• Student


The Dark Side of Machine Learning: An Amazon Case Study
Posted on November 13, 2018 at 4:52 pm
From startups to Big Tech, everyone loves to tout their strategy for leveraging machine learning. These companies promise that machine learning is our 21st century savior; it will not only liberate us from the tedious drudgery of administrative tasks but [...]

Activity Feed

On November 14, 2018, nicivey commented on Made In Space. Literally. :

Thank you for this insightful piece. Like many commenters, I question the quality standards and market potential of this technology.

Common sense and this blog from NASA strongly suggest that high quality and reliability are paramount for rockets ( This is only increasingly important where the technology is used in manned rockets, especially commercial ones of the future where one tragedy could derail the entire industry. That said, I think there is opportunity to continue their current strategy of using the tech for non-critical things until the technology develops further.

Additionally, the market potential for this company seems to be very small, so small that it may be hard for them to have enough volume to be profitable. There were 89 space flights in the past year, many of which were unmanned ( This is not a broad market. It is also unlikely that commercial space flight will widen the market in the near term. Despite the hype, most companies are mainly focused on deploying supplies to the international space station and launching astronauts ( Accessible space travel for consumers seems far out which hurts the potential demand/profitability for this company. All in all, this is intriguing company but it has many hurdles on its path for success.

On November 14, 2018, nicivey commented on The Benefits of Additive Manufacturing at BIA :

Thank you for this insightful essay. I think issue of whether to outsource or vertically integrate is an important one, especially since it will likely shape the company’s suite of competitive advantages that are key to marketshare. This will be especially critical as experts project 3D printing in the automotive space to be a $8B business by 2024 so rapid adoption and change area likely ( Before BIA jumps blindly on the bandwagon, I think there are two key challenges they need to consider.

1. BIAs Ability To Invest: What is BIAs ability to tolerate PP&E costs associated with buying rapidly outdated machinery? Will the depreciation present tax breaks that offset the costs? Is there any regulations that need to be considered (tax credits?) Does the company have enough EBIT to support these sorts of continuous investments? How long do experts forecast until the industry reaches a relative steady state?

2. The Costs of Outsourcing: What are the costs and advantages to BIA of having another supplier? Will this lead to further coordination problems? Will the mark-up BIA now has to pay be worth the upside?

I think exploring and answering these overall questions will be key. While it is difficult to speak to the company’s specific financials (a google search could not produce them), market research suggests that many companies are developing their in-house 3D printing capabilities ( This may make it prudent for BIA to keep these tech in house, but the final decision will need to take into consideration how they answer the questions posed above.

On November 14, 2018, nicivey commented on Crowdsourcing legacy: How Rio is changing policymaking :

I love this piece and found the idea of applying OI in the public sector extremely thought provoking. I think your question of how sustainable engagement with OI is across variable political conditions is an important one. I think one of the most detrimental ways this could manifest is citizens engage with the process but are disgruntled when they don’t see their idea come to fruition and do not know why. I think your solution of having a more robust feedback loop and bipartisan buy-in is very helpful to mitigating those issues. I also think it would be helpful to push a stronger relationship aspect to these platforms. A recent HBR article entitled Open Innovations Next Challenge: Itself ( posits that a strong way to improve the efficacy of OI is to ensure it helps create more relationships and these relationships and subsequent brainstorms could lead to stronger innovation, community connection, and more as people connect and build on each others ideas.

I think there are two key ways that engagement and reputation can be preserved during these scenarios. The first s

Thank you for this thought-provoking piece. There are two things I’d like to add to the discussion on what makes OI successful.

