Reflections on “Inside Amazon’s artificial intelligence flywheel” [Levy]

Inside Amazon’s artificial intelligence flywheel — from wired.com by Steven Levy
How deep learning came to power Alexa, Amazon Web Services, and nearly every other division of the company.

Excerpt (emphasis DSC):

Amazon loves to use the word flywheel to describe how various parts of its massive business work as a single perpetual motion machine. It now has a powerful AI flywheel, where machine-learning innovations in one part of the company fuel the efforts of other teams, who in turn can build products or offer services to affect other groups, or even the company at large. Offering its machine-learning platforms to outsiders as a paid service makes the effort itself profitable—and in certain cases scoops up yet more data to level up the technology even more.

It took a lot of six-pagers to transform Amazon from a deep-learning wannabe into a formidable power. The results of this transformation can be seen throughout the company—including in a recommendations system that now runs on a totally new machine-learning infrastructure. Amazon is smarter in suggesting what you should read next, what items you should add to your shopping list, and what movie you might want to watch tonight. And this year Thirumalai started a new job, heading Amazon search, where he intends to use deep learning in every aspect of the service.

“If you asked me seven or eight years ago how big a force Amazon was in AI, I would have said, ‘They aren’t,’” says Pedro Domingos, a top computer science professor at the University of Washington. “But they have really come on aggressively. Now they are becoming a force.”

Maybe the force.

 

 

From DSC:
When will we begin to see more mainstream recommendation engines for learning-based materials? With the demand for people to reinvent themselves, such a next generation learning platform can’t come soon enough!

  • Turning over control to learners to create/enhance their own web-based learner profiles; and allowing people to say who can access their learning profiles.
  • AI-based recommendation engines to help people identify curated, effective digital playlists for what they want to learn about.
  • Voice-driven interfaces.
  • Matching employees to employers.
  • Matching one’s learning preferences (not styles) with the content being presented as one piece of a personalized learning experience.
  • From cradle to grave. Lifelong learning.
  • Multimedia-based, interactive content.
  • Asynchronously and synchronously connecting with others learning about the same content.
  • Online-based tutoring/assistance; remote assistance.
  • Reinvent. Staying relevant. Surviving.
  • Competency-based learning.

 

The Living [Class] Room -- by Daniel Christian -- July 2012 -- a second device used in conjunction with a Smart/Connected TV

 

 

 

 

 

 

 

We’re about to embark on a period in American history where career reinvention will be critical, perhaps more so than it’s ever been before. In the next decade, as many as 50 million American workers—a third of the total—will need to change careers, according to McKinsey Global Institute. Automation, in the form of AI (artificial intelligence) and RPA (robotic process automation), is the primary driver. McKinsey observes: “There are few precedents in which societies have successfully retrained such large numbers of people.”

Bill Triant and Ryan Craig

 

 

 

Also relevant/see:

Online education’s expansion continues in higher ed with a focus on tech skills — from educationdive.com by James Paterson

Dive Brief:

  • Online learning continues to expand in higher ed with the addition of several online master’s degrees and a new for-profit college that offers a hybrid of vocational training and liberal arts curriculum online.
  • Inside Higher Ed reported the nonprofit learning provider edX is offering nine master’s degrees through five U.S. universities — the Georgia Institute of Technology, the University of Texas at Austin, Indiana University, Arizona State University and the University of California, San Diego. The programs include cybersecurity, data science, analytics, computer science and marketing, and they cost from around $10,000 to $22,000. Most offer stackable certificates, helping students who change their educational trajectory.
  • Former Harvard University Dean of Social Science Stephen Kosslyn, meanwhile, will open Foundry College in January. The for-profit, two-year program targets adult learners who want to upskill, and it includes training in soft skills such as critical thinking and problem solving. Students will pay about $1,000 per course, though the college is waiving tuition for its first cohort.

 

 

 

A cyber-skills shortage means students are being recruited to fight off hackers — from technologyreview.com by Erin Winick
Students with little or no cybersecurity knowledge are being paired with easy-to-use AI software that lets them protect their campus from attack.

Excerpt:

There aren’t enough cybersecurity workers out there—and things are getting worse. According to one estimate, by 2021 an estimated 3.5 million cybersecurity jobs will be unfilled. And of the candidates who apply, fewer than one in four are even qualified.

