Technology

Few things in the world of technology can really ever be said to be &done,& and certainly exoskeletons are not among their number. They exist, but they are all works in progress, expensive, heavy, and limited. So itgreat to see this team working continuously on their TWIICE robotic wearable, improving it immensely with the guidance of motivated users.

TWIICE made its debut in 2016, and like all exoskeletons it was more promise made than promise kept. Ita lower-half exoskeleton that supports and moves the legs of someone with limited mobility, while they support themselves on crutches. Itfar from ideal, and the rigidity and weight of systems like this make them too risky to deploy at scale for now.

But two years of refinement have made a world of difference. The exoskeleton weighs the same (which doesn&t matter since it carries its own weight), but supports heavier users while imparting more force with its motors, which have been integrated into the body itself to make it far less bulky.

Perhaps most importantly, however, the whole apparatus can now be donned and activated by the user all by herself, as Swiss former acrobat and now handcycling champion Silke Pan demonstrated in a video. She levers herself from her wheelchair into the sitting exoskeleton, attaches the fasteners on her legs and trunk, then activates the device and stands right up.

She then proceeds to climb more stairs than I&d rather attempt. She is an athlete, after all.

That kind of independence is often crucially important for the physically disabled for a multitude of reasons, and clearly achieving the capability has been a focus for the TWIICE team.

Although the exoskeleton has been worked on as a research project within the Ecole Polytechnique Federale de Lausanne (EPFL), the plan is to spin off a startup to commercialize the tech as it approaches viability. The more they make and the more people use these devices — despite their limitations — the better future versions will be.

At this moment in history itimpossible not to see the problems that arise from humanbias. Now magnify that by compute and you start to get a sense for just how dangerous humanbiasvia machine learning can be. The damage can be twofold:

• Influence. If theAIsaid so it must be true… people trust outputs ofAI,so if humanbiasis missed in the training it could compound the problem by infecting more people;
• Automation. SometimesAImodels are plugged into a programmatic function, which could lead to the automation ofbias.

But there is potentially a silver machine-learned lining. BecauseAIcan help expose truth inside messy data sets, itpossible for algorithms to help us better understandbiaswe haven&t already isolated, and spot ethically questionable ripples in human data so we can check ourselves. Exposing human data to algorithms exposesbias,and if we are considering the outputs rationally, we can use machine learningaptitude for spotting anomalies.

But the machines can&t do it on their own. Even unsupervised learning is semi-supervised, as it requires data scientists to choose the training data that goes into the models. If a human is the chooser,biascan be present. How the heck do we tackle such abiasbeast We will attempt to pick it apart.

The landscape of ethical concerns withAI

Bad examples abound. Consider thefindingfrom Carnegie Mellon that showed that women were shown significantly fewer online ads for high-paying jobs than men were. Orrecall the sad case of Tay, Microsoftteen slang Twitter bot that had to be taken down after producing racist posts.

In the near future, such mistakes could result in hefty fines or compliance investigation, a conversation thatalreadyoccurring in the U.K. parliament. All mathematicians and machine learning engineersshould considerbiasto some degree, but that degree varies from instance to instance. A small company with limited resources will often be forgiven for accidentalbiasas long as the algorithmic vulnerability is fixed quickly; a Fortune 500 company, which presumably has the resources to ensure an unbiased algorithm, will be held to a tighter standard.

Of course, an algorithm that recommends novelty T-shirts does not need nearly as much oversight as an algorithm that decides what dose of radiation to give to a cancer patient. Itthese high-stakes decisions that will become the most pronounced when legal liability enters the discussion.

Itimportant for builders and business leaders to establish a process for monitoring the ethical behavior of theirAIsystems.

Three keys to managing biaswhen building AI

There are signs of existing self-correction in theAIindustry: Researchers arelooking at waysto reducebiasand strengthen ethics in rule-based artificial systems by taking human biases into account, for example.

These are good practices to follow; itimportant to be thinking proactively about ethics regardless of the regulatory environment. Lettake a look at several points to keep in mind as you work on yourAI.

1.Choose the right learning model for the problem.

Therea reason allAImodels are unique: Each problem requires a different solution and provides varying data resources. Thereno single model to follow that will avoidbias, but there are parameters that can inform your team as itbuilding.