As @Charlotte Chang noted, one condition to make OI successful is that the company should likely have a legitimate blind spot where consumers can help. I agree and think that OI can help companies rethink strategies, but I also think a successful condition for OI is that companies be able to discern which suggestions have mass appeal. I fear that sometimes OI solicits recommendations from a niche of power users or people who may have their own blindspots when making recommendations. You cite an example of OI as Modelez’s Cherry Cola Oreo, but Amazon and Target reviews suggest that the product was neither like nor broadly bought (

The other idea I’d like to posit is that company could likely do more to foster relationships within it’s OI community. We know from cases like IDEO and IBM Watson, that work culture and being able to brainstorm and bounce ideas off other people can lead to better innovations. This is also a central thesis of a recent HBR article title Open Innovations Next Challenge:Itself (

This is such an interesting piece about an interesting product. I agree with the other commenters concerns about machine learning applications in healthcare. The issues raised there can help address your second question of the best go-to-market strategy for Mindstrong. Namely, the go-to-market strategy will need to address the concerns of data security and the efficacy of the analysis of that data. I think one way to address those challenges is by partnering with a mental healthcare provider. There are over 550K+ providers in the US alone ( so there is significant opportunity for partnership. Furthermore, partnering with providers could add a sense of legitimiacy in the crowded field of mental health apps (there are over 10K according to this site: I also think it helps protect the company as the providers can help mediate the results and treatment protocol. All in all, I think this could be a helpful tool if deployed correctly.

On November 13, 2018, nicivey commented on Predictive Policing: Promoting Peace or Perpetuating Prejudice? :

Thank you for this very important essay. I think you are absolutely correct in highlighting the many potential pitfalls with machine learning. Unfortunately, these issues are not getting enough air time and could lead to catastrophic results. I fundamentally challenge the notion that machine learning can be unbiased in its prediction of crime. As you point out, it relies on too much biased data and has an unchecked feedback loop. Because of these issues, I wholeheartedly agree that it would be more powerful to repurpose ML to help prevent the use of excessive force or to connect people with needed resources. There is very little downside in these cases, whereas the downside in the current applications is enormous and has a generational impact. I also wonder if there needs to be government intervention. Perhaps laws need to be created to prevent certain machine learning/artificial intelligence applications because of the implications and severity of biased data.

On November 9, 2018, nicivey commented on The North Face & IBM Watson: A Winning E-Commerce Combination? :

This is a very intriguing post that points out many of the potential pitfalls with machine learning and how humans may have higher standards for machine vs. man. I think that humans will continue to have a higher standard for machines, but I am not sure that is a bad thing. If we are going to lose the je ne sais quoi of human interactions (and the potential jobs that come with a move from humans to machine), I think it is a reasonable tradeoff that machines perform better than humans in this context. That said, this line of thinking becomes tricky when error is much more catastrophic. We see this play out with self-driving cars. They are essentially held to an error rate of zero. Any accident results in a ton of bad PR and usually a scaling back of the pilot program. This may be unfair as the car may have a significantly lower error rate than its human counterpart, but we are less willing to tolerate those minimal errors because they come from a machine. Overall, I think this shift to automation will require a better understanding of the psychology behind how humans view machines and AI/machine learning likely demands a shift in that perspective if it is to be adopted widely.

This is a thought-provoking piece. I particularly like the well-thought out action plan anchored in rapid prototyping and celebrity sponsorship. You mention “The reduction of time to produce this shoe comes at a limited cost to Nike as the process improvements, such as using specific lines of material, do not deter from global construction and processes currently in place.” I agree that there is likely a net-positive impact from Nike’s utilization of 3D printing, but I disagree with the assertion that it will be minimally disruptive to other industries. 3D printing threatens to disrupt vulnerable populations that currently work in Nike’s shoe factories. Over 1M+ contractors currently help make Nike shoes and I wonder what happens to those jobs as 3D printing becomes the preferred method of production ( On the one hand, this could be net-positive given the horrific workplace conditions like sub-standard pay that have been reported over the years ( On the other hand, this could be net negative given many of these contractors are located in the developing world and there are limited alternative opportunities. This does not change many of the conclusions you asserted in your article, but does beg the question of how disruptive this innovation will be to the world economy and who will bear the brunt of the negative externalities.