That’s why many large corporations are investing in longer-term solutions like mobile training trucks and apprenticeship programs. But Texas A&M University has found a way to solve its labor shortage in the short term. It’s pairing student security beginners with AI software.

The college’s Security Operations Center deals with about a million attempts to hack the university system each month. While the center does have some full-time employees, the majority of its security force is made up of students. Ten students currently work alongside AI softwareto detect, monitor, and remediate the threats.

 

 

 

 

In the 2030 and beyond world, employers will no longer be a separate entity from the education establishment. Pressures from both the supply and demand side are so large that employers and learners will end up, by default, co-designing new learning experiences, where all learning counts.

 

OBJECTIVES FOR CONVENINGS

  • Identify the skills everyone will need to navigate the changing relationship between machine intelligence and people over the next 10-12 years.
  • Develop implications for work, workers, students, working learners, employers, and policymakers.
  • Identify a preliminary set of actions that need to be taken now to best prepare for the changing work + learn ecosystem.

Three key questions guided the discussions:

  1. What are the LEAST and MOST essential skills needed for the future?
  2. Where and how will tomorrow’s workers and learners acquire the skills they really need?
  3. Who is accountable for making sure individuals can thrive in this new economy?

This report summarizes the experts’ views on what skills will likely be needed to navigate the work + learn ecosystem over the next 10–15 years—and their suggested steps for better serving the nation’s future needs.

 

In a new world of work, driven especially by AI, institutionally-sanctioned curricula could give way to AI-personalized learning. This would drastically change the nature of existing social contracts between employers and employees, teachers and students, and governments and citizens. Traditional social contracts would need to be renegotiated or revamped entirely. In the process, institutional assessment and evaluation could well shift from top-down to new bottom-up tools and processes for developing capacities, valuing skills, and managing performance through new kinds of reputation or accomplishment scores.

 

In October 2017, Chris Wanstrath, CEO of Github, the foremost code-sharing and social networking resource for programmers today, made a bold statement: “The future of coding is no coding at all.” He believes that the writing of code will be automated in the near future, leaving humans to focus on “higher-level strategy and design of software.” Many of the experts at the convenings agreed. Even creating the AI systems of tomorrow, they asserted, will likely require less human coding than is needed today, with graphic interfaces turning AI programming into a drag-and-drop operation.

Digital fluency does not mean knowing coding languages. Experts at both convenings contended that effectively “befriending the machine” will be less about teaching people to code and more about being able to empathize with AIs and machines, understanding how they “see the world” and “think” and “make decisions.” Machines will create languages to talk to one another.

Here’s a list of many skills the experts do not expect to see much of—if at all—in the future:

  • Coding. Systems will be self-programming.
  • Building AI systems. Graphic interfaces will turn AI programming into drag-and-drop operations.
  • Calendaring, scheduling, and organizing. There won’t be need for email triage.
  • Planning and even decision-making. AI assistants will pick this up.
  • Creating more personalized curricula. Learners may design more of their own personalized learning adventure.
  • Writing and reviewing resumes. Digital portfolios, personal branding, and performance reputation will replace resumes.
  • Language translation and localization. This will happen in real time using translator apps.
  • Legal research and writing. Many of our legal systems will be automated.
  • Validation skills. Machines will check people’s work to validate their skills.
  • Driving. Driverless vehicles will replace the need to learn how to drive.

Here’s a list of the most essential skills needed for the future:

  • Quantitative and algorithmic thinking.  
  • Managing reputation.  
  • Storytelling and interpretive skills.  
  • First principles thinking.  
  • Communicating with machines as machines.  
  • Augmenting high-skilled physical tasks with AI.
  • Optimization and debugging frame of mind.
  • Creativity and growth mindset.
  • Adaptability.
  • Emotional intelligence.
  • Truth seeking.
  • Cybersecurity.

 

The rise of machine intelligence is just one of the many powerful social, technological, economic, environmental, and political forces that are rapidly and disruptively changing the way everyone will work and learn in the future. Because this largely tech-driven force is so interconnected with other drivers of change, it is nearly impossible to understand the impact of intelligent agents on how we will work and learn without also imagining the ways in which these new tools will reshape how we live.