For example, supervised and unsupervised learning models have their respective pros and cons. Unsupervised models that cluster or do dimensional reduction can learnbiasfrom their data set. If belonging to group A highly correlates to behavior B, the model can mix up the two. And while supervised models allow for more control overbiasin data selection, that control can introduce humanbiasinto the process.

Non-biasthrough ignorance — excluding sensitive information from the model — may seem like a workable solution, but it still has vulnerabilities. In college admissions, sorting applicants by ACT scores is standard, but taking their ZIP code into account might seem discriminatory. But because test scores might be affected by the preparatory resources in a given area, including the ZIP code in the model could actually decreasebias.

You have to require your data scientists to identify the best model for a given situation. Sit down and talk them through the different strategies they can take when building a model. Troubleshoot ideas before committing to them. Itbetter to find and fix vulnerabilities now — even if it means taking longer — than to have regulators find them later on.

2. Choose a representative training data set.

Your data scientists may do much of the leg work, but itup to everyone participating in anAIproject to actively guard againstbiasin data selection. Therea fine line you have to walk. Making sure the training data is diverse and includes different groups is essential, but segmentation in the model can be problematic unless the real data is similarly segmented.

Itinadvisable — both computationally and in terms of public relations — to have different models for different groups.When there is insufficient data for one group, you could possibly use weighting to increase its importance in training, but this should be done with extreme caution. It can lead to unexpected new biases.

For example, if you have only 40 people from Cincinnati in a data set and you try to force the model to consider their trends, you might need to use a large weight multiplier. Your model would then have a higher risk of picking up on random noise as trends — you could end up with results like &people named Brian have criminal histories.& This is why you need to be careful with weights, especially large ones.

3. Monitor performance using real data.

No company is knowingly creating biasedAI, of course — all these discriminatory models probably worked as expected in controlled environments. Unfortunately, regulators (and the public) don&t typically take best intentions into account when assigning liability for ethical violations. Thatwhy you should be simulating real-world applications as much as possible when building algorithms.

Itunwise, for example, to use test groups on algorithms already in production. Instead, run your statistical methods against real data whenever possible. Ask the data team to check simple test questions like &Do tall people default onAI-approved loans more than short people& If they do, determine why.

When you&re examining data, you could be looking fortwo types of equality: equality of outcome and equality of opportunity. If you&re working onAIfor approving loans, result equality would mean that people from all cities get loans at the same rates; opportunity equality would mean that people whowould have returned the loan if given the chance are given the same rates regardless of city.Without the latter, the former could still hide if one city has a culture that makes defaulting on loans common.

Result equality is easier to prove, but it also means you&ll knowingly accept potentially skewed data.While itharder to prove opportunity equality, it is at least valid morally. Itoften practically impossible to ensure both types of equality, but oversight and real-world testing of your models should give you the best shot.

Eventually, these ethicalAIprinciples will be enforced by legal penalties. IfNew York Cityearly attemptsat regulating algorithms are any indication, those laws will likely involve government access to the development process, as well as stringent monitoring of the real-world consequences ofAI. The good news is that by using proper modeling principles,biascan be greatly reduced or eliminated, and those working onAIcan help expose accepted biases,create a more ethical understanding of tricky problemsand stay on the right side of the law — whatever it ends up being.

The holiday season is here again, touting all sorts of kids& toys that pledge to pack ‘STEM smarts& in the box, not just the usual battery-based fun.

Educational playthings are nothing new, of course. But, in recent years, long time toymakers and a flurry of new market entrants have piggybacked on the popularity of smartphones and apps, building connected toys for even very young kids that seek to tap into a wider ‘learn to code& movement which itself feeds off worries about the future employability of those lacking techie skills.

Whether the lofty educational claims being made for some of these STEM gizmos stands the test of time remains to be seen. Much of this sums to clever branding. Though thereno doubt a lot of care and attention has gone into building this category out,you&ll also find equally eye-catching price-tags.

Whatever STEM toy you buy therea high chance it won&t survive the fickle attention spans of kids at rest and play. (Even as your childrenappetite to be schooled while having fun might dash your ‘engineer in training& expectations.) Tearing impressionable eyeballs away from YouTube or mobile games might be your main parental challenge — and whether kids really need to start ‘learning to code& aged just 4 or 5 seems questionable.

Buyers with high ‘outcome& hopes for STEM toys should certainly go in with their eyes, rather than their wallets, wide open. The ‘STEM premium& can be steep indeed, even as the capabilities and educational potential of the playthings themselves varies considerably.

