While the October news was dominated by ongoing Facebook (sorry, meta) problems (you’d think they’d get tired of the apology tour), the most interesting news comes from the AI world. I am fascinated by the use of large language models to analyze the “language” of whales and to preserve endangered human languages. It’s also important that machine learning seems to have taken a step forward (pun slightly intended), with robots teaching themselves to walk through trial and error, and robots learning to assemble themselves to do specific tasks to execute.
- The design studio Artefact has developed a game to teach middle school students about algorithmic biases.
- Researchers are building large natural language models, possibly as large as GPT-3, to decipher the “language” of whales.
- A group in Berkeley has built a robot that uses reinforcement learning to teach itself to walk from the ground up – that is, through trial and error. They used two levels of simulation before loading the model into a physical robot.
- AI reinvents computers: AI drives new types of CPUs, new “out of the box” form factors (doorbells, devices), decision-making instead of conventional calculations. The “computer” as the computing device known to us could be on the way out.
- Strange Creatures: Unimals, or universal animals, are robots that use AI to develop their body shapes so that they can solve problems more efficiently. Future generations of robotics may not be designed with solid bodies, but have the ability to adjust their shape as needed.
- Would a National AI Cloud be a subsidy to Google, Facebook, etc., a privacy threat, or a valuable academic research tool?
- I was skeptical of digital twins; they seem to be a technology looking for an application. However, digital twins (AI models of real systems used to predict their behavior) appear to be a useful technology for optimizing the performance of large batteries.
- Digital twins could provide a way to predict supply chain problems and bypass bottlenecks. They could enable manufacturers to compromise between just-in-time stocking processes that are prone to bottlenecks and resilience.
- Modulate is a startup currently testing language change software in real time. They provide realistic, human-sounding voices that replace the user’s own voice. They are aimed at games, but the software is useful in many situations where harassment is a risk.
- Voice copying algorithms were able to fool both people and voice-activated devices about 50% of the time (30% for Azure’s voice recognition service, 62% for Alexa). This is a new front in the deep fakery.
- Facebook AI Research has created a series of Ego4D (Ego4D) videos to train AI. They want to build AI models that can “see the world as a person sees it” and answer questions such as “Where did I leave my keys”. In essence, it means they have to literally collect everything a subscriber does. Although Facebook denies considering commercial applications, there are obvious connections to Ray-Ban Stories and their interest in augmented reality.
- DeepMind is working on a deep learning model that can emulate the output of any algorithm. This is known as neuro-algorithmic thinking; it could be a step towards “general AI”.
- Microsoft and NVIDIA announce a natural language model with 530 billion parameters called the Megatron-Turing NLG 530B. That is bigger than GPT-3 (175B parameter).
- Can machine learning be used to document endangered indigenous languages and help language recovery?
- Beethoven’s 10th symphony completed by AI: I’m not convinced that Beethoven would have written this, but that’s better than other (human) attempts to complete the 10th that I’ve heard. For the most part it sounds like Beethoven, but it quickly becomes aimless.
- I am still fascinated by techniques to thwart face recognition. Here’s an article on an AI system that designs minimal, natural-looking makeup that reshapes the parts of the face that facial recognition algorithms are most sensitive to without significantly altering a person’s appearance.
- Thoughtworks’ Responsible Tech Playbook is a curated collection of tools and techniques to help businesses become more aware of the bias and become more inclusive and transparent.
- Kerla is a Linux-like operating system kernel written in Rust that can run most Linux executables. I doubt this will ever be built into Linux, but it’s another sign that Rust made it big.
- Turing Complete is a game about computer science. That’s roughly what it says …
- wasmCloud is a runtime environment that can be used to set up distributed systems with wasm in the cloud. WebAssembly was designed as a programming language-neutral virtual machine for browsers, but it also seems to be increasingly finding a home on the server side.
- Adobe Photoshop now runs in the browser using wasm and Emscripts (the C ++ toolchain for wasm). In addition to compiling from C ++ to Wasm, Emscripten also translates POSIX system calls into Web API calls and converts OpenGL to WebGL.
- JQL (JSON Query Language) is a Rust based language for querying JSON (what else?).
- Microsoft has started efforts to train 250,000 cybersecurity workers in the United States by 2025. These efforts will work with community colleges. They estimate that it will only account for 50% of the security talent shortage.
- Integrating zero trust security into the software development lifecycle is really the only way forward for companies that depend on secure and available systems.
- A supply chain attack on a Node.js library (UA-Parser-JS) installs crypto miners and Trojans to steal passwords on Linux and Windows systems. The normal function of the library is to parse user agent strings and identify the browser, operating system, and other parameters.
- A cybercrime group has set up penetration testing consultancies whose goal is to acquire customers, then gather information and launch ransomware attacks against these customers.
- A federated cryptographic system enables medical data to be shared without compromising patient privacy. This is an essential element of “predictive, preventive, personalized and participatory” medicine (also known as P4).
- The European Parliament has taken steps to ban surveillance based on biometric data, private facial recognition databases and predictive policing.
- Is it possible to reverse engineer the data on which a model was trained? An attack on a fake face generator was able to identify the original faces in the training data. This has important privacy and security implications as it seems to generalize to other types of data.
- Adversarial attacks on machine learning systems pose different cybersecurity challenges. Models are not code and have their own vulnerabilities and attack vectors. Atlas is a project to define the machine learning threat landscape. Tools for hardening machine learning models against attacks include the Adversarial Robustness Toolbox from IBM and Counterfit from Microsoft.
- Researchers have found that you can encode malware in DNA that will attack sequencing software and give the attacker control of your computer. This attack has not (yet) been found in the wild.
- Masscan is an extremely fast next generation port scanner. It’s similar to nmap, but much faster; It claims to be able to scan the entire internet in 6 minutes.
- ethr is an open source tool for cross-platform network performance measurement that was developed by Microsoft in Go. Right now it looks like the best network performance tool out there.
- Confident systems constantly monitor themselves and are able to detect (and even repair) attacks.
Infrastructure and operations
Devices and things
- Amazon is working on an internet-enabled refrigerator that will keep track of its contents and notify you when supplies are running low. (And there are similar products out there already.) Remember when that was a joke?
- Consumer-oriented AI: On the one hand, “smart gadgets” offer many challenges and opportunities. On the other hand, it needs better results than “smart” doorbells. Smart hearing aids that can be upgraded on-site as a subscription service?
- A drone was used to deliver a lung for an organ transplant. This is only the second time a drone has been used to transport organs for transplant.
- Intel has released its next generation neuromorphic processor, Loihi. Neuromorphic processors are based on the structure of the brain in which neurons send signals to each other asynchronously. While they’re still a research project, they seem to use a lot less power than traditional CPUs.
- ipleak and dnsleaktest are websites that tell you what information your browser is leaking through. They are useful tools when you are interested in protecting your privacy. The results can be scary.
- Dark design is the practice of designing user interfaces that encourage users to do things they may not want to do, whether it be agreeing to reveal information about their web usage or buying a product. Dark patterns are already common and are becoming more common.
- Black Twitter has become the new “Green Book”, a virtual place for tips on how to deal with a racist society. The original Green Book was a Jim Crow era publication that told black people where they could travel safely, what hotels they would accept, and where they would likely be victims of racial violence.
- A group at Duke University has made significant strides in correcting errors in quantum computers. You have created a “logical qubit” that has a 99.4% probability of being read correctly. (Still well below what is needed for practical quantum computing.)
- There are now two claims of quantum superiority from Chinese quantum computing projects.
- Would our response to the COVID pandemic be better if viewed as a technical problem rather than scientific research?