On Wednesday last week we hosted the fifth event in our #AdrenalinSessions series which encourages knowledge-sharing and collaboration within the digital industry. Wednesday’s talk was focused on the topic of Artificial Intelligence and its relevance to brands.
We heard from the following speakers:
Maurice Pagnucco, Professor, Deputy Dean (Education), UNSW Engineering and Head of School, School of Computer Science and Engineering
Henry Cho, Head of UX and AI, Upwire
Magda Cortez, Senior Product Manager at eBay Classifieds Group, Currently working on Gumtree
Here is a summary of the slides presented:
The doors opened at 7.30am when the audience members made their way in, grabbed a coffee and some breakfast ready for the 8am start. Maurice Pagnucco kicked off the morning by discussing the history of artificial intelligence and when the term AI was coined.
Image: Maurice Pagnucco, UNSW
McCarthy (1956) coined the term AI
He then moved on to explain what artificial intelligence is. It breaks into three main areas:
- Symbolic AI
- Sub-symbolic AI
- Statistical AI
The key differences between Machine Learning and Recommender Systems (like Amazon and Spotify use to recommend other things you might like).
- Learning patterns from data.
- Many different approaches
- One current popular method: deep learning
- Used by AlphaGo
- Used for image processing and computer vision
- Labelling images
- Reverse image search
- Collaborative filtering
- Content-based recommender systems
- Other types of recommender systems:
- Opinion-based recommender systems
- Risk aware recommender systems
The future is about augmented intelligence
Image: Henry Cho, Upwire
Henry Cho spoke next about the four main roles of AI. He described these as:
Support – Generate – Enhance – Deliver
He provided examples of how brands are already using AI in these ways.
Artificial Intelligence is freaking us out!
Support – Humans in the creation of experience
Henry provided an example of how Air BnB uses AI to turn sketches into code, for example turning wireframes into designs for pages. He also mentioned Adobe Sensei and Google’s Autodraw – which if you haven’t tried, you should (although we still think it’s got a little while before it could be joining our creative team!)
Generate – experiences without human interaction
He spoke about how Netflix automates localisation with AI selecting relevant screenshots from programs and inserting the title of the show into a suitable section of the image. Also WIBBITZ which uses AI-driven production software to create short videos for USA Today. The Washington Post has also leveraged automated storytelling to cover high school football stories. He raised the interesting point that AI can take some jobs out of the equation:
Could this eliminate juniors?
If junior roles are taken over by AI it would make it harder to break into a junior level job.
Enhance – experiences through personalisation augmentation and anticipation
Three common enhancements are:
- Identify behaviours, individuals, trends for targeting, personalisation or insight generation
- Predict future outcomes or anticipate customer needs
- Understand – allow natural interaction with humans via voice, text or visual input
Deliver – intrinsically intelligent experiences
Apple’s calendar reminders combine first, third and network data to provide value by reminding you about an appointment and recommending when to leave to arrive on time.
Artificial Intelligence is like running with scissors - we want to run with it, but we’re unsure of what it might do and the power we have with it!
Image: Magda Cortez, Gumtree Australia
Lastly we heard fromMagda Cortez who discussed MVP (Minimum Viable Personality).
It's important to understand human movement on a website and when the right moment is for Artificial Intelligence to interject with customer experience.
Gumbot is the Gumtree robot which is a Messaging system on the platform which enables buyers and sellers to connect. Gumbot is a chatbot which operates inside this Messaging service to help customers trade.
Magda described how predetermined ‘canned’ responses inside the Messaging function helps their customers communicate at the tap of a button and uses neural networks based on common conversation phrases.
Last year Gumtree launched the Facebook Messenger chatbot which helps users search for items on Gumtree and get Customer Support. Magda said that they found that customers used it more for customer support than for searching for items.
Magda summarised by saying: How can you create a bot that’s more intelligent? Giving it an MVP - Minimum Viable Personality. The reality is, AI is in its infancy. We have got a long way to go before they [bots] can pick up on nuances, humour or emotion. This is where you can inject your brand's tone of voice. Fynd have created FiFy and she responds in a fun conversive way (for example she uses emojis and exclamations).
Building a Chat Bot is easy - building an intelligent Bot is hard
If you would like to find out about how Adrenalin works with AI, or if you have project you would like to discuss – get in touch.
Our next #AdrenalinSessions will take place on Wednesday 13th December and will focus on helping you further you digital career.
Sign up to ‘How to Land Your Dream Job in Digital’ here.