Espen – James 3(4)-1

James May has bought a new car – a blue Tesla Model S.

Well, I never thought the day would come when I would be ahead of Top Gear, but I am actually on my third blue Tesla (fourth if you count the toy car) as this picture will testify:

99512322-305C-414F-8C5E-62E8E5A21C83

(Incidentally, the Model S is for sale. Interested? You would have the same car as James May…)

Not sure I qualify as a petrolhead, since I have never owned an Alfa Romeo, but at least I can congratulate James May on a car purchase I fully agree with.

Avoid UPS Norway at all costs

UPS – United Parcel Service – is a paragon of operational excellence in the United States. For all I know, they are excellent in other countries as well. In Norway, they are the worst transportation company you can think of – at least when it comes to administration and customer service.

I order quite a few things from abroad, and whenever the company I order from tell me they have sent it with UPS I groan – because I know the package will be late, the paperwork will be cumbersome, and the customer service slow.

Here is the latest example: I order an electric bicycle from Sweden, from Kringla.com. The price is given including Norwegian customs and taxes. I pay the net sum to the bicycle company, and UPS is supposed to send me a link with payment information, since they cannot send the package out before customs has been paid.

Except they did not send it until I called them after 10 days (“because someone forgot to send it” according to customer service.) I paid, the packaged was delivered (the driver was great, incidentally, carried the heavy package behind the house since we were away.)

Then, two weeks later, I get an invoice for the customs and taxes. I call in, and they say (after a long wait on the phone) that it is paid, but they send out an invoice not marked as paid in case I need it for documentation.

I have interviewed top management in UPS in the States, the company is known for excellent delivery services and innovation in logistics. In Norway – check out this page, with 33 reviews, all of them one star…

Can someone in top management in the US please contact the Norwegian organization and tell them to get a grip?

From notepad: The power and limits of deep learning – Yann LeCun

Warning: These are my notes from an ACM webcast. Misunderstandings, skips, jumps and errors (probably) abound. Caveat emptor.

Notes from
The Power and Limits of Deep Learning,” presented on Thursday, July 11 at 1 PM ET/10 AM PT by Yann LeCun, VP & Chief AI Scientist at Facebook, Silver Professor at NYU, and 2018 ACM A.M Turing Award Laureate.

Abstract:
Deep Learning (DL) has enabled significant progress in computer perception, natural language understanding, and control. Almost all these successes rely on supervised learning, where the machine is required to predict human-provided annotations, or model-free reinforcement learning, where the machine learns policies that maximize rewards. Supervised learning paradigms have been extremely successful for an increasingly large number of practical applications such as medical image analysis, autonomous driving, virtual assistants, information filtering, ranking, search and retrieval, language translation, and many more. Today, DL systems are at the core of search engines and social networks. DL is also used increasingly widely in the physical and social sciences to analyze data in astrophysics, particle physics, and biology, or to build phenomenological models of complex systems. An interesting example is the use of convolutional networks as computational models of human and animal perception. But while supervised DL excels at perceptual tasks, there are two major challenges to the next quantum leap in AI: (1) getting DL systems to learn tasks without requiring large amounts of human-labeled data; (2) getting them to learn to reason and to act. These challenges motivate some the most interesting research directions in AI.

Notes:

  • supervised learning works, but requires too many samples
  • convolutional networks: using layers to tease out compositional hierarchy
  • other approaches: reinforcement learning,
    • use convolutional networks and a few other architectural concepts, requires huge number of interactions with clearly defined universe – takes 80 hours to reach performance a human uses 15 minutes to reach. In the end, it does better than the human, but it takes a long time
    • impractical for non-electronic settings (self-driving car would need to crash thousands of times
  • better approach: (deep) multi-layer neural nets
    • alternates linear/non-linear layers
  • supervised machine learning, such as stochastic gradient descent
  • figure out tweaking by computing gradients by back-propagation (automatic differentiation)
  • architecture of neural networks – figure out sparse networks, not using all connections, based on research on visual cortex
    • first using simple cells, then combining them
  • convolutional neural network builds on this idea, but introduces back propagation
    • turn on/off each neuron based on the portion it sees, then combine
  • shows examples through the nineties, such as recognising numbers (for checks)
  • neural networks out of fashion with AI researchers, realized that they could recognize multiple objects
  • research on moving robots, did not need training data
  • moving on to autonomous driving by classifying pixels
  • 2010: Deep learning revolution, driven by speech recognition community
    • largely responsible for lowering of errors in SR
  • 2012: (Alexnet) Krizhevsy et al, NIPS 2012, other nets, large networks
  • better and better performance, dramatic increase in number of layers
    • current record: 84% image recognition
    • trying to find the minimal architecture that gives performance
    • Facebook: billions of pictures, each goes through 6 convnets
  • Mask R-CNN: instance segmentation, two stage detection system, identifies areas of interest and send them to new networks
  • RetinaNet: One-pass object recognition
  • other works, recognizing background,
  • Applications:
    • image recognition, such as finding femurs (for hip ops) by taking in the whole 3D picture rather than using layers
    • autonomous driving
    • everyone uses convnets
  • Limitation:
    • good for perception, not for reasoning
    • for this: introducing working memory (differentiable associative memory), need to maintain a number of facts, “memory network”, a neural net with an attached network for memory, essentially soft RAM
    • transformer networks, every unit is itself a neural network, works with translation (dynamic convolution)
    • Facebook; dynamic neural nets: networks that put out networks
  • Challenge: How can humans and animals learn so quickly?
    • children learn largely by observation
      • learn about gravity between 6 and 9 months, just by observation
    • solution(?) self-supervised networks
      • not task-directed, comprises most of our own learning (cake example)
      • very large networks (see slide on process)
      • works for speech recognition and text, filling in 15-20% of blanks in text
      • does not work for filling in missing parts of images (yet)
      • works partly for speech recognition
      • summary: works with discrete data (text, partly speech), much more difficult with continuous data, because we do not have good ways of parameterization
        • predicts the average of all possible futures, results in blurry images…
    • Adversarial training: prediction under uncertainty:
      • generator that makes prediction, discriminator that determines whether it is good or not
      • works well for generating images of people that don’t exist, clothes that has not been designed yet
      • important with video prediction for self-driving cars, that is where the demand is
    • Self-supervised forward models: training self-driving cars to predict it environment by adding latent variables, randomly sampled
    • Final slide: Theory follows invention, will deep learning result in a theory of intelligence?

(did not take notes during question session, should have don (might add them later), talk available at learning.acm.com)

Teaching case teaching

IMG_5933For the past two days I have run a seminar on case teaching for faculty at the University of Stavanger Business School, by initiative and invitation from Ken Wathne. (The picture above was taken at 11pm, which tells you something about summer in Norway and the gardening at UiS.)

I love case teaching (obviously) and think it is not only a very useful tool to use and skill to have – but also a way for business schools to future-proof their business model. In a time of Youtube and learning-on-demand, case teaching offers a way to analyse and learn real leadership skills in complex decision situation – something that cannot easily be automated, like standard business courses.

Here are some links to things I have used in the seminar:

  • A useful article on what case teaching is (because many people get this wrong). [My comments on this one is in Norwegian, but by all means read the original]: Garvin, David (2003): “Making the case“, Harvard Magazine.
  • A useful article: Christensen, Clayton C. and Michael Raynor (2003) “Why hard-nosed executives should care about management theory”Harvard Business Review. Well-written article on how to create theory – and why it is important to understand what works in some contexts does not work in others. I use this article as a rationale for using case teaching to test theory (simulated) practice – case teaching is simulation of decision situations, and is great for finding out if theories are theories or mere beliefs and buzzwords.
  • Case sources: I primarily use the Harvard Business Press case collection. Anyone teaching at a degree-granting institutions has access (you will have to register with an email address belonging to an institution and be able to point to an affiliation, such as a personal web page at the institution.) There are other case sources too – here is a list.
  • Material for students on how to prepare. I like Bill Ellet’s Case Study Handbook, but it can be a bit much for smaller courses or if you use cases only sparingly. HBSP allows you to buy just some its chapters, so that is an alternative. With my colleague Hanno Roberts, I have made a video series for students, originally for BIs MBA program with Fudan University in Shanghai. The videos are a bit long and rambling, but you can always make your own. For complicated cases I have a note on how to analyze complex case using a structured time line.
  • A blog post on how to do effective student feedback.
  • Course design: You can find many examples on the Interwebs, so I steal with pride. Here are some of my own: GRA6834 Business Development and Innovation Management (M.Sc., one semester, weekly sessions), GRA68175 Technology Management and Disruptive Technology (two day module, EMBA), and my memo to students for my course Strategic Technology and Information Management in Shanghai (four day module, more structured and instrumental than a usual case course.)
  • Writing cases: A blog post on how to write a teaching case. You can, of course, use anything – articles, blog posts, videos – as a basis for discussion. Most important is that there is room for real discussion – that there is enough detail, that the material illustrates what you want to get across in a way that is not obvious and allows for interpretations and complexity.