At the cheaper end of the price spectrum, a ‘developmental toy& might not really be so very different from a more basic or traditional building block type toy used in concert with a kidown imagination, for example.

While, at the premium end, there are a few devices in the market that are essentially fully fledged computers — but with a child-friendly layer applied to hand-hold and gamify STEM learning. An alternative investment in your childfuture might be to commit to advancing their learning opportunities yourself, using whatever computing devices you already have at home. (There are plenty of standaloneappsofferingguided coding lessons, for example. And tons and tons of open source resources.)

For a little DIY STEM learning inspiration read this wonderful childhood memoirby TechCrunchvery own John Biggs — a self-confessed STEM toy sceptic.

Italso worth noting that some startups in this still youthful category have alreadypivoted more toward selling wares direct to schools — aiming to plug learning gadgets into formal curricula, rather than risking the toys falling out of favor at home. Whichdoes lend weight to the idea that standalone ‘play to learn& toys don&t necessarily live up to the hype. And are getting tossed under the sofa after a few days& use.

We certainly don&t suggest there are any shortcuts to turn kids into coders in the gift ideas presented here.Itthrough proper guidance — plus the power of their imagination — that the vast majority of children learn. And of course kids are individuals, with their own ideas about what they want to do and become.

The increasingly commercialized rush towards STEM toys, with hundreds of millions of investor dollars being poured into the category, might also be a cause for parental caution. Therea risk of barriers being thrown up to more freeform learning — if companies start pushing harder to hold onto kids& attention in a more and more competitive market. Barriers that could end up dampening creative thinking.

At the same time (adult) consumers are becoming concerned about how much time they spend online and on screens. So pushing kids to get plugged in from a very early age might not feel like the right thing to do. Your parental priorities might be more focused on making sure they develop into well rounded human beings — by playing with other kids and/or non-digital toys that help them get to know and understand the world around them, and encourage using more of their own imagination.

But for those fixed on buying into the STEM toy craze this holiday season, we&ve compiled a list of some of the main players, presented in alphabetical order, rounding up a selection of what they&re offering for 2018, hitting a variety of price-points, product types and age ranges, to present a market overview — and with the hope that a well chosen gift might at least spark a few bright ideas…

Adafruit Kits

Product: Metro 328 Starter PackPrice: $45Description: Not a typical STEM toy but a starter kit from maker-focused and electronics hobbyist brand Adafruit. The kit is intended to get the user learning about electronics and Arduino microcontrollers to set them on a path to being a maker. Adafruit says the kit is designed for &everyone, even people with little or no electronics and programming experience&. Though parental supervision is a must unless you&re buying for a teenager or mature older child. Computer access is also required for programming the Arduino.

Be sure to check out Adafruit&sYoung Engineers Categoryfor a wider range of hardware hacking gift ideas too, from $10 for a Bare Conductive Paint Pen, to $25 for theDrawdio fun pack, to $35 for thisKonstruktor DIY Film Camera Kit or $75 for theSnap Circuits Green kit— where budding makers can learn about renewable energy sources by building a range of solar and kinetic energy powered projects. Adafruit also sells a selection of STEM focused childrenbooks too, such asPython for Kids($35)Age: Teenagers, or younger children with parental supervision

[inline-ads]

Anki

Product: CozmoPrice: $180Description:The animation loving Anki team added a learn-to-code layer to their cute, desktop-mapping bot last year — called Cozmo Code Lab, which was delivered via free update — so the cartoonesque, programmable truck is not new on the scene for 2018 but has been gaining fresh powers over the years.

This year the company has turned its attention to adults, launching a new but almost identical-looking assistant-style bot, called Vector, thatnot really aimed at kids. That more pricey ($250) robot is slated to be getting access to its code lab in future, so it should have some DIY programming potential too.Age: 8+

Dash Robotics

Product: Kamigami Jurassic World RobotPrice: ~$60Description: Hobbyist robotics startupDash Robotics has been collaborating with toymaker Mattel on the Kamigami line of biologically inspired robots for over a year now. The USB-charged bots arrive at kids& homes in build-it-yourself form before coming to programmable, biomimetic life via the use of a simple, icon-based coding interface in the companion app.

