This simple spreadsheet of machine learning foibles may not look like much but it’s a fascinating exploration of how machines “think.” The list, compiled by researcher Victoria Krakovna, describes various situations in which robots followed the spirit and the letter of the law at the same time.
For example, in the video below a machine learning algorithm learned that it could rack up points not by taking part in a boat race but by flipping around in a circle to get points. In another simulation “where survival required energy but giving birth had no energy cost, one species evolved a sedentary lifestyle that consisted mostly of mating in order to produce new children which could be eaten (or used as mates to produce more edible children).” This led to what Krakovna called “indolent cannibals.”
It’s obvious that these machines aren’t “thinking” in any real sense but when given parameters and a the ability to evolve an answer, it’s also obvious that these robots will come up with some fun ideas. In other test, a robot learned to move a block by smacking the table with its arm and still another “genetic algorithm [was] supposed to configure a circuit into an oscillator, but instead [made] a radio to pick up signals from neighboring computers.” Another cancer-detecting system found that pictures of malignant tumors usually contained rulers and so gave plenty of false positives.
Each of these examples shows the unintended consequences of trusting machines to learn. They will learn but they will also confound us. Machine learning is just that – learning that is understandable only by machines.
One final example: in a game of Tetris in which a robot was required to “not lose” the program pauses “the game indefinitely to avoid losing.” Now it just needs to throw a tantrum and we’d have a clever three-year-old on our hands.
I’ll admit that I’ve been caught up in the Bose hype. I’ve worn Bose headphones …
It might have taken longer than expected (the event has historically taken place in March), but Apple announced its spring product unveiling event this week. Executives from the company will take to the stage in a livestream from Apple’s Cupertino headquarters, where they’ll introduce and discuss new products that Apple plans to ship in the near future.
As has become the custom, we’ll do our best to outline what you should or shouldn’t expect in terms of announcements from the unveiling.
While each Apple event’s accompanying graphic and name—this one is called “Spring Loaded,” and you can see the graphic above—can sometimes (not always) act as a sort of teaser for what kind of announcements are to come, the name doesn’t tell us much this time around. That said, the scribbled lines that make up the Apple logo above look like the work of an Apple Pencil, one of the key accessories for the iPad.
And indeed, all signs point toward this being an iPad-centric event. So let’s start there.
If there’s one product we’re relatively certain we’re going to see on April 20, it’s a new iPad Pro.
Rumor has it that the new iPad Pro will come in the same two sizes as before—11 and 12.9 inches—and that it won’t have many, if any, visible design changes. So what will be new?
Well, for starters, the new tablets are sure to include a new system-on-a-chip from Apple that will include an updated CPU, GPU, Neural Engine, and so on. We don’t know what it will be designated, though “A13X” seems like a strong possibility given past naming conventions. Reports say it will be similar in at least some ways to the M1 processor Apple put in low-end Macs last year, but we don’t know exactly what that means yet.
In any case, the Pro is sure to offer improved performance for games and creative apps. We’ll have to wait to see how big that performance bump is.
Below: Photos of the 2020 iPad Pro from our review. The new iPad Pro is expected to look similar.
Perhaps the bigger story is that the 12.9-inch iPad Pro (and not the 11-inch, apparently) will have a new display technology: Mini LED. It will still have an LCD screen, but this new approach will enable better contrast, among other improvements. Mini LED has already started shipping in high-end 4K TVs.
Supplies of Mini LED panels, however, may be limited, so the 12.9-inch model might be hard to get a hold of for a while if this rumor is true.
Apple may also choose to update the cameras in the iPad Pro, and a third-generation Apple Pencil accessory is not outside the realm of possibility. In fact, that’s the only possible hint we can imagine coming from the event’s name—there have been some rumors of spring-loaded components finding their way into future Pencil designs.
The iPad mini hasn’t been updated in a while, and it shows. The design of the product is ancient at this point, with enormous bezels wasting a significant amount of the device’s limited space instead of the better screen-to-body ratio seen in other recent iPads. The story is better with regards to the internals, but it’s still worth noting that the A12 chip included in the device will be 3 years old this year.
