Happenings
1️⃣ Starbucks is trying out a mobile order-only store; 2️⃣ "Traveler" and "Bon Appétit" magazines join "Billboard" in using the iPhone 7 Plus Portrait Mode to shoot their covers, claiming it's indistinguishable from Pro cameras for the 99.9%; 3️⃣ Apple added some interesting iOS remote management capabilities (incl. remote reboot/shutdown), brought Live Photos to the web (community is important) and expanded App Store ads internationally, 4️⃣ ✨Facebook stole the week✨ by making AR more accessible to creators and a perceived commodity for consumers:
- AR Camera Effects Platform, similar to Snapchat's World Lenses but open to artists/developers, reaching far more users (2B vs 160M) and situations (Facebook/Messenger/Instagram/WhatsApp). Effects range from static stickers to interactive AR (which you can look at from different positions with your phone - basic Inside-out Positional Tracking), associated with faces/places/time/objects (e.g. plant, coffee);
- Mobile Computer Vision: "Caffe2" for recognizing images/videos/sound/text on low-power devices (e.g. on raspberry pi, mobile phones - likely powering AR Camera Effects);
- Facebook's Assistant (dubbed "M", released to all of the US last week and powered by wit.ai natural language processing) understands food ordering intent in group chats and offers its help to order and pay (Food Ordering is also available by talking to directly M);
AI and ML
Research Blog: Federated Learning: Collaborative Machine Learning without Centralized Training Data
Interesting description of how on-device machine learning works. Useful reminder you should question some of your assumptions from time to time (ML doesn’t always need a server).
googleblog.com
Google’s adversarial AIs could lead to less reliance on real-world data | TechCrunch
Even though Machine Learning require far less time to train than a Human, they require the same or more stimuli (e.g. images). That's why even though you can use ML recognition on mobile, it's sometimes faster to train it on the server where you have access to a lot more stimuli. Funnily enough, that is true for general features (e.g. Faces, Couches) but doesn't apply to specifics (e.g. Your Face, Your Couch), which can be learned on the server but why upload when the capability is already on your phone?
techcrunch.com
Fast Drawing for Everyone
For consumers, this is 🤖🎨 Autocorrect for Drawings, for Google, it's free Machine Learning training. Similar symbiotic relationship Duolingo has, where by learning a new language you help translate content.
Fast Drawing short-circuits the creation process, providing quick access to a limited set of outcomes (i.e. the finished drawings in their database), which is a good enough goal for most, and allowing for customization if desired.
🤔 Imagine short-circuiting other limited-outcome processes, you don't even have to end on the same domain where you started (e.g. draw to get a drawing vs write to get a drawing), and you could achieve multiple diverse outcomes (e.g. speak to improve contrast and layout throughout an app – very Star Trek, for now).
blog.google
Why Machine Learning Will Do More For Your Company Than For Your Car
Sam Gerstenzang (A16Z investor, product/team/design manager at Google/Imgur) on how Machine Learning (software) unlocks hardware innovation:
- Design validation through simulations instead of costly prototyping;
- Operational optimization through data insights (e.g. supply chain, quality assurance);
- Personalized consumer experience at scale (e.g. efficient marketing, loyalty).
You'll notice these are applicable not only to cars (hardware) but software products as well. 👷♀️
newcartographer.com
"AI and HCI: two fields divided by a common focus" argues that HCI flourishes when AI declines.
When we can't get Artificial Intelligence to work, we fall back to manual User Interfaces.
twitter.com
Chatbots
Amazon rolls out chatbot tools in race to dominate voice-powered tech | Reuters
Commodity chatbots, also see Wit.ai for a similar service used by Facebook M.
reuters.com
Design
📞🔐 Microsoft replaces the password with a phone-based log-in
Something I've advocated for about 10 years now 🎥 (the login code was sent via Twitter DM and then received via SMS, used a URL instead of an email/username). One of the challenges you have is the cost of sending the SMS (unless you're using an additional code generator, but that’s clunky), and questions such as "What if I loose my phone, how do I log in?". It's good to question these things, and we have to remember: 🚨not all accounts are bank accounts, many simply offer convenience.
engadget.com
Adding Jobs-to-be-Done to a Persona-Based Product Development Process
An interesting conversation with a Senior UX Researcher about Jobs-to-be-Done and Personas
Summary:
- 👩 Personas being used for capturing Attitudes, Jobs-to-be-Done for Situations, and Journey Maps for Experiences.
- 👩 Personas being used to define copy and emotion for different audiences.