The latest addition to the range is dinosaur bot series Jurassic World, currently comprised of a pair of pretty similar looking raptor dinosaurs, each with light up eyes and appropriate sound effects. Using the app kids can complete challenges to unlock new abilities and sounds. And if you have more than one dinosaur in the same house they can react to each other to make things even more lively.Age: 8+

Kano

Product: Harry Potter Coding KitPrice: $100Description: British learn-to-code startup Kano has expanded its line this year with a co-branded, build-it-yourself wand linked to the fictional Harry Potter wizard series. The motion-sensitive e-product features a gyroscope, accelerometer, magnetometer and Bluetooth wireless so kids can use it to interact with coding content on-screen. The company offers 70-plus challenges for children to play wizard with, using wand gestures to manipulate digital content. Like many STEM toys it requires a tablet or desktop computer to work its digital magic (iOS and Android tablets are supported, as well as desktop PCs including KanoComputer Kit Touch, below)Age: 6+

Product: Computer Kit TouchPrice: $280Description: The latest version of Kanobuild-it-yourself Pi-powered kids& computer. This yearcomputer kit includes the familiar bright orange physical keyboard butnow paired with a touchscreen. Kano reckons touch is a natural aid to the drag-and-drop, block-based learn-to-code systems itputting under kids& fingertips here. Although its KanoOS Pi skin does support text-based coding too, and can run a wide range of other apps and programs — making this STEM device a fully fledged computer in its own rightAge: 6-13

Lego

Product:Boost Creative ToolboxPrice: $160Description: Boost isLegorelatively recent foray into offering a simpler robotics and programming system aimed at younger kids vs its more sophisticated and expensive veteranMindstorms creator platform (for 10+ year olds). The Boost Creative Toolbox is an entry point to Lego + robotics, letting kids build a range of different brick-based bots — all of which can be controlled and programmed via the companion app which offers an icon-based coding system.

Boost components can also be combined with other Lego kits to bring other not-electronic kits to life — such as its Stormbringer Ninjago Dragon kit(sold separately for $40). Ninjago + Boost means = a dragon that can walk and turn its head as if itabout to breathe fireAge: 7-12

littleBits

Product: Avengers Hero Inventor KitPrice: $150Description: This Disney co-branded wearable in kit form from the hardware hackers over at littleBits lets superhero-inspired kids snap together all sorts of electronic and plastic bits to make their own gauntlet from the Avengers movie franchise. The gizmo features an LED matrix panel, based on Tony Starkpalm Repulsor Beam, they can control via companion app. There are 18 in-app activities for them to explore, assuming kids don&t just use amuse themselves acting out their Marvel superhero fantasiesAge: 8+

Itworth noting that littleBits has lots more to offer — so if bringing yet more Disney-branded merch into your home really isn&t your thing, check out its wide range of DIY electronics kits, which cater to various price points, such as this Crawly Creature Kit($40) or anElectronic Music Inventor Kit($100), and much more… No major movie franchises necessary

Makeblock

Product: Codey RockyPrice: $100Description: Shenzhen-based STEM kit maker Makeblock crowdfunded this emotive, programmable bot geared towards younger kids on Kickstarter. Thereno assembly required, though the bot itself can transform into a wearable or handheld device for game playing, as Codey (the head) detaches from Rocky (the wheeled body).

Despite the young target age, the toy is packed with sophisticated tech — making use of deep learning algorithms, for example. While the companyvisual programming system, mBlock, also supports Python text coding, and allows kids to code bot movements and visual effects on the display, tapping into the 10 programmable modules on this sensor-heavy bot. Makeblock says kids can program Codey to create dot matrix animations, design games and even build AI and IoT applications, thanks to baked in support for voice, image and even face recognition… The bot has also been designed to be compatible with Lego bricks so kids can design and build physical add-ons tooAge: 6+

Product: AirblockPrice: $100Description:Another programmable gizmo from Makeblockrange. Airblock is a modular and programmable drone/hovercraft so this is a STEM device that can fly. Magnetic connectors are used for easy assembly of the soft foam pieces. Several different assembly configurations are possible. The companion appblock-based coding interface is used for programming and controlling your Airblock creationsAge: 8+

Ozobot

Product: EvoPrice: $100Description: This programmable robot has a twist as it can be controlled without a child always having to be stuck to a screen. The Evosensing system can detect and respond tomarks made by marker pens and stickers in the accompanying Experience Pack — so this is coding via paper plus visual cues.

