“Jin and colleagues devised an algorithm called TextFooler capable of deceiving an AI system without changing the meaning of a piece of text. The algorithm uses AI to suggest which words should be converted into synonyms to fool a machine.
This caused the algorithm to classify the review as “positive,” instead of “negative.” The demonstration highlights an uncomfortable truth about AI—that it can be both remarkably clever and surprisingly dumb.”
“What are your top 3 tips for legacy brands who are starting to explore AI/ML in e-commerce marketing? How can marketers integrate AI and ML from inception through to measuring performance?
1) Get informed about what ML can do for marketers and what you need to get started. Like all exciting technologies, there is a lot of noise out there and some over-inflated claims. A dose of realism will help guide you forward.
2) Start small! For example, implement AI-driven upsell recommendations based on the page of one of your more popular products rather than your entire catalog.
3) Define a clear hypothesis about which metrics indicate improvement and how much improvement you expect to see. It’s crucial to follow a scientific method to avoid getting lost in the complexities and possibilities of ML. And, automate collecting the performance data so the results are transparent and easily measurable.
4) Outsource for expertise – it is unrealistic (and time-consuming) to spin up a data science and ML group for these kinds of complex initiatives. When outsourcing or purchasing off the shelf tools, get a proof-of-concept tailored to your use case before committing to long term costs.”
“Algorithms were used to generate creative descriptions of each product via optical character recognition (OCR) and image processing. With this artificial-intelligence (AI)-enabled solution and natural language processing (NLP), automated product descriptions were generated. This reduced the time spent on this task and the retailer was able to reclaim lost online sales worth $6.9 mn, annually.”