Sam's Profile

This seems like a great space for open source innovation. People who are using these libraries are typically highly educated and attuned to the issues facing them when using products for research. Academics, heavy library resource users, are increasingly required to learn how to code in traditionally non-technical areas (for example, accountants and political scientists needing to learn natural language processing to do research), and allowing these people to contribute to the systems that allow them to succeed as academics seems like a natural step, and one that many users would aid in developing. We have seen user tools develop over the past few years that help save time (e.g. automated citations), but I am sure that opening up coding to many users would produce innovative solutions