Canadian camouflage manufacturer Hyperstealth Biotechnology has applied for patents on its ‘Quantum Stealth’ material.

“The reason no government wants you to think about all this is because there is no serious humanitarian response possible to a nuclear explosion. There’s no way to really help the immediate victims of a nuclear attack. This is not a hurricane, wildfire, earthquake, or nuclear accident — it is all of these things at once, but worse. No nation on earth is prepared to deal with it.

“This is a regularly updated guide to abstract strategy games: what they are, which are the best, and where to learn, play, buy, and discuss them.

This GPT-2 based text generator, which can generate relatively plausible-sounding text from a short prompt, has been upgraded with a larger model. Would you even notice if I let the generator write the rest of this link description? 

“Over the space of two months Pavur sent out 150 GDPR requests in his fiancée's name, asking for all and any data on her. Of the responses, 24 per cent simply accepted an email address and phone number as proof of identity and sent over any files they had on his fiancée. A further 16 per cent requested easily forged ID information.

 Excellent and comprehensive C++ guide.

Claims to be "the world’s best machine translation".

 Summaries for hundreds of non-fiction books.

“OpenAI's GPT-2 scored a 70.7% back in February, and that was 7 points ahead of second place. At the time, that was a scarcely believable jump forward in the state of the art and one reason why everyone was so amazed by the network.

So a 70.7% success rate is great compared to what AI could do. It's still severely crippled compared to what humans can do. The human success rate for the Winograd Schema challenge is 95%.

Microsoft's latest endeavor, however, completely smashes through everything we thought we could do, resting at about 89%. This is unbelievable.

“We’ve created MuseNet, a deep neural network that can generate 4-minute musical compositions with 10 different instruments, and can combine styles from country to Mozart to the Beatles. MuseNet was not explicitly programmed with our understanding of music, but instead discovered patterns of harmony, rhythm, and style by learning to predict the next token in hundreds of thousands of MIDI files. MuseNet uses the same general-purpose unsupervised technology as GPT-2, a large-scale transformer model trained to predict the next token in a sequence, whether audio or text.

An interactive demonstration of various strategies for the prisoner's dilemma, and the implications for trust in general.

“Next year, we’re aiming to donate fifteen percent of our income instead of ten; by coincidence, that means we would be able to save fifteen children instead of ten. Half a classroom. You’re a hero, if you rescue half a classroom from a fire.

And we can do that next year, and the year after that, and the year after that.

(...)

Whenever that voice in my head that talks about how I’m a worthless stupid failure who doesn’t deserve to exist gets too loud, I count up the children I’ve saved.

“We’ve trained a large-scale unsupervised language model which generates coherent paragraphs of text, achieves state-of-the-art performance on many language modeling benchmarks, and performs rudimentary reading comprehension, machine translation, question answering, and summarization — all without task-specific training.