Cataloging the problems facing AI researchers is a cross between a parenting manual and a management book

animation (1)

Concrete Problems in AI Safety, an excellent, eminently readable paper from a group of Google AI researchers and some colleagues, sets out five hard problems facing the field: robots might damage their environments to attain their goals; robots might figure out how to cheat to attain their goals; supervising robots all the time is inefficient; robots that are allowed to try novel strategies might cause disasters; and robots that are good at one task might inappropriately try to apply that expertise to another unrelated task. Read the rest

Algorithms to Live By: what computer science teaches us about everyday decisions

algorithms-to-live-by-brain
Brian Christian and Tom Griffiths' Algorithms to Live By: The Computer Science of Human Decisions is pitched as a combination of personal advice and business book grounded in the lessons of computer science, but it's better than that: while much of the computer science they explain is useful in personal and management contexts, the book is also a beautifully accessible primer on algorithms and computer science themselves, and a kind of philosophical treatise on what the authors call "computational kindness" and "computational stoicism."

Let's teach programming as a tool for analyzing data to transform the world

Globaloria_students_working (1)

Data-scientist Kevin H Wilson argues that computers are tools for manipulating data -- from companies' sales data to the input from games controllers -- but we teach computer programming as either a way to make cool stuff (like games) or as a gateway to "rigorous implementation details of complicated language," while we should be focusing on fusing computer and math curriciula to produce a new generation of people who understand how to use computers to plumb numbers to find deep, nuanced truths we can act upon. Read the rest

Intel x86s hide another CPU that can take over your machine (you can't audit it)

PIC12C508-HD
Recent Intel x86 processors implement a secret, powerful control mechanism that runs on a separate chip that no one is allowed to audit or examine. When these are eventually compromised, they'll expose all affected systems to nearly unkillable, undetectable rootkit attacks. I've made it my mission to open up this system and make free, open replacements, before it's too late.

Emojibot uses deep learning to synthesize expressive new nonverbal communications

emojiland2

Dango is a personal assistant that feeds its users' messages into a deep-learning neural net to discover new expressive possibilities for emojis, GIFs and stickers, and then suggests never-seen combinations of graphic elements to your text messages that add striking nuances to them. Read the rest

Twitterbot that produces endless entries in an imaginary daemonological grimoire

Ckge6x1UYAENpbN

The Lesser Bot is a twitterbot that is writing a machine-generated grimoire, complete with summoning runes, which is timely, given that we're entering the age of demon-haunted computers. Read the rest

Award-winning short sf film written by an AI is pretty good

056c026d-1c66-4d42-9fae-a8e96df290c5-1020x1089

Director Oscar Sharp and AI researcher Ross Goodwin trained a machine-learning system with a huge pile of classic science fiction screenplays and turned it loose to write a short film. What emerged was an enigmatic 9-minute movie called Sunspring, which has just won Sci-Fi London's 48-hour challenge. Read the rest

Password hashing demystified

1200px-Double-alaskan-rainbow

The password breaches are getting stronger and worser, and hardly a week goes by without a dump that's a couple zeroes bigger than the biggest to date -- but not all password breaches are created equal, and a lot depends on whether and how the passwords were hashed. Read the rest

Deep learning AI "autoencodes" Blade Runner, recreates it so faithfully it gets a takedown notice

1-LNPjTuTmTW3Wzeg-3RzlZg

Artist and researcher Terence Broad is working on his master's at Goldsmith's computing department; his dissertation involved training neural networks to "autoencode" movies they've been fed. Read the rest

EFF: FBI & NIST's tattoo recognition program exploited prisoners, profiled based on religion, gave sensitive info to private contractors

why_tattoos_slide

Dave Maass from EFF says, "Right now, NIST researchers are working with the FBI to develop tattoo recognition technology that police can use to learn as much as possible about people through their tattoos. But an EFF investigation has found that these experiments exploit inmates, with little regard for the research's implications for privacy, free expression, religious freedom, and the right to associate. And so far, researchers have avoided ethical oversight while doing it." Read the rest

Generative, collaging architecture system designs impossible, Inception-like cities

26639435895_cfbc2e876c_b

London's Daniel Brown created a generative design system that designs beautiful, brutalist cityscapes that are part Blade Runner Hong Kong, part Inception; he then manually sorts through the results, picks the best, and publishes them in a series called "Travelling by Numbers." Read the rest

Jury hands Oracle its ass, says Google doesn't owe it a penny for Java

056c026d-1c66-4d42-9fae-a8e96df290c5-1020x1069

When Oracle bought Sun Microsystems, they acquired Java, Sun's popular programming language, pitched from its inception as an open standard for the networked computer. Read the rest

Tor Project is working on a web-wide random number generator

hs_montreal_4

Random number generators are the foundation of cryptography -- that's why the NSA secretly sabotaged the RNG standard that the National Institute for Standards and Technology developed. Read the rest

Pastejacking: using malicious javascript to insert sneaky text into pasted terminal commands

DEC_VT100_terminal

When a computer stops behaving, the solution often involves looking up an obscure command and pasting it into the terminal -- even experienced administrators and programmers aren't immune to this, because remembering the exact syntax for commands you use once every couple years is a choresome task. Read the rest

Programmers' stress levels can accurately predict the quality of their code

security-1163108_960_720

In Using (bio)metrics to predict code quality online, presented at the ACM's 38th International Conference on Software Engineering, two Swiss researchers presented their work on monitoring programmers' biometrics to predict the quality of the code they were writing.

Read the rest

A software developer's version of the CIA's bureaucratic sabotage manual

What_is_sabotage^_Sabotage_is_treason^_-_NARA_-_535191

The Simple Sabotage Field Manual was published in 1944 by the Office of Strategic Services, the agency that came to be the CIA: it outlined simple tactics for putting bureaucratic grit in the wheels of occupied countries, for example, by referring key decisions to committees and then obstructing the work of those committees. Read the rest

Algorithmic cruelty: when Gmail adds your harasser to your speed-dial

inbox-speed-dial

Inbox by Gmail combs through your email looking for frequent correspondents and puts the people who email you the most in a "speed dial" sidebar (that you can't edit) that puts their names and pictures front-and-center for you every time you go to your email. Read the rest

More posts