Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: information science

Brain-computer interfaces (BCIs) are tools that can connect the human brain with an electronic device, typically using electroencephalography (EEG). In recent years, advances in machine learning (ML) have enabled the development of more advanced BCI spellers, devices that allow people to communicate with computers using their thoughts.

So far, most studies in this area have focused on developing BCI classifiers that are faster and more reliable, rather than investigating their possible vulnerabilities. Recent research, however, suggests that algorithms can sometimes be fooled by attackers, whether they are used in computer vision, speech recognition, or other domains. This is often done using , which are tiny perturbations in data that are indistinguishable by humans.

Researchers at Huazhong University of Science and Technology have recently carried out a study investigating the security of EEG-based BCI spellers, and more specifically, how they are affected by adversarial perturbations. Their paper, pre-published on arXiv, suggests that BCI spellers are fooled by these perturbations and are thus highly vulnerable to adversarial attacks.

Read more | >

Forget the Thighmaster. Someday you might add a spring to your step when walking or running using a pair of mechanically powered shorts.

Step up: The lightweight exoskeleton-pants were developed by researchers at Harvard University and the University of Nebraska, Omaha. They are the first device to assist with both walking and running, using an algorithm that adapts to each gait.

Making strides: The super-shorts show how wearable exoskeleton technology might someday help us perform all sorts of tasks. Progress in materials, actuators, and machine learning has led to a new generation of lighter, more powerful, and more adaptive wearable systems. Bulkier and heavier commercial systems are already used to help people with disabilities and workers in some factories and warehouses.

Read more | >

TrackML was a Kaggle competition in 2018 with $25 000 in cash prizes where the challenge was to reconstruct particle tracks from 3D points left in silicon detectors. CERN (the European Organization for Nuclear Research) provided data over particles collision events. The rate at which they occur over there is in the neighborhood of hundreds of millions of collisions per second, or tens of petabytes per year. There is a clear need to be as efficient as possible when sifting through such an amount of data, and this is where machine learning methods may be of help.

Particles, in this case protons, are boosted to high energies inside the Large Hadron Collider (LHC) — each beam can reach 6.5 TeV giving a total of 13 TeV when colliding. Electromagnetic fields are used to accelerate the electrically charged protons in a 27 kilometers long loop. When the proton beams collide they produce a diverse set of subatomic byproducts which quickly decay, holding valuable information for some of the most fundamental questions in physics.

Detectors are made of layers upon layers of subdetectors, each designed to look for specific particles or properties. There are calorimeters that measure energy, particle-identification detectors to pin down what kind of particle it is and tracking devices to calculate the path of a particle. [1] We are of course interested in the tracking, tiny electrical signals are recorded as particles move through those types of detectors. What I will discuss is methods to reconstruct these recorded patterns of tracks, specifically algorithms involving machine learning.

Read more | >

He called the visits his “tail” and his “leash.” Eventually, his leash was stretched to every two weeks. Later, it became a month. Mr. Gates wasn’t told why. He complained that conversations with his probation officers were cold and impersonal. They rarely took the time to understand his rehabilitation.

He didn’t realize that an algorithm had tagged him high risk until he was told about it during an interview with The New York Times.

Read more | >

The drug, known as DSP-1181, was created by using algorithms to sift through potential compounds, checking them against a huge database of parameters, including a patient’s genetic factors. Speaking to the BBC, Exscientia chief executive Professor Andrew Hopkins described the trials as a “key milestone in drug discovery” and noted that there are “billions” of decisions needed to find the right molecules for a drug, making their eventual creation a “huge decision.” With AI, however, “the beauty of the algorithm is that they are agnostic, so can be applied to any disease.”

We’ve already seen multiple examples of AI being used to diagnose illness and analyze patient data, so using it to engineer drug treatment is an obvious progression of its place in medicine. But the AI-created drugs do pose some pertinent questions. Will patients be comfortable taking medication designed by a machine? How will these drugs differ from those developed by humans alone? Who will make the rules for the use of AI in drug research? Hopkins and his team hope that these and myriad other questions will be explored in the trials, which will begin in March.

Read more | >

The hidden secret of artificial intelligence is that much of it is actually powered by humans. Well, to be specific, the supervised learning algorithms that have gained much of the attention recently are dependent on humans to provide well-labeled training data that can be used to train machine learning algorithms. Since machines have to first be taught, they can’t teach themselves (yet), so it falls upon the capabilities of humans to do this training. This is the secret achilles heel of AI: the need for humans to teach machines the things that they are not yet able to do on their own.