Given that Apple has more recently updated the other iPads in the lineup, the iPad mini is the only non-Pro iPad we see as a possibility for this event. That said, we’re not sure exactly what to expect from it.
Below: The current iPad mini, from our most recent review. Those are some chunky bezels by today’s standards.
If Apple does update it, the mini will probably have a more recent chip—most likely the A14 we saw in last year’s iPhone lineup. But what we’re really hoping for is a redesign that increases the screen size without changing the actual footprint of the device.
It’s also possible that Apple could add support for newer, better version of the Apple Pencil, as the current mini only supports the previous iteration.
Apple Silicon Macs
We believe this event will focus primarily on the iPad, but Apple did commit last summer to update its entire Mac product line to replace Intel’s chips with Apple Silicon within two years from that date. We’re coming up on one year in, and so far Apple has just updated the low-end configurations of the MacBook Air, MacBook Pro, and Mac mini.
Various reports have claimed that Apple plans to introduce higher-end MacBook Pro models and a new iMac this year, with the Mac Pro desktop tower getting an Apple Silicon refresh sometime in 2022. There’s also talk of a redesigned MacBook Air, though that computer already got the M1 treatment last year.
Most of those reports have said that the MacBook Pro and Air are slated for the second half of the year, so if we see any new Mac at this event (that’s definitely not a given, to be clear), it might be a new iMac.
Apple is believed to be working on a totally redesigned iMac with more screen real estate, a different body shape, Apple Silicon, and other improvements.
Below: The M1-equipped Mac mini from our review last year.
We’re just speculating here, but it’s possible that Apple will do the same thing it did with the laptops last year: start with an M1-equipped lower-end iMac before bringing out a much faster high-end model with an “M1X”-type chip alongside the new MacBook Pro models later. But we don’t know for sure.
It seems worth mentioning that Apple just discontinued the iMac Pro, which you might read as a hint that a big, high-end update for the standard iMac is right around the corner to take its place. But we’re not totally confident in that reading, because the current specced-out standard Intel iMac already matches or beats the languishing iMac Pro.
Whatever form it takes, a new iMac might be accompanied by a new Apple-designed display that would offer some of the benefits of the professional studio-targeted Pro Display XDR but at a more consumer-friendly price point. Apple used to design and sell excellent consumer desktop monitors, but for the past few years, it has loosely partnered with LG, which makes Mac-focused monitors that are variants of the South Korean company’s other products.
We don’t know what a new Apple monitor would look like at this stage, but we’re hoping it will be a triumphant return on that front.
The Rich Communication Services (RCS) rollout continues to be a hopeless disaster. A year and a half ago, the cellular carriers created the “Cross-Carrier Messaging Initiative (CCMI),” a joint venture between AT&T, Sprint, T-Mobile, and Verizon that would roll out enhanced messaging to the masses in 2020. Now, Light Reading is reporting that initiative is dead, meaning that the carriers have accomplished basically nothing on the RCS front in the past 18 months.
RCS is a carrier-controlled GSMA standard introduced in 2008 as an upgrade for SMS, the ancient standard for basic carrier messaging. SMS (which started in 1992!) has not kept up with the feature set of over-the-top messaging services like WhatsApp, Facebook Messenger, and iMessage, and while RCS still wouldn’t be able to keep up with services like those, it can bring slightly more messaging functionality to carrier messaging. RCS includes things like typing indicators, presence information, read receipts, and location sharing.
Verizon confirmed the news to Light Reading, saying, “The owners of the Cross Carrier Messaging Initiative decided to end the joint venture effort. However, the owners remain committed to enhancing the messaging experience for customers including growing the availability of RCS.”
What is the motivation for RCS?
With the carriers in charge of RCS, everything about the rollout has moved at an absolutely glacial pace. The problem is that there’s no motivation for carriers to actually roll out RCS: free messaging is the norm, so there’s no clear way to make money off an RCS rollout. Even if you could snap your fingers and instantly make every phone on every carrier RCS compatible today, it still would not be a viable competitor to an over-the-top service. RCS is a decade-old specification, and it feels like it—the spec lacks things you would want in a modern messaging app, like encryption.
