The printing press enabled mass production of books which allowed ideas to flow freely throughout Europe. It dramatically changed economics and finance (because better book keeping was possible), it transformed education (which was previously only for the elite). It enabled a scientific revolution and ultimately the democratisation of knowledge itself. So it was fairly fundamental.

Attending the conference gave me a chance to reflect on how AI technologies are developing. Here are three key takeaways from my experience at AI Day:

1. AI and the innovation associated with it has matured a lot from 12 months ago.

People seem to have tried, failed and learned from their implementations of chat bots and robotic automation. They have realised the importance of careful design to allow AI systems to work seamlessly with human systems.

Natural Language Processing (AI tools, like Siri, that can understand and sythesise speech) seems to be an area that has advanced rapidly; a very smooth real-time translation from Chinese to English was demonstrated. 

Deep Learning (a type of AI that is trained on huge data sets) is still driving much of the innovation. It was amusing to see an example of a deep learning algorithm watching cricket clips and providing the commentary. That’s a big step up from where we were this time last year.

The chief data scientist from Homes.co.nz had a useful message. That company understood they owned a valuable data set about real estate in NZ. They decided to collaborate with an Barcelona-based AI start-up that specialises in processing real estate images with AI. The plan is to automatically tag real estate photos so customers can do more customised searches to look for properties that have particular features (like “pizza oven“ or whatever). Sometimes strategic partnering is a great way to go.

Dr Mahsa Mohaghegh from AUT gave my favourite talk of the day which explored the future of the role of teachers. She argued rather than asking “will teachers be replaced by AI?”, the more informed question is “how much AI will be used by teachers in the future?” Citing some interesting examples that included Squirrel AI Learning and Amy the maths tutor (Japipunda), she concluded that “AI is going to change the shape of education”, but she reminded us that the role of teachers includes teaching wisdom, inspiring and motivating us. And it turns out humans like to learn from humans, from someone reachable and tangible, who can understand our imperfections. Someone with empathy. 

2. The discussion on ethics and fairness is still hot.

The trade-off between privacy and public safety is a hot issue. One of the speakers used the example that AI amplifies the Chinese social credit system. He described it as AI processing video images from a camera on the street that can identify you and then recognise that you are jay-walking for instance. In this case you automatically lose points in the system. The suggestion was that you can eventually end up on a black list, at which point, your social connections are warned before interacting with you. To me that sounds like a Black Mirror episode, but it could be happening, right now. 

And there remain other important issues. The world is experiencing ‘techlash’, and even tech employees are voicing concerns. To be honest I found myself feeling quite sceptical listening to several high-powered executives from big tech expounding their ethical design process.  

One speaker suggested we could go back to first principles (The Universal Declaration of Human Rights) to help determine what we should and shouldn’t do with AI. Many feel that regulators will have to play a critical role to keep people safe – for example from facial recognition software. Issues of bias, fairness and sexism are still very much in play: “facial recognition works really well if you are a white guy”. And by the way, why are nearly all the digital avatars attractive women?

3. Convergence of AI with other technology is particularly exciting.

The most jaw-dropping moment for me was a demonstration of real-time realistic rendering of an exact 3D replica of a human, in augmented reality, live on stage (via a smart phone.). You can read more about that here.

Finally, in the keynote talk, Sean Gourley quoted a study (Wang et. al, 2017) that compared human intelligence with AI for identifying metastatic breast cancer. In the study, the best AI deep learning algorithm scored 92.5 percent at identifying metastatic breast cancer. An independent pathologist reviewing the same images scored 96.6 percent. BUT when you combine the AI predictions with the human pathologist diagnoses, the result was 99.5 percent. When humans collaborate with AI, the results can be life saving.

So is it fair to compare AI to the printing press? Well, at AI day we explored how it is transforming education, health, communication and real estate. I’m OK with the comparison for now.