Aaron Reich, senior director of technology innovation at Avanade, told iTWire in an interview that what was powerful about AI and its cognitive computing capability was that it was more than just direct language translation.
"Machines today can understand intent and nuances, making translation much more natural than in the past," he said. "However, it’s not just about the tool; it’s also about how it is embedded in an organisation’s culture."
Reich (below) is part of the innovation and incubation team at Avanade, a global professional services company providing IT consulting and services focused on the Microsoft platform.
For the past few years he has been responsible for growth of Avanade's Windows Azure business.
iTWire: How does artificial intelligence avoid the bias that is inherent in humans?
Aaron Reich: We are seeing a shift with AI to machines that are able to adapt to the natural situations and experiences fundamental to humans. Artificial intelligence is a combination of understanding (interpreting the meaning of data), reasoning (forming conclusions based on unstructured data and learning over time), and a natural experience (intuitive engagement of machines/apps with humans).
Andrew Ng, former chief scientist at Baidu, has said that if a typical person can do a mental task with less than one second of thought, we can probably automate it using AI either now or in the near future. That is a great vision, but what we see with clients today is if a person can make a decision with less than one second of thought, a ton of relevant data needs to be gathered, trained (continuously) and housed in the right domain to replicate the human outcome.
Most bias in AI is based on how organisations are guiding decision-making in the models they are building or inherently in the data that is being used to build the models. AI and its associated algorithms may provide an opportunity to eliminate some bias, but organisational transparency is key. Each organisation needs to make sure it is doing everything it can to acknowledge that bias does exist and be transparent about how it will appropriately manage it. At Avanade, we counsel clients to develop a digital ethics framework to address how data and algorithms are governed.
The developed world is overly skewed towards the use of English and is biased in favour of this language. How does AI overcome this limitation?
AI won’t solve language bias, which requires cultural and behavioural change. However, AI can provide the tooling to enable the shift. Language translation is one area where the technology is ready and we are going to see this as a first area of cognition that will be applied quickly. For example, Microsoft has Skype today which has real-time translation built in. Similarly, using cognitive services from Microsoft, Google or IBM, I can take a document or video and instantly see or hear it in English, whilst my colleague sees or hears it in Spanish. A start-up company called Waverly Labs is also due to release a real-time translation hearable later this year.
What is powerful about AI and its cognitive computing capability, is that it is more than just direct language translation – machines today can understand intent and nuances, making translation much more natural than in the past. However, it’s not just about the tool; it’s also about how it is embedded in an organisation’s culture.
The language that is in common use contains words that are, again, inherently biased. How does AI tackle this?
It is possible to build systems that detect bias and root it out, but in the short term, we should be focused on bringing organisations and agencies together to make sure what is being built with AI is best for society.
Is the bias removed – which would make it less efficient than humans? Or is the bias built in so that human efforts are recreated?
By augmenting employees’ capabilities using AI, organisations can achieve far more, faster, with intelligent action and better results. Starting with simple or repetitive tasks, we expect to see an increase in people’s time and available cognitive capacity to focus on the items that are meaningful both to them and organisational profitability. Organisations that aren’t automating the most mundane areas of their workforce risk boring — and ultimately losing — their most important resource: their people.
Organisations that aren’t already thinking of some element of workforce automation are lagging. Avanade recommends organisations start experimenting with AI now, most likely for simple, straight-forward processes. Regardless of how an organisation starts, they must be mindful of bias and provide guardrails to manage it.
Is AI meant to replace the function a human does, or to create a newer, better system?
To compete for the best and brightest minds, businesses will need to leverage technology that makes work more exciting, and more focused on meaningful, higher-order tasks – things that are more human, less robotic by nature. We see that humans will be more focused on strategic decision-making, with all the tedious tasks of collecting relevant information outsourced to machines.
As an example, a financial services client that Avanade works with is using text recognition to scan PDF application forms, exponentially speeding up the task for reviewers. The key for organisations is transparency and discussing with your workforce what AI means for your business and your workers. The human touch will always be essential to the workforce.
Some AI experiments of the recent past — like Microsoft's Chatbot and Google's self-driving cars, for example — have shown that there is more hype than reality around AI. Would you agree?
These days, it’s hard to pick up a newspaper or analyst report that doesn’t includes some kind of forecast on how AI will reshape businesses and industries. As a result, it’s challenging for organisations to know where to focus their energy and efforts. At Avanade, we counsel clients that now is the time to learn, explore your business for opportunities, and experiment. When you take a strategic approach first, you can identify the areas that make the most sense to start applying AI technologies.
Amongst our clients, the elements of AI that are real today are: robotic process automation (optimising current business processes using intelligent bots); advanced analytics (using machine learning technologies to translate vast amounts of data into actionable insights), and digital agents (where smart bots enable more intuitive engagement with humans).
The possibilities of artificial intelligence are immense, but organisations must carefully navigate the associated ethical dilemmas and prepare for unintended secondary consequences.