The "A" in AI?

There’s really only one possible interpretation, and it’s “artificial”, isn’t it?

For a long time, people would have given me funny looks for even debating this, and there are plenty even now who might think it a little crackpot to ask the question “is artificial the best we can do?”. In January, I was lucky enough to hear a talk by Garry Kasparov that chimed with a different point of view. Maybe we can take the step from “artificial” to “augmented”. 

At the heart of modern AI research, there’s a crisis that is getting closer to the surface. We have all these amazing new tools that are approaching human capability in image recognition and natural language tasks, but we have little idea how they work. Studies have shown that many image processing algorithms focus on strange and unexpected parts of pictures to recognise their contents, and natural language models make strange, un-human-like errors. How can we really trust something we don’t understand? Moreover, without a complete map of human cognition, we might wonder if we can ever hope to make these solutions more human-like.

Enter, stage-left “advanced chess”. Fresh from a bruising defeat at the hands of the brute-force computing power of “Deep Blue” in 1997, Kasparov saw an opportunity for cooperative competition. Rather than pitting man and machine against one another, why not bring them together? Allowing each human player in the game to use a computer to check for human blunders, and at the same time allowing humans overall control of the long-term strategic aspects of the game played to the strengths of both. A new and exciting form of competition was born, where chess engines can augment human problem solving ability. 

Back in the 1990s, chess engines worked by computational power alone, doing a tree-search of all possible moves that a player could make, and selecting the best amongst all possible options. Now, with the advent of the likes of DeepMind’s alpha zero, there are much more powerful, but less well-understood chess playing tools around - ones which will make better moves, but are unable to show you their working. Kasparov argues that advanced chess should be the paradigm of 21st century coexistence with intelligent algorithms - cooperation, not competition. 

At Kare, our underlying technology is based on this exact view. We want to give you, our customers, a tool that can search through your documents and your knowledge to find the right answer to a natural language question. As a company, however, you have unique skills and knowledge, and that’s what makes you different. We want to enable our clients to control the way their customers interact with that knowledge, and to express their singular view of the world in a way that keeps them apart from the rest. We provide a solution that can answer the overwhelming majority of questions automatically, but can also ask for expert human input when it matters most - when your customers need to know. It’s time to think about “augmenting” your CX, rather than just making it “artificial”...