Why humans will teach computers instead of programming them
Scyfer CTO explains how learning machines shape our world
Tijmen Blankevoort has always enjoyed studying how humans can make their own lives easier. In this TEDx talk at the Nyenrode University he makes a prediction about the next 20 years of Artificial Intelligence: We will all be teaching computers how to better help us out. Yes, “teaching” computers, instead of “programming” them.
Fistly, we can now safely show how, with the right amount of examples and training, that neural networks can learn to improve our lives and offer customised solutions. We, here at Scyfer, see in our artificial intelligence consultancy work that this is possible. Moreover, it’s possible in a wide range of industries, from medical to finance, media and more.
But what is a neural network, you may ask?
Put in more simple terms, it is a computer system built to mimic the human brain. That means the more examples you give it, the smarter it gets.
In recent times we’ve seen algorithms based on neural networks. They have enabled us to have machines learn from complex data. The next step in this evolution is a more organic way of learning. This will mean that we tell the system what to do, correct it when it makes mistakes, encourage it when it gets the information right.
We are actually close to being able to teach robots and computers to do useful tasks, much like the way we teach children. Pretty cool, right?
In the next twenty years, Artificial Intelligence will be as ubiquitous as the internet, and it will greatly improve our lives.
We may not all need a pancake-making robot in our lives, like Tijmen does. All of us can, however, benefit from, for example, machines detecting cancer cells better than the human eye. We shouldn’t be afraid about AI can bring, because we’ll be teaching it to benefit society as a whole.
Key takeaways from the presentation
- Machine learning is a tool we can use to greatly improve our lives. No Terminator-like future in sight (we’re pretty sure)
- People are in control of the tool, since they are the ones teaching it. And teaching it will become easier to do in the future for non-technical people.
- The future of AI consists of human-in-the-loop systems with active learning
- Programming leads to machine learning which in turn evolves into deep learning, after which active learning is the next step. That means we go from hard programming to having a computer learn from complicated data, to us being ultimately able to teach a system.