Practical business development

I have come to learn that there are no boring industries – one always finds something interesting in what at first may looks fairly mundane. And that is something I am trying to teach my students, as well.

Andrew Camarata is a young man who works for himself with excavators, bulldozers, gravel, stone, earthworks and so on. He lives and runs his business in the Hudson Valley just south of Albany, New York and in the winter he does, among a lot of other things, snow plowing.

In this video, he will tell you almost everything there is to know about how to plow snow commercially in rural United States and make money from it.

The interesting point about this video (and a lot of other videos he has made, he has a great following on Youtube) is that he provides a very thorough understanding of business design: In the video, he talks about acquiring and maintaining resources, understanding customers (some are easy, others difficult, you need to deal with both), administration and budgeting, ethics (when to plow, when not to), and risk reduction (add the most complicated jobs with the greatest risk of destroying equipment last in the job queue, to reduce the consequences of breakdowns).

For a business student, this is not a bad introduction to business, and Camarata is certainly a competent businessman. In fact, I see nothing here that is not applicable in any industry.

When it also comes in a pedagogically and visually excellent package, what’s not to like?

Teaching with cases online

teaching_with_case_onlineCase teaching is not just for the classroom – increasingly, you can (and some schools do) offer discussion-based (or, at least, interaction-courses online.) My buddy Bill Schiano and I wrote a long note on how to do this for Harvard Business Publishing back in 2017 – and I then completely forgot about it.

Recently, I rediscovered it online, published by HBP – so, here goes: Schiano, Bill and Espen Andersen (2017): Teaching with Cases Online, Harvard Business Publishing. Enjoy. (More resources available at the HBP Teaching Center page.)

Incidentally – should you (as a school) consider doing courses this way, beware that this is not a cost reduction strategy. You probably will need to pay the online teachers more than those doing regular classrooms, simply because it is more work and quite a bit more design, at least in the beginning. But it may be an excellent way of reaching student groups you otherwise could not reach, for geographical or timing reasons.

Student cases of digitalization and disruption

I teach a M.Sc. class called “business development and innovation management”, and challenge students to write Harvard-style cases about real companies experiencing issues within these areas. The results are always fun and provide learning opportunities for the students: You may not provide the answer for the company, but you get a chance to really learn about one company or one industry and dive into the complexities and intricacies of their situation. That knowledge, in itself is valuable when you are hitting the job market.

Here is a (fairly anonymized) list of this year’s papers:

  • disruption in the analytics industry: One group is studying SAS Institute and how their closed software and architecture model is being challenged by open-source developments
  • disruption in the consulting industry: One group wants to study a small consulting company and how they should market some newly developed software that allows for automated, low-cost analysis
  • establishing a crypto-currency exchange: One group wants to study strategies for establishing a payment and exchange service for crypto-currencies
  • marketing RPA through a law firm: One group wants to study how a large law firm can market their internal capabilities for RPA (robotics process automation) in an external context
  • fast access to emergency services: One group wants to write a case on Smarthelp and how that service can be spread and marketed in a wider context
  • using technology to manage sports club sponsorship: One group wants to study how to develop strategy for a startup company that helps participation sports clubs with gain corporate sponsorships
  • electronic commerce and innovation in the agricultural equipment sector: One group wants to study how a vendor of farm equipment and supplies can extend their market and increase their innovative capability through ecommerce and other digital initiatives
  • machine learning in Indian banking: One group wants to study how machine learning could be used to detect money laundering in a large Indian bank
  • social media analysis in consumer lending: One group wants to study an Indian startup company that uses digital indicators from users’ online behavior to facilitate consumer financing for online purchases

Al in all, a fairly diverse set of papers – I am looking forward to reading them.