for W3c validation
Friday Faves is our weekly blog series highlighting a few select pieces from the REG team’s reading lists. You can catch up on past Friday Faves on the archive.
How to teach Artificial Intelligence some common sense
Anne says: I have to admit… the title of this article drew me in! If we can teach artificial intelligence (AI) common sense, then could we potentially apply that to people as well??!!
Teaching AI is called deep learning – basically it’s about providing data and patterns, then eventually the AI learns the patterns and starts to distinguish for themselves. The article starts off with an example of how DeepMind, a London based AI company, watched as their AI taught itself to play the Atari game, Breakout. Impressive, but it did take 600 or so games before the algorithms could anticipate the bounce of the ball.
However, as the examples in the article illustrate, it’s also rather easy to confuse or manipulate the AI outcomes. Another company replicated DeepMind’s Breakout AI but tweaked a few items including moving the paddle a little higher and an unbreakable block. It failed. As humans, we can use generalisation to adapt quickly to these sorts of anomalies – we understand cause and effect, we use inference and past knowledge to apply context and adapt to new situations. AI didn’t understand the game of Breakout, it was just following a pattern. When the pattern changed, too bad, so sad!
Although AI is now used in many mainstream devices and systems – self-driving cars, speech recognition like Alexa and Siri – it seems to have hit some challenges. Replicating human intelligence – our ability to generalise – goes beyond pattern recognition. Again, the examples where images are slightly altered create flawed results in the AI. Something that as humans we can do quickly, can take AI hundreds of examples to adapt to. The question is, can we ever develop AI with common sense or the ability to generalise? And what would that involve?
How do we provide the data that we have stored since childhood? Fire burns, balls bounce, relative sizes – eg. elephants don’t fit in bathtubs!
And now, back to the big question – what is common sense? The definition in the article is wonderful:
“… all the knowledge about the world that we take for granted but rarely state out loud.”
The scientists are using questions such as: “If I put my socks in a drawer, will they be there tomorrow?” (This is starting to sound like my first year university philosophy class!)
The process to achieve this in AI is rather amazing, but also logical: Ask people, lots of them, their response to a set of everyday situation questions. Even then, the AI was only able to achieve a 50% correct response.
It sounds overwhelming, and the number of projects described attempting to crack this code reinforces how difficult this challenge is. Navigating through the concepts of rogue AI machines that try to kill us, the article concludes with a thought-provoking question:
“… maybe it’s worth keeping artificial intelligence a little bit artificial.”
Warning: this is a long read but is really valuable for grasping the tasks that face us with AI – it’s just not common sense!
Virtual Restaurants – On the Rise
Helen says: Uber Eats has been around since 2015 and is enjoying strong growth in the rapidly expanding online food delivery business. Some of this growth is attributed to the emergences of “ghost”, or virtual, restaurants. This digital disruptor has used its data to identify demands for particular foods in specific areas then approached restaurants to take on the role of the supplier, and the model is meeting with success. An Italian restaurant, for example, might take on the production of foods far removed from their standard Italian cuisine, market them under a completely different brand and make these items available only to online delivery food delivery businesses.
Some businesses have even built ‘restaurants’ specifically to service this market and you won’t see a table or chair in sight. You may not even be able to find the restaurant – it doesn’t have to be visible to be successful. The low overheads in setting up these product lines not only contributes to improved profit margins but also allows the business to be more flexible and responsive to market demands. The exciting part of this is their ability to experiment with new and creative ideas without risking their entire business.
I enjoy following the changes, challenges and new products and services that come with digital disruption and look forward to seeing what the VP of UberEverything comes up with next! What about in your industry – could your organisation be disrupted by “ghost” versions of your business?
Harry Potter coding kit brings its magic to Apple stores
Joel says: I’m a big advocate for teaching kids to code at a young age. Even if they don’t want to be computer programmers one day it’s still a fantastic way to teach problem-solving skills that can be used for the rest of their lives. That’s why I was excited to see that kids can now learn to code and express their fondness for Harry Potter all at the same time now.
Kano, the creator of the interactive kit that lets you code with a wand will be bringing their product to some Apple stores in the US, UK, Australia and New Zealand. And in select stores, there will be staff performing demonstrations with the kit in December and January.
“Our goal is to demystify technology,” said Alex Klein, co-founder and CEO of Kano, in a release. “The rich gestural technology of our coding wand and iconic Wizarding World moments combine in the Harry Potter Kano Coding Kit.”