 

 

 

What will be important in the learn and work ecosystem in 2030? How do we prepare? — from evolllution.com by Holly Zanville | Senior Advisor for Credentialing and Workforce Development, Lumina Foundation

Excerpt:

These seven suggested actions—common to all scenarios—especially resonated with Lumina:

  1. Focus on learning: All learners will need a range of competencies and skills, most critically: learning how to learn; having a foundation in math, science, IT and cross-disciplines; and developing the behaviors of grit, empathy and effective communication.
  2. Prepare all “systems”: Schools will continue to be important places to teach competencies and skills. Parents will be important teachers for children. Workplaces will also be important places for learning, and many learners will need instruction on how to work effectively as part of human/machine teams.
  3. Integrate education and work: Education systems will need to be integrated with work in an education/work ecosystem. To enable movement within the ecosystem, credentials will be useful, but only if they are transparent and portable. The competencies and skills that stand behind credentials will need to be identifiable, using a common language to enable (a) credential providers to educate/train for an integrated education/work system; (b) employers to hire people and upgrade their skills; and (c) governments (federal/state/local) to incentivize and regulate programs and policies that support the education/work system.
  4. Assess learning: Assessing competencies and skills acquired in multiple settings and modes (including artificial reality and virtual reality tools), will be essential. AI will enable powerful new assessment tools to collect and analyze data about what humans know and can do.
  5. Build fair, moral AI: There will be a high priority on ensuring that AI has built-in checks and balances that reflect moral values and honor different cultural perspectives.
  6. Prepare for human/machine futures: Machines will join humans in homes, schools and workplaces. Machines will likely be viewed as citizens with rights. Humans must prepare for side-by-side “relationships” with machines, especially in situations in which machines will be managing aspects of education, work and life formerly managed by humans. Major questions will also arise about the ownership of AI structures—what ownership looks like, and who profits from ubiquitous AI structures.
  7. Build networks for readiness/innovation: Open and innovative partnerships will be needed for whatever future scenarios emerge. In a data-rich world, we won’t solve problems alone; networks, partnerships and communities will be key.

 

 

Also see:

 

 

Three shifts as big as print to digital — from gettingsmart.com by Tom Vander Ark

Excerpts (emphasis DSC):

We just lived through the biggest shift in learning since the printing press—a 25-year shift from print to digital. While it extended access and options to billions, it didn’t prove as transformational as many of us expected. It did, however, set the stage for three shifts that will change what and how people learn.

  1. Basic to broader aims.
  2. Passive to active learning.
  3. Time to demonstrated learning.

 

 

 

U.S. students spend more time working paid jobs than going to class — from bloomberg.com by Riley Griffin
Facing mounting debt, U.S. college students spend double the time working paid jobs than in the library.

Excerpts:

Haunted by costly degrees and insurmountable student debt, American college students now spend more time working paid jobs than in lectures, the library or studying at home.

The vast majority of current students—85 percent—work while enrolled, according to an HSBC survey published Thursday. Students spend an average of 4.2 hours a day working paid jobs, which is more than double the time they spend in the library, nearly two hours more than they spend in class and 1.4 hours more time than they spend studying at home.

Haunted by costly degrees and insurmountable student debt, American college students now spend more time working paid jobs than in lectures, the library or studying at home.

The vast majority of current students—85 percent—work while enrolled, according to an HSBC survey published Thursday. Students spend an average of 4.2 hours a day working paid jobs, which is more than double the time they spend in the library, nearly two hours more than they spend in class and 1.4 hours more time than they spend studying at home.

 

“The economics of the debt crisis have become a major distraction to students’ education,” said John Hupalo, founder and chief executive officer of Invite Education, an education financial planner. “Students’ first priority should be to get value out of their education, not squeezing out hours at a job in order to make money to sustain that education.”

 

 


From DSC:
Obviously, this could be a major problem for many students — depending upon whether their work experiences are paying off in terms of other kinds of learning/experiences/skills development/obtaining jobs later on. But this need to work to get through school is also why I think online education needs to be more prevalent in higher ed. If students need to work to obtain a degree, then they need the flexibility to make their class schedules jibe with their work schedules. As with healthcare, I’d also like to see us find ways to bring the costs down.

 

Also see:

One HBCU Hopes Its ‘$10,000 Degree Pathway’ Will Win Over Students Considering For-Profit Alternatives — from edsurge.com by Jeff Young

Excerpt:

A public university in North Carolina has teamed up with six community colleges to offer a program that promises students they will pay no more than $10,000 out of pocket for their four-year degree.