RCS’s second major problem is Apple, which will never support RCS unless the company has a major change of strategy. Earlier this month, the Epic Games lawsuit revealed internal Apple communication that made it clear the company views iMessage’s exclusion of Android users as a competitive advantage, and RCS would poke holes in the walls of Apple’s walled garden.
RCS also has all the standard carrier-centric problems you get from phone number-driven products like Google Allo and the new Google Pay, which envision your phone number as your online identity and the center of your communication universe. Like those other products, RCS lacks robust multi-device support for devices that don’t have a phone number, like laptops, desktops, tablets, and watches.
Less opposition to Google’s RCS efforts?
Speaking of Google, the company is actually the biggest player in RCS messaging, thanks to its purchase of Jibe—a middleware company offering RCS solutions to carriers—in 2015. Google killed Google Allo (Google messaging app from 2016) in 2018, with the plan to push RCS over Google Messages (Google messaging app from 2014). The carriers’ creation of the Cross Carrier Messaging Initiative (CCMI) was seen as a snub of Google’s RCS plans at the time, and in response Google started rolling out RCS without the carriers by enabling RCS in the UK and France, provided both users were on Google Messages and had the “Chat” setting turned on. The whole point of RCS is its default-ness, though, as a lowest-common-denominator replacement for SMS, so it didn’t make a ton of sense to have it as an optional extra in Google’s messaging app. Google has been able to experiment with delivering services on top of RCS, like end-to-end encryption, provided both users are on Google Messages (and for one-on-one messages only).
Google’s plans started making a bit more sense when it signed a deal with T-Mobile in May 2020, which made Google Messages the company’s default SMS/RCS app across all Android devices. As Light Reading notes, Google doesn’t manage T-Mobile’s backend with Jibe (T-Mobile uses a company called Mavenir), but T-Mobile can sidestep the CCMI and push a standard RCS messaging app now. T-Mobile has always been the easiest carrier to work with, though. Verizon said it still wants to pursue RCS, but it declined to say how, while AT&T didn’t return any of Light Reading’s questions about what’s next. It all sounds just like the usual status quo that has resulted in the past 10+ years of RCS nothingness.
On one hand, the founding of the CCMI in 2019 and the dissolution of it in 2021 means the carriers have accomplished absolutely nothing on the RCS front for the past year and a half. On the other hand, this could mean there is less opposition to Google’s RCS platform, which is a working, functional solution. Historically, though, the right move has been to bet against RCS, and we’ll most likely continue to see nothing happen.
Last year’s Oculus Quest 2 VR headset remains one of the cheapest—though not necessarily recommended—ways to jump into virtual reality. But even I must admit its sales proposition became more tantalizing on Tuesday with a late-night announcement from reps at Facebook: two disabled features inside the headset are now being unlocked as a default option.
The first is a wireless-VR mode, which Facebook is calling Oculus Air Link, coming “soon” to headset-and-PC combos that run compatible Oculus software. The short version: you will soon be able to connect your Oculus Quest 2 to a gaming PC using nothing more than a local Wi-Fi connection. This feature will be supported within stock headset software, no extra apps required. And it will essentially make connecting to your PC’s VR apps work the same as the VR apps built directly into Quest 2’s storage.
“Not every network and PC setup will be ideal”
“We know gamers want to use Link without a wire,” the announcement says, and sure enough, that cry tends to be the loudest in our VR hardware reviews. No more wires in VR, the readers complain, and Facebook has responded with no more wires. But, gosh, do you really want to use this feature, folks?
Keep in mind that Oculus Air Link is “experimental.” As Facebook says, “Not every network and PC setup will be ideal.” Connecting your headset to a PC via a cable (as in, Oculus Link) is “the way to go” for most users and offers the “highest fidelity visuals possible.” Heck, its “known issues” list says that AMD GPUs can only wirelessly stream via Air Link at half the rate of Nvidia GPUs, even if you have AMD’s newest, highest-end products.
That abundance of caution for average users isn’t surprising, since wireless VR runs up against a significant burden of comfort and fidelity. If your local network can’t consistently deliver 72 fps or 90 fps of high-res images directly to your face, any blurriness or control lag can feel all the more severe. Some Oculus Quest owners already know this because they’ve tested wireless-VR modes as enabled through the third-party Virtual Desktop app, which has always required jumping through at least one hoop to get it working.
