Scientist explains what it will take to make AI work for marketers

Most of the time, "artificial intelligence", "machine learning", and "automation", familiar terms that are said to be transforming the technological landscape, still seem like abstract concepts to marketers.Marketing spoke to Richard Socher, adjunct professor at Stanford's computer science department, chief scientist at Salesforce as well as founder and CEO/CTO at MetaMind, who doesn't think the misunderstanding is altogether surprising. Although the impact of AI has already surfaced in the marketing and advertising world, few know that AI is the key driver for those opportunities, he said.What Socher advised, if marketers wish to leverage the technology, is to understand what contributes to a successful AI model, which helps marketers to identify opportunities easily.The three basic ingredients include: data, algorithms and workflow integration."The very first piece is always about acquiring the data.""Fortunately for now, some of the data in some cases is public data, but then in other cases like Facebook, it's harder to acquire the data," explained Socher. "When you want to connect marketing to other important aspects like services, you actually need to have access to all the service data, as well as the marketing data, as that can be a very powerful combination."When you want to connect marketing to other important aspects like services, you actually need to have access to all the service data, as well as the marketing data, as that can be a very powerful combination.Algorithm: keep natural language and image-language data in the loopThe second piece of the key ingredients is the algorithms, which, Socher explained, are mostly based on deep-learning, both on the computer vision side and on the natural language learning side."You want to be able to understand the sentiment and identify company logos. You want to be really accurate in your understanding of both how your marketing messages are perceived, but also predicting how likely your messages engaging to specific audience," he said.For instances, Socher said the cloud computing company is now equipped with the technology to identify differences in hashtag used for products, or the scenery of the picture that people are usually in when interacting with a specific product. "Let's say the brand sells running shoes. You may identify that people take more pictures with that shoe in the mountain than when they are in the beach, which helps the brand to improve its next marketing campaign, and understand customers better."Let's say the brand sells running shoes. You may identify that people take more pictures with that shoe in the mountain than when they are in the beach, which helps the brand to improve its next marketing campaign, and understand customers better.On the natural language processing side, Socher said it is very important to track the sentiments that users have regarding the brand's product or the company overall. Brands may also want to keep an eye on sentiment improvement after, for example, a marketing campaign."When someone take a photo of a pair of shoes and say, 'this thing sucks', they don't hashtag the company," he said. "But with local detection, marketers  can identify the company logo, which trigger the server automatically, allowing the marketing organisation to proactively reach out based on that public complaint, and make them happy again."When someone take a photo of a pair of shoes and say, 'this thing sucks', they don't hashtag the company. But with local detection, marketers  can identify the company logo, which trigger the server automatically.Workflow integration is often underestimatedSocher said there is a lot of interesting work on the algorithm side, but as a technologist, he often sees people underestimating the work-flow integration."It's really important to understand how marketers work, and it is great if you have a cloud-based software on a browser, so you can actually track how people use your software and make it better for them."