case, the data available, the compute resources and other considerations.

Understanding how to use your data to take advantage of the latest machine learning algorithms is a challenge. There are many questions to consider:

• What do I want to achieve?
• Is my data suitable for machine learning, how much data do I need and how do I prepare my data?
• What algorithms, models, languages and libraries should I use?
• What infrastructure should I use to train my model?
• Should I train my model from scratch or fine-tune a trained model or use transfer learning?
• How do I deploy trained models and monitor performance?
• How do I undertake inference and what is my measure of success?
• What is the cost?

Organisations can also take advantage of generative AI models including ChatGPT, DALL-E and Copilot. These can be integrated into business processes as is or fine tuned with the organisation's data.

Machine learning is a challenging landscape to navigate. We can assist in helping your organisation develop and implement a machine learning and AI strategy to achieve tangible outomes.