Up to a dozen assignment expressions per TC definition sounds reasonable. In practice it depends on the layer types. There are basically two scaling limiters: compilation and especially autotuning time grows very fast as a function of a number of expressions; the amount of exploitable parallelism may decrease with the increasing number of inter-dependent operations. One of the goals of TC is to make it easy to define new layers and look at practically achieved performance. It should be as easy as moving the line between two TC defs and changing input/output tensor lists.

Music streaming giant Spotify is cracking down on users who employ apps to bypass restrictions and access a modified version of its premium services without paying. The Swedish company, which last week filed for a public listing with the US Securities and Exchange Commission, has begun emailing listeners it believes to be using illegal software saying it has "detected abnormal activity" and warning that their accounts have been deactivated and will not be restored until the offending programmes have been uninstalled. "If we detect repeated use of unauthorised apps in violation of our terms, we reserve all rights including suspending or terminating your account," non-paying subscribers are cautioned. The free version of Spotify has 88m subscribers but inserts audio adverts every three tracks and ensures songs can only be played on shuffle mode, conditions the premium service removes at a cost of £9.99 a month. The apps in question do not provide the high-quality streaming a premium account would offer but do allow users to sidestep the adverts and restrictions that many find annoying.

Tumblr has been banned in Indonesia for providing people with access to pornographic content. The communications ministry said the block – which means nobody in the country can access the blogging platform – came after complaints about hundreds of accounts that provided pornography. The country has stepped up efforts to police the kind of content available online. It has been trying to stem a rise in hoax stories and hate speech, and anti-pornography laws pushed by Islamic parties in the country have led tens of thousands of adult websites to be taken offline. The I.F.O. is fuelled by eight electric engines, which is able to push the flying object to an estimated top speed of about 120mph.

Kiwi startup Thematic uses machine learning technology to make sense of customer feedback forms. One of its customers is Sky TV. When you see Artificial Intelligence in the news, the topic is typically robots taking over our jobs or the dangers of super-intelligent computers. You don't see Elon Musk or Bill Gates opining on how AI is improving the way we do data analytics, for example. But Kiwi startup Thematic is doing exactly that. It's using machine learning technology to make sense of customer feedback forms.

Even if you don't care for the actual song, the music video for Hardcore Anal Hydrogen's track Jean-Pierre is a visual feast, taking advantage of several recent breakthroughs in artificial-intelligent image processing that's quickly gone from creating nightmares, to revolutionizing how modern visual effects are created. The behind-the-scenes processes that went into making this music video are just as fascinating as its visuals, so Hardcore Anal Hydrogen shared a lengthy breakdown of all the tools they used, including AI software like Deep Dream, Neural Style Transfer where artistic styles can be copied from other pieces, and Optical Flow, which creates seamless transitions between the processed frames.

Apple planted its HomePod smart speaker deep inside a walled garden: You must have an Apple-branded mobile device just to set it up, and you can pretty much use it only with Apple's own services if you want voice control for music. The company then adds insult to injury by leaving its older iPhones, iPads, and iPods outside that garden wall. Unless you're an ardent Apple fan who upgrades to new iOS devices with relative frequency, you won't even be able to set up a HomePod. If you own an iPhone, iPad, or iPod model that Apple essentially rendered obsolete last fall when it established the hardware requirements for iOS 11, you're out of luck. You can stream music from various devices to a HomePod using AirPlay, but not until you've set up the speaker using an iOS device running iOS 11 or higher.

AI is also generating its own content with little or no human supervision. AI that generates text is widespread in journalism and used by publishers to automate some news production. The Associated Press has used AI to free up around 20% of reporters' time while increasing output tenfold. The Washington Post developed a tool to cover sports and political news, producing about 70 articles a month in its first year. In music, AI has produced instrumental sounds that humans have never heard before.

Spike Jonze essentially makes two types of commercials. There's the distraction model, in which a frenetic, unrelated-to-the-product conceit takes over the ad, like his KENZO World commercial from 2016 starring Margaret Qualley and the Dorothy Chandler Pavillion: In this category, we can put his ad for Levi's fly weight jeans (like the KENZO World ad, it's mostly about choreography) as well as that Nissan Frontier commercial that is mostly about a dog pushing a recliner through traffic. Then there are the Spike Jonze ads that would be anti-capitalist if they weren't ads. See, e.g., his IKEA ad, which made fun of viewers for becoming emotionally attached to products (while simultaneously encouraging them to buy more products): Or his Gap ad, which harnessed people's bottled up rage at monotonous chain retail stores (while simultaneously encouraging them to shop at the Gap): His newest ad for Apple HomePod combines the two forms: it's a strikingly beautiful short film, in which the Apple HomePod transforms FKA Twigs' depressing apartment into a rainbow-striped wonderland by choosing exactly the right song at exactly the right time. It also makes owning an Apple HomePod seem supremely depressing and possibly dangerous on an interdimensional level.

In this special report on the future of artificial intelligence, we explore the technology's implications. Are people ready to trust their lives to driverless cars? Who's to blame when price-setting algorithms work together to collude?

Like a film critic asked if the Oscars got it right this year, one has to feel a sense of standing too close to the frame, the field of vision too narrow to provide the context necessary for proper judgment. After spending an afternoon among the various installations that comprise "Thinking Machines: Art and Design in the Computer Age: 1959–1989," I wonder if this anxiety applied to the team tasked with creating this exhibit. In this case, I think not. Here, closeness to the frame is a virtue, not a vice. The thread running through the exhibit's diverse elements is both lucid and enjoyable: celebrating the collaborative possibilities between human artists and computational machines, and the energies unleashed when the humanistic meet the systematic.