Participating students will attend a two-year college in the state to get their Associate’s degree, then transfer to an online program at Fayetteville State University to finish their bachelor’s. The students will continue to have access to mentors and resources at the local community college to help them stay on track.

 

Making College Affordable Remains a High Priority in Washington — from campustechnology.com by Sara Friedman
More states are providing free college tuition, but equity concerns remain when it comes to the costs of textbooks, transportation and housing.

 

 

 

You spend weeks studying for an important test. On the big day, you wait nervously as your teacher hands it out. You’re working your way through, when you’re asked to define “ataraxia.” You know you’ve seen the word before, but your mind goes blank. What just happened? Elizabeth Cox details the complex relationship between stress and memory.

 

 

Some of the ways to reduce stress that was mentioned include:

  • Getting regular exercise
  • Getting enough sleep
  • Doing practice tests — especially under similar conditions; under time pressure for example
  • On the day of the test, take deep breaths

 

 

 

From DSC:
Can we please see a Saturday Night Live skit on this? It would be really interesting to see what happens to the AI based on certain facial expressions!  🙂


 

 

Excerpt (emphasis DSC):

Graduates are spending thousands of pounds on training to beat tough emotion-scanning robot interviewers for top City jobs.

Firms such as Goldman Sachs and Unilever are using artificial intelligence (AI) software to weed out candidates, as single advertised positions attract thousands of graduates.

Via a webcam, the software remotely asks preliminary-round candidates 20 minutes of questions and brain-teasers, and records eye movements, breathing patterns and any nervous tics.

 


From DSC:
But on a more serious note, getting by the Applicant Tracking Systems and AI’s of the world — in order to actually talk to a human being — is getting harder and harder to do.  


 

 

 

Why giving kids a roadmap to their brain can make learning easier — from edsurge.com by Megan Nellis

Excerpts:

Learning, Down to a Science
Metacognition. Neuroplasticity. Retrieval Practice. Amygdala. These aren’t the normal words you’d expect to hear in a 15-year-old rural South African’s vocabulary. Here, though, it’s common talk. And why shouldn’t it be? Over the years, we’ve found youth are innately hungry to learn about the inner workings of their mind—where, why and how learning, thinking and decision-making happens. So, we teach them cognitive science.

Over the next three years, we teach students about the software and hardware of the brain. From Carol Dweck’s online Brainology curriculum, they learn about growth mindset, memory and mnemonics, the neural infrastructure of the brain. They learn how stress impacts learning and about neuroplasticity—or how the brain learns. From David Eagleman and Dan Siegel, they learn about the changing landscape of the adolescent brain and how novelty, emotionality and peer relationships aid in learning.

Pulling from books such as Make It Stick and How We Learn, we pointedly teach students about the science behind retrieval practice, metacognition and other strategies. We expressly use them in our classes so students see and experience the direct impact, and we also dedicate a whole class in our program for students to practice applying these strategies toward their own academic learning from school.

 

 

 

Choice: The key to reaching every student — from flr.flglobal.org by Terra Graves

Excerpt:

Who doesn’t like to have a choice?  This seems like a no-brainer to me. Whenever teachers can give their students choices in their learning process, everyone wins. When we have options, we tend to have more ownership of that experience. It also provides us with a sense of control, which most students do not experience in school. In her article on facultyfocus.com, Elizabeth Betsy Lasley EdD writes, “When students are asked to interpret, construct, and demonstrate their concepts or ideas regarding specific course concepts from a selection of product or performance options, content retention, commitment, motivation, and creativity increase.” Flipped Learning environments are ripe for offering choices to students in how they consume content and how they express their learning outcomes. Giving students choice allows us to reach every student, every day because it honors their individuality. Cassie Shoemaker explains it simply in her article Let it go: Giving students choices, “When teachers give students choices as to how they will show what they have learned, students become better problems solvers, more creative, and more engaged.” Problem-solving: It’s not just for math! Students NEED to have opportunities to make decisions in school to learn to make decisions in life. If we continue to spoonfeed and micromanage our students, they won’t learn to figure things out on their own.  Teachers by nature tend to be control freaks (including me). However, when we allow our students to try/fail/try again, we support their growth and confidence.

 

 

 

Looking for something?

Use the form below to search the site:

Still not finding what you're looking for? Drop a comment on a post or contact us so we can take care of it!

© 2018 | Daniel Christian