- 👨🏫 Jobs-to-be-Done goals used to measure satisfaction (e.g. measure against: I just want to tick the box; I want to actually learn; etc)
- 👨🏫 Jobs-to-be-Done also used to capture paths beyond the “happy path” (this is where the interview became inconsistent)
There's a bigger lesson here though, the answer I got after querying the host on the possibility of capturing attitudes in Jobs-to-be-Done was that this wasn't done due to 🏢 organizational legacy, the difficulty of removing Personas altogether as it represents a big culture change. It's definitely something to have in mind, big changes will very often have to be introduced gradually and with a potential compromise here and there. One thing I couldn't figure out was whether Job Stores were being used as the only term I read was Jobs-to-be-Done.
jobstobedone.org
The Digital Agency That Is Quietly Crushing The Client Game
Interesting to see how frequently sharing an iterative prototype is doing wonders for this agency’s client retention (70% YoY growth 📈, 90% client retention 🎉, 200 employees 👫, NYC/PDX/GIG 🌎). Process seems to downplay user research and validation, instead it starts with multiple ers ideating in isolation and then converging to pick the best ideas. Clients include Apple, Facebook, Google, Nike, Marriott, Virgin America, the method has earned recognition by Forrester Research. It works, and not surprisingly this approach is spearheaded by an ex-Marketeer (i.e. make it flashy ✨).
framer.com
Health
Advances in AI and ML are reshaping healthcare
The cost of personalized health is reducing based on two factors: A) More and richer data (sometimes in our phones); B) Cheaper processing of this data (sometimes in our phones).
techcrunch.com
Reality
Lytro's Latest VR Light-field Camera is Huge, and Hugely Improved – Road to VR
Check out Lytro if you haven't heard of them before, they pretty much quit on consumer products but are developing some rather impressive professional hardware.
roadtovr.com
"Why do we need this @Facebook?"
Perception is the hardest challenge for VR. You don't know how valuable it is (or not) before you try it. The other thing is: Facebook is using consumers for their demos, doing banal things like sharing holiday photos with a lot of ceremony, but the workplace has more limitations in terms of space/time/social — the constraints VR removes and perhaps where it could make a bigger impact.
twitter.com
The end of smartphone innovation — Benedict Evans
Benedict Evans argues smartphones have matured and AR is to become the next big thing.
ben-evans.com
Research
Alan Kay's Research Principles
Alan Kay simply has a different point of view which he says "is worth 80 IQ points". He's shaped some of my understanding around thinking, collaboration, learning, invention.
Read the whole thing, but here's a summary:
- Have a long-term vision;
- Invest in people;
- Take a step back and find the real problem;
- Milestones not deadlines;
- Failure is part of success;
- Cross-domain knowledge;
- Don't let dependencies slow you down;
- "Every invention has to be [documented for and usable by] 100 users".
Security & Privacy
Strategy
Getting Human Resources Right – AVC
HR on a Design newsletter? Well, I if there's one thing a Product Designer needs to be good at is looking beyond a specific domain and around the work desk. Sometimes you come across articles like this which I'll sum up for you (but feel free to read it in its entirety).
- Do HR as early as possible, evolve in tandem with the company;
- Team health is paramount, HR should report to CEO (and CEO to HR!);
- HR isn't hiring and on-boarding/off-boarding;
- HR is culture, leadership, performance, goals for everyone including CEO;
- Clear values and consequences of not following them;
- On-boarding is more than "here's your laptop, desk, and boss", it's "here's our plan, strategy, goals";
- Handbooks should include mechanisms for honest, anonymous feedback and whistleblowing.
Voice UI
Google Home is rolling out support for multiple users
Very interesting, likely using vocal fingerprinting like we have on Android/iOS. Would be interesting if any Home, not just your own, recognized you but I guess our voices aren’t that unique 😉. Rumor has it Apple is exploring facial recognition for the same purpose, hopefully not 2D facial recognition.
thenextweb.com
Ad forces Google Home to advertise a Burger
- Burger Chain ad, just before it ends, asks Google Home to read the Burger's wikipedia page for free advertising;
- People learn about this and start editing the page so it contains odd ingredients and health warnings;
- Google patches Google Home against that specific ad.
Blast From The Past
Bias in Mental Testing, Arthur Jensen, 1980."Before thermometers: philosophers mocked the idea of temperature ever being measurable, with all its nuance, complexity and subjectivity"
Spent hours boiling down 120 links into just 24, each with a summary and personal comment on its importance so you don't have to actually click them (I'm destroying my own click through rate 🖱 along with my free time 🌴).