Toggle light / dark theme

Why Neural Networks can learn (almost) anything

A video about neural networks, how they work, and why they’re useful.

My twitter: https://twitter.com/max_romana.

SOURCES
Neural network playground: https://playground.tensorflow.org/

Universal Function Approximation:
Proof: https://cognitivemedium.com/magic_paper/assets/Hornik.pdf.
Covering ReLUs: https://proceedings.neurips.cc/paper/2017/hash/32cbf687880eb…tract.html.
Covering discontinuous functions: https://arxiv.org/pdf/2012.03016.pdf.

Turing Completeness:
Networks of infinite size are turing complete: Neural Computability I & II (behind a paywall unfourtunately, but is cited in following paper)
RNNs are turing complete: https://binds.cs.umass.edu/papers/1992_Siegelmann_COLT.pdf.
Transformers are turing complete: https://arxiv.org/abs/2103.

More on backpropagation:

Artificial intelligence deciphers detector ‘clouds’ to accelerate materials research

X-rays can be used like a superfast, atomic-resolution camera, and if researchers shoot a pair of X-ray pulses just moments apart, they get atomic-resolution snapshots of a system at two points in time. Comparing these snapshots shows how a material fluctuates within a tiny fraction of a second, which could help scientists design future generations of super-fast computers, communications, and other technologies.

Resolving the information in these X-ray snapshots, however, is difficult and time intensive, so Joshua Turner, a lead scientist at the Department of Energy’s SLAC National Accelerator Center and Stanford University, and ten other researchers turned to artificial intelligence to automate the process. Their machine learning-aided method, published October 17 in Structural Dynamics, accelerates this X-ray probing technique, and extends it to previously inaccessible materials.

“The most exciting thing to me is that we can now access a different range of measurements, which we couldn’t before,” Turner said.

This breakthrough AI can detect lung diseases with 98% accuracy

Researchers at the University of West Scotland (UWS) believe that groundbreaking artificial intelligence (AI) could help reduce winter stresses and demands on hospitals. The innovative approach, using AI, would automatically diagnose lung diseases, such as pneumonia and tuberculosis.

The research was published in the journal Computer Methods and Programs in Biomedicine.

AI could lead to surge in lawsuits as people fight over ownership

COURTS across the globe are braced for a surge in cases related to AI in years to come as the technology develops at a rapid pace.

Officials are scrambling to come up with new laws on how the advanced systems should be kept in check and ultimately who is responsible for what they do or create.

We recently heard about AI-generated art that won a competition in Colorado, leading to a backlash and questions about fairness.

CISA Warns of Critical Vulnerabilities in 3 Industrial Control System Software

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has published three Industrial Control Systems (ICS) advisories about multiple vulnerabilities in software from ETIC Telecom, Nokia, and Delta Industrial Automation.

Prominent among them is a set of three flaws affecting ETIC Telecom’s Remote Access Server (RAS), which “could allow an attacker to obtain sensitive information and compromise the vulnerable device and other connected machines,” CISA said.

Teleportation Anyone? XPRIZE Winner Is The Closest You Can Get

Teleportation became a bit more real on Saturday when a German team of data scientists and engineers won the ANA Avatar XPRIZE competition in Long Beach, California, with a four-wheeled, humanoid robot named NimbRo.

But in this form of teleportation, rather than transporting a human to a remote location, vision, hearing, and a sense of touch were wirelessly transmitted from a humanoid robot to a remote human operator who then directed the robot to complete a series of complex tasks.

“Telepresence and avatar technology will be an essential part of human progress in the decades to come,” said David Locke, ANA Avatar XPRIZE’s executive director said in a statement following the conclusion of the four-year competition.

Meta AI creates first ever database of 600 million metagenomic structures

‘These structures provide an unprecedented view into the breadth and diversity of nature,’ say the researchers.

In a world first, Meta’s artificial intelligence (AI) has produced the structures of the metagenomic world at the scale of hundreds of millions of proteins, according to a blog by the company published on Tuesday.

“Proteins are complex and dynamic molecules, encoded by our genes, that are responsible for many of the varied and fundamental processes of life. They have an astounding range of roles in biology,” wrote the Meta research team who also published a paper on the matter in the preprint database bioRxiv.


It is common knowledge that a vast number of proteins exist beyond the ones that have been catalogued and annotated in well-studied organisms and now these proteins are coming to the surface.

Women get to speak less in TV debates than men, AI analyzes 625,409 dialogues

It’s a small comparison percentage, however, it matters.

The big idea.


Simonkr/iStock.

I’m a computer scientist who uses AI to study social science questions. In collaboration with student AI researchers from Carnegie Mellon University, we developed AI methods that reliably distinguish intrusive and unfriendly interruptions from those that are benign. Intrusive interruptions aim to take over a conversation or stifle the speaker, and benign interruptions aim to support the speaker with helpful information or indications of agreement.

/* */