There is also a digital, block-based coding interface for controlling Evo, called OzoBlockly (based on GoogleBlockly system). This has a five-level coding system to support a range of ages, from pre-readers (using just icon-based blocks), up to a ‘Master mode& which Ozobot says includes extensive low-level control and advanced programming featuresAge: 9+

Pi-top

Product: Modular LaptopPrice: $320 (with a Raspberry Pi 3 Model B+), $285 withoutDescription: This snazzy 14-inch modular laptop, powered by Raspberry Pi, has a special focus on teaching coding and electronics. Slide the laptopkeyboard forward and it reveals a built in rail for hardware hacking. Guided projects designed for kids include building a music maker and a smart robot. The laptop runspi-toplearn-to-code oriented OS— which supports block-based coding programs like Scratch and kid-friendly wares like Minecraft Pi edition, as well as its homebrew CEEDUniverse: A Civilization style game that bakes in visual programming puzzles to teach basic coding concepts. The pi-top also comes with a full software suite of more standard computing apps (including apps from Google and Microsoft). So this is no simple toy. Not a new model for this year — but still a compelling STEM machineAge:8+

Robo Wunderkind

Product: StarterKitPrice: $200Description:Programmable robotics blocks for even very young inventors. The blocks snap together and are color-coded based on function so as to minimize instruction for the target age group. Kids can program their creations to do stuff like drive, play music, detect obstacles and more via a drag-and-drop coding interface in the companion Robo Code app. Another app — Robo Live — lets them control what they&ve built in real time. The physical blocks can also support Lego-based add-ons for more imaginative designsAge: 5+

Root Robotics

Product: RootPrice: $200Description: A robot that can sense and draw, thanks to a variety of on board sensors, battery-powered kinetic energy and its central feature: A built-in pen holder. Root uses spirographs as the medium for teaching STEM as kids get to code what the bot draws. They can also create musical compositions with a scan and play mode that turns Root into a music maker. The companion app offers three levels of coding interfaces to support different learning abilities and ages. At the top end it supports programming in Swift (with Python and JavaScript slated as coming soon). An optional subscription service offers access to additional learning materials and projects to expand Rooteducational valueAge: 4+

Sphero

Product: BoltPrice: $150Description: The app-enabled robot ball makerlatest STEM gizmo. Itstill a transparent sphere but now has an 8×8 LED matrix lodged inside to expand the programmable elements. This colorful matrix can be programmed to display words, show data in real-time and offer game design opportunities. Bolt also includes an ambient light sensor, and speed and direction sensors, giving it an additional power up over earlier models. The Sphero Edu companion app supports drawing, Scratch-style block-based and JavaScript text programming options to suit different agesAge: 8+

Tech Will Save Us

Product: Range of coding, electronics and craft kitsPrice: From ~$30 up to $150Description: A delightful range of electronic toys and coding kits, hitting various age and price-points, and often making use of traditional craft materials (which of course kids love). Examples include asolar powered moisture sensor kit($40) to alert when a pot plant needs water; electronic dough($35); a micro:bot add-on kit($35) that makes use of the BBC micro:bit device (sold separately); and the creative coder kit($70), which pairs block-based coding with a wearable that lets kids see their code in action (and reacting to their actions)Age:4+, 8+, 11+ depending on kit

UBTech Robotics

Product: JIMU Robot BuilderBots Series: Overdrive KitPrice: $120Description:More snap-together, codable robot trucks that kids get to build and control. These can be programmed either via posing and recording, or using Ubtechdrag-and-drop, block-based Blockly coding program. The Shenzhen-based company, which has been in the STEM game for several years, offers arange of other kitsin the same Jimu kit series — such as this similarly pricedUnicornBotand its classic MeeBot Kit, which can be expanded via the newerAnimal Add-on KitAge: 8+

Wonder Workshop

Product:Dot Creativity KitPrice: $80Description:San Francisco-basedWonder Workshop offers a kid-friendly blend of controllable robotics and DIY craft-style projects in this entry-level Dot Creativity Kit. Younger kids can play around and personalize the talkative connected device. But the startup sells a trio of chatty robots all aimed at encouraging children to get into coding. Next in line thereDash ($150), also for 6+ year olds. ThenCue ($200) for 11+. The startup also has a growing range of accessories to expand the bots& (programmable) functionality — such as thisSketch Kit($40) which adds a few arty smarts to Dash or Cue.

With Dot, younger kids play around using a suite of creative apps to control and customize their robot and tap more deeply into its capabilities, with the apps supporting a range of projects and puzzles designed to both entertain them and introduce basic coding concepts.Age:6+