The $100 kit is available on Kano’s website, Apple.com and some Apple stores. You can build the wand yourself, then cast spells with it on iOS and Android tablets, and on Windows and Mac computers. It can detect more than 30 spell gestures, including making a feather float (it’s levy-OH-sa, not levy-oh-SAH!).
Kano have also partnered with Weasley twin actors James and Oliver Phelps from the Harry Potter films to help advertise the coding wand, and you can find out more about it on the official Kano website. It’s an interesting approach to making learning engaging – maybe the kids who learn this way will come to work expecting a learning wand to do their compliance training.
Voice tech like Alexa and Siri hasn’t found its true calling yet: Inside the voice assistant ‘revolution’
Jakkii says: This is a very interesting long read on where we really are with voice technology, amongst the hype of Alexa and Google and Cortana and Siri and, of course, Facebook’s new Portal.
Spoiler alert: we’re much earlier in the stages of voice technology development – and use – than the hype might suggest.
There’s some interesting data provided in the article about a range of different aspects of voice assistance, such as how frequently a smart speaker is used after purchase. There’s also a graph indicating the frequency of use of smart assistance for a variety of activities, the top five of which were:
- Music – 70%
- Weather forecast – 64%
- Fun questions – 53%
- Online search – 47%
- Alarms/reminders – 46%
One of the particularly interesting insights was regarding advertising, particularly in relation to listening to podcasts. It seems we’re a lazy lot – it’s more of a pain to get the voice assistant to skip an ad than listen to it, so users are much less likely to skip ads when using a smart speaker than listening through more traditional methods.
“Smart speaker listeners are much more passive,” she added. “People with voice interfaces tend to accept what’s given to them.”
Ultimately, the article contents, as we’re only in the early days of voice technology there’s a bright future ahead – but it probably won’t be in smart speakers. Instead, the important part of voice technology is the voice assistant, and that’s where the future will be. What exactly that might look like, however, remains to be seen.
On a completely different note, with algorithms and AI also high on hype I wanted to share this piece about Amazon’s sexist hiring algorithm with you as well this week. Between the team we’ve talked a lot about bias over the course of our Friday Faves – most recently when Anne discussed deliberate bias in AI – so it came as little surprise to learn Amazon recently shut down an experimental AI after it came to light that it was discriminating against women.
The company created the tool to trawl the web and spot potential candidates, rating them from one to five stars. But the algorithm learned to systematically downgrade women’s CV’s for technical jobs such as software developer.
In exploring this issue of bias, AI and human judgement further, the article gives a nice bit of background on bias in algorithms, before going on to discuss a 2018 study which looked at what happens when we allow AI to make decisions without human intervention.
No spoiler alerts for this one though, as it’s worth having a read and a good think about bias more broadly, about how and why we need to manage it in algorithms and their resultant AI, and maybe even ponder a bigger question: how much does bias itself contribute to what it means to be human?
This week in social media
- French regulators will ’embed’ with Facebook to monitor how it combats hate speech
- 6 ways to make your social media posts accessible for people with a visual impairment
- Facebook Portal isn’t listening to your calls, but may track data
- Design for life: Pinterest chief out to inspire as it gets personal with users
- Facebook to end forced arbitration for sexual-harassment claims
- Twitter is thinking about an edit button to fix typos in tweets
- Facebook launches Lasso, its music and video TikTok clone
- #StartWithThem: What Makes Twitter’s Aussie Audience Unique
- LinkedIn continues to re-focus on Groups, outlines new Groups infrastructure
- Snapchat adds Friendship Profiles to highlight your relationships
Sydney Business Insights – The Future This Week Podcast
This week: satire sells science, rats are not human, and bacterial bio bricks. Sandra Peter (Sydney Business Insights) and Kai Riemer (Digital Disruption Research Group) meet once a week to put their own spin on news that is impacting the future of business in The Future, This Week.
The stories this week:
Satire can be an effective tool for communication
Massive study into link between cell phones and cancer is almost immediately irrelevant
Other stories we bring up:
Ars technica on John Oliver’s science rant
Using creative mediums to showcase research
Satire is more informative than actual news
Hans Rosling’s TED Talk, The best stat’s you’ve ever seen
Metro reports ‘clear link’ between cellphone use and cancer (not)
‘Clear evidence’ mobile phones ARE linked to cancer, landmark study finds (not)
Landmark study finds cellphones are linked to cancer (not)