It can be overwhelming to find successful uses of AI in business. The pace of innovation in academia far exceeds the pace at which companies can process the new technology and evaluate its utility. To get started, here are ways that AI is being used today.
Image related AI.
The field of computer vision focuses on understanding images. Anything from the actual object itself, to a cognitive concept represented in the image. To identify use cases where computer vision can help, think about places where extracting abstract representations about images can be useful.
- Recommender systems: E-commerce websites can use a convolutional neural network to contextualize images for recommendation systems.
- User shopping behavior: Analyze customer purchases to find similarities. Does a user prefer certain cuts? Materials?
- Visual search: Find items similar to one selected by a user.
- Retail analytics: Track most visited areas within a store.
- Magic mirror: Allow users to view your products on them.
- Object detection: Find and track objects in an image.
Text related AI.
The field concerned with anything text related is called natural language processing (NLP). Some current use cases that have worked well are:
- Machine translation: An example would be translating from English to French. The trick here is to think about anything in your business that could be framed as a translation task. For example, a stock movement over some period, translated into a stock movement into the next period.
- Data extraction: If you have unstructured data written in text, NLP has become really good at extracting that data. For example: "My cat Peter, ran into the supermarket." Current systems can recognize that Peter is a name and supermarket a location.
- Text classification: AI does a pretty good job nowadays at categorizing sequences of texts. A sentence like "I really loved that bad movie" would be reasonably classified as positive.
- Document summarization: Although this is still not a solved problem, newer systems can extract the most relevant sentence from a longer passage.
A main point to take away is to think about solutions to text and vision as recipes: the abstract ideas can transfer to other domains. Working with stock data? look at text translation. Working with detecting user behavior after seeing certain patterns? look at text classification.
