Artificial Intelligence: MIT creates ‘liquid’ machine learning – ‘future of robot control’

Artificial Intelligence: MIT creates ‘liquid’ machine learning – ‘future of robot control’

01/28/2021

Artificial intelligence: Expert discusses research on future crime

Massachusetts Institute of Technology (MIT) engineers have arrived at a new solution to advance AI – a neural network fluid enough to learn on the job. Such agile algorithms, nicknamed ‘liquid’ networks, can continuously change their underlying equations to adapt immediately to new data inputs.

This latest advance in AI may transform aid decision-making based on evolving data, including cutting-edge tech used in medical diagnosis and autonomous driving.

This is a way forward for the future of robot control

Dr Ramin Hasani

Dr Ramin Hasani, the MIT research’s lead author claims “The potential is really significant.”

He said: “This is a way forward for the future of robot control, natural language processing, video processing — any form of time series data processing.”

Dr Hasani added how time series data are now considered both ubiquitous and vital to our understanding of the world.

We will use your email address only for sending you newsletters. Please see our Privacy Notice for details of your data protection rights.

He said: “The real world is all about sequences. Even our perception — you’re not perceiving images, you’re perceiving sequences of images.

“So, time series data actually create our reality.”

Video processing, financial data, and medical diagnostic applications as everyday examples of time series already central to today’s society.

And although such constantly altered data streams can be unpredictable, analysing these data in real-time, and using them to anticipate future behaviour boasts the potential for further improving technology used every day.

The MIT team built a neural network engineered to adapt to the variability of real-world systems.

Neural networks are complex algorithms used to recognise patterns after studying sets of “training” examples.

They are often conceived as mimicking our brains’ processing pathways, with Dr Hasani in this instance reportedly inspired by the microscopic nematode, C. elegans.

He said: “It only has 302 neurons in its nervous system, yet it can generate unexpectedly complex dynamics.”

DON’T MISS
Stephen Hawking’s ‘black hole time machine’ proposal to NASA [REVEALED]
Stonehenge breakthrough: Julius Caesar letter exposes ‘secret’ [VIDEO]
Antarctica discovery: Century-old letter reveals shock find [PICTURES]

His neural network was coded with careful attention to how C. elegans neurons activate and communicate with each other via electrical impulses.

The elegant equations used to structure his neural network allowed the parameters to change over time based on the results of a nested set of differential equations.

This flexibility is considered crucial – with most neural networks’ behaviour fixed after the training phase, meaning they are poor at adjusting to change.

However, Dr Hasani says the fluidity of his “liquid” network makes it “more robust” and “more interpretable” to unexpected or “noisy” data, such as when bad weather obscures the view of an autonomous vehicle’s rear-view camera.

He added how this liquid network avoids the inscrutability common to other neural networks.

Dr Hasani said: “Just changing the representation of a neuron, you can really explore some degrees of complexity you couldn’t explore otherwise.”

The next stage involves continuing to fine-tune the system and prepare it for industrial application.

He said: “We have a provably more expressive neural network that is inspired by nature.

“But this is just the beginning of the process. The obvious question is how do you extend this?

“We think this kind of network could be a key element of future intelligence systems.”

Source: Read Full Article