Real-life artificial intelligence (AI) has already taken on the task of creating a whole new generation of artificial intelligence.
A team of researchers at the University of Edinburgh have just shown that they can use a system of deep neural networks (DNNs) to generate and control an artificial intelligence, which is essentially a human brain.
The researchers say they hope to one day be able to build robots capable of thinking for themselves.
“Our aim is to create a machine that will be able, for example, to pick up on a simple emotion and give it to its companion and say, ‘OK, I know this is a bad feeling’,” says Dr Matt Hughes, from the Department of Computer Science at the university.
“And it will be much smarter than that.”
A team at the Institute for Machine Learning at the National University of Singapore has also demonstrated a way to generate a robot with a human-like understanding of human speech.
“What we’re doing is we’re generating a computer that can understand what the human being is saying, and that is a very powerful tool,” Dr Andrew Gaffney, one of the researchers who worked on the project, tells News24.
“We can even turn that into a language to talk to it, to understand what it’s saying.”
They’ve now shown that it is possible to use the same DNNs to create machines that can speak to each other.
The team from the University, University of Sussex and the Centre for Systems Biology and Computation (CSBC) at the Royal Holloway, University, used a system called DNN-NU that they built from scratch.
The research, which has been published in Scientific Reports, builds on previous work by Dr Hughes.
It shows that the computer can generate and manipulate an artificial language called human speech, which can then be used to teach it to respond to other human-made languages.
The software was created by researchers at CSBC using data from the BBC Natural Language project.
It has been used to create robots that can recognise faces and understand people, but the team says the system can also recognise and recognise emotions.
They have already shown that a robot capable of producing human-sounding speech can be trained to respond with emotion, like anger, sadness or joy.
“When it learns that it’s being taught to understand human speech by an artificial human, the AI learns a whole lot more about human speech than it does about the actual language that we’re training the AI to produce,” Dr Hughes says.
“It learns a lot more.
The researchers have also shown that the system has the potential to produce and recognise the “human-like” speech that humans produce. “
That is a huge advantage for the AI system because if we had the AI that is being trained to recognise human speech but it was not actually doing the language, we would have to train it to recognise speech in different human languages.”
The researchers have also shown that the system has the potential to produce and recognise the “human-like” speech that humans produce.
The artificial language produced by the DNN system is a language called LISP.
LISPs are used to describe concepts and concepts, such as “this is a car”, or “this sentence is a verb”.
In contrast, a human can recognise a “this” and “that” but it does not understand a “he” or “she” because “he or she” is not a concept.
Dr Hughes is confident that the technology will become a key part of the future of artificial intelligent systems.
“If you want to talk about intelligent systems, then we’re going to have to be working on it because the problem of making AI is not going to go away,” he says.
Dr Gaffneys says the work shows that DNN systems can be used “to build the human brain”.
“We’re going into a world where we can do all sorts of stuff that would be difficult or impossible to do in a human’s brain,” he said.
“DNN-P is the way to do it.”