Still, with the right network conditions, Virtual Desktop has proven Quest 2’s ability to stream higher-end PC VR games to the cheaper Oculus Quest 2 with acceptable performance. And having those features built directly into the firmware could prove to be even more efficient—though Oculus’ notes suggest a maximum of 200Mbps of upstream-and-downstream via local wireless networks, which is far lower than Virtual Desktop’s maximum of 1,200Mbps. We look forward to testing and comparing the two options.
120 Hz: “Soon,” but when? And for what?
The other big-deal feature announced on Tuesday is Quest 2’s panel jumping to a whopping 120 Hz refresh, up from its current maximum of 90 Hz. As it turns out, Quest 2’s single LCD panel was rated for 120 Hz refresh rates all along—meaning it was likely sourced from production lines that were making displays at the same speed as a new standard for smartphone screens.
After a tease from one Facebook executive in February, longtime Oculus contributor John Carmack confirmed in March that the feature would eventually arrive. “Only a few existing games will be tweaked for 120 [Hz], but some new titles will consider it an option in their design phase,” he wrote on Twitter.
Facebook’s official 120 Hz announcement confirms this plan, albeit in different language: “Not many apps will support 120 Hz just yet,” according to the statement, and it won’t apply to the hardware’s default “home” environment. There’s no release date beyond “soon” for this feature’s rollout. When asked by Ars Technica, Facebook declined to offer a list or hints of what existing software may receive a 120 Hz refresh update at that point.
In Quest 2’s default use case, as a wholly wireless headset running internally installed software, 120 Hz mode may have limited impact. Quest 2 hardware is already pushed pretty hard by 90 Hz speeds, which is why many Quest 2 games, including the wildly popular and Facebook-owned Beat Saber, stick to its lowest 72 Hz refresh rate. Jumping further not only cranks the SoC (and its cooling system) that much more but will also hammer the system’s already capricious battery life.
Dreaming of updates for PC, plus productivity
As a connected PC-VR option, on the other hand, 120 Hz mode could be a serious treat, especially for PCs that are equipped to run VR games at such speeds. In particular, higher refresh rates seriously impact long-term VR comfort when sessions exceed 30 minutes at a time. My earliest tests of the Valve Index, which natively supports 120 Hz and 144 Hz modes, hinged on using that headset as a virtual work monitor for hours at a time, and what I said at the time still holds: higher refresh rates make juggling multiple, floating work screens and panels all the easier on the eyes.
But even though Facebook is vague about 120 Hz modes for default Oculus Quest 2 use, it’s somehow even more vague about the same coming to connected PC-VR (aka Oculus Link). That’s coming in a “future release,” as opposed to “soon.” I wish it were the other way around, assuming one will take longer than the other.
Everything I said above about using VR to run a virtual office is clearly on Facebook’s mind, as well, as Tuesday’s blog post included hints of productivity boosts built into the Quest 2’s “home” interface. Among those: the headset will soon allow you to place “a virtual desk on your real furniture,” and it will start adding support for “Bluetooth-enabled keyboard tracking” while inside of VR.
This will require a compatible keyboard to start, with Logitech’s K830 being the first supported model. Facebook’s latest sample GIF shows a 3D-rendered keyboard appearing in your virtual world, along with a black-and-white glimmer of your real-life hands typing on it. Go back and forth between tapping on the keyboard and gesturing in mid-air with your fingers to control VR windows and interfaces like a mouse.
This is clearly Facebook building upon its finger-tracking system, which launched as a beta within the Oculus Quest 2 firmware in late 2019, and it’s a good hint of the company’s aspirations to make VR part of a balanced work-from-home diet—even if such features feel entirely too late to the pandemic party. I’ve gotten in touch with Logitech about the K830 and plan to test it for a future article about whether Quest 2 might fit my remote-office needs. Why buy a zillion monitors when a single VR headset, as paired with smart office hardware, could produce them virtually for cheaper? (Albeit with Facebook’s embedded Oculus cameras watching the entire time.) I’ll test and follow up.