How AI is revolutionising crisis communication
Covid-19 has impacted companies all around the world with massive operational changes as leaders scramble to respond, survive and innovate. It has shown us the importance of agility in our business plans but also how critical internal and external communication is to businesses, so they thrive rather than simply survive.
Many businesses who have weathered the storms and are seeing the calmer waters ahead are doing so thanks to a combination of technology and the human touch. Together, these two factors will be the secrets to success moving forwards.
Giants of technology
Microsoft Teams, Workplace by Facebook, Google, AWS and Zoom are everyday working terms today, and implementing and tailoring these tech giants’ solutions within our own structure, we were able to provide a near seamless service to our customers and look after the wellbeing of our people. Adopting it early is all well and good, but maintaining communication is the key now.
Over 2020 we have been working with state of the art technologies to help us better understand those conversations. We leverage the work from Google, Facebook, Baidu and OpenAI, who are all making significant breakthroughs in this field. Over the past few years Natural Laguage Processing technology has evolved rapidly, starting with Google’s Attention Is All You Need paper that presented the Transformer model, which achieved state-of-the-art performance on almost all Natural Language tasks. That paper opened the flood gates for increasingly sophisticated and increasingly large transformer-based models to be produced.
Today, companies announce cutting edge NLP performance based on the transformer approach on an almost weekly basis constantly pushing the boundaries of what is possible. Improvements using Transformer based models will only continue and the recently announced Performer model from Google may prove to be another step-change in NLP accuracy. This technology is helping us to move closer to a point where machines can have context aware and human-like interactions with us.
Computer guesswork
Another area that is helping us automate our communication processing is the ability for computers to guess. Historically the lack of good training data has been a barrier for all but the largest organizations, however that barrier has been reduced with the introduction of zero shot classification. This approach uses general purpose pre-trained models to classify text the machine has never seen before, taking contextual information from pre-trained general models to predict a classification of something it’s never seen, essentially guessing. For certain tasks, the models are already amazingly accurate, removing the need to train new models from scratch removes a significant barrier for companies wanting to adopt these technologies, essentially democratizing capabilities that were historically only available to the largest tech organizations.
These are just two examples where technology is advancing quicker than ever before, and it is especially true in the field of communication. It has given us more options to communicate, improved the ways we communicate and generally pulled us into a new world of opportunity. Covid-19 has brought to the fore what can be achieved and what is on the cusp of being achieved in communication. It has shown that we crave human interaction facilitated by technology and together these two elements will shape the way the world communicates as we step into our new future.
Pete Hanlon is CTO of Moneypenny