Machine learning is what powers today’s AI systems. Organizations are implementing one or more of the seven patterns of AI, including computer vision, natural language processing, predictive analytics, autonomous systems, pattern and anomaly detection, goal-driven systems, and hyperpersonalization across a wide range of applications. However, in order for these systems to be able to create accurate generalizations, these machine learning systems must be trained on data. The more advanced forms of machine learning, especially deep learning neural networks, require significant volumes of data to be able to create models with desired levels of accuracy. It goes without saying then, that the machine learning data needs to be clean, accurate, complete, and well-labeled so the resulting machine learning models are accurate. Whereas it has always been the case that garbage in is garbage out in computing, it is especially the case with regards to machine learning data.

According to analyst firm Cognilytica, over 80% of AI project time is spent preparing and labeling data for use in machine learning projects:

Read more | >

If you’re interested in mind uploading, then I have an excellent article to recommend. This wide-ranging article is focused on neuromorphic computing and has sections on memristors. Here is a key excerpt:

“…Perhaps the most exciting emerging AI hardware architectures are the analog crossbar approaches since they achieve parallelism, in-memory computing, and analog computing, as described previously. Among most of the AI hardware chips produced in roughly the last 15 years, an analog memristor crossbar-based chip is yet to hit the market, which we believe will be the next wave of technology to follow. Of course, incorporating all the primitives of neuromorphic computing will likely require hardware solutions even beyond analog memristor crossbars…”

Here’s a web link to the research paper:

Computers have undergone tremendous improvements in performance over the last 60 years, but those improvements have significantly slowed down over the last decade, owing to fundamental limits in the underlying computing primitives. However, the generation of data and demand for computing are increasing exponentially with time. Thus, there is a critical need to invent new computing primitives, both hardware and algorithms, to keep up with the computing demands. The brain is a natural computer that outperforms our best computers in solving certain problems, such as instantly identifying faces or understanding natural language. This realization has led to a flurry of research into neuromorphic or brain-inspired computing that has shown promise for enhanced computing capabilities. This review points to the important primitives of a brain-inspired computer that could drive another decade-long wave of computer engineering.

Read more | >

If you are interested in mind uploading, then I have a research paper for you to consider. One of the serious issues with mind uploading is the computer substrate. Simulating the brain will require a new and incredible computing capability. New techniques and new hardware are going to be required to make it practical. Of course, there is currently zero demand for mind uploading hardware, so the market is not going to provide this capability. However, there is incredible market demand for cutting edge hardware for machine learning and artificial intelligence. And it turns out that one potential technique for artificial intelligence simulates the way that the brain works: neuromorphic computing. And there is a relatively new type of electronic component that seems to mimic some of the functions of a brain’s neuron: the memristor. Memristors are relatively new, having only been fabricated for the first time by HP in 2008. So I am trying to keep up with the latest developments in memristive technology.

Here are some excerpts from the paper:

“…Artificial Neural Network (ANN) algorithms offer fast computations by mimicking the neuronal network of brains. A weight matrix is used in neural networks (NNs) for parallel processing that makes computing faster…The memristor has attracted much attention because of its potential to have linear multilevel conductance states for vector-matrix multiplication (output = weight × input), corresponding to parallel processing…”

Here is a web link to the research paper:

Neuromorphic computation is one of the axes of parallel distributed processing, and memristor-based synaptic weight is considered as a key component of this type of computation. However, the material properties of memristors, including material related physics, are not yet matured. In parallel with memristors, CMOS based Graphics Processing Unit, Field Programmable Gate Array, and Application Specific Integrated Circuit are also being developed as dedicated artificial intelligence (AI) chips for fast computation. Therefore, it is necessary to analyze the competitiveness of the memristor-based neuromorphic device in order to position the memristor in the appropriate position of the future AI ecosystem. In this article, the status of memristor-based neuromorphic computation was analyzed on the basis of papers and patents to identify the competitiveness of the memristor properties by reviewing industrial trends and academic pursuits. In addition, material issues and challenges are discussed for implementing the memristor-based neural processor.

Read more | >

The world’s first programming language based on classical Chinese is only about a month old, and volunteers have already written dozens of programs with it, such as one based on an ancient Chinese fortune-telling algorithm.

The new language’s developer, Lingdong Huang, previously designed an infinite computer-generated Chinese landscape painting. He also helped create the first and so far only AI-generated Chinese opera. He graduated with a degree in computer science and art from Carnegie Mellon University in December.

After coming up with the idea for the new language, wenyan-lang, roughly a year ago, Huang finished the core of the language during his last month at school. It includes a renderer that can display a program in a manner that resembles pages from ancient Chinese texts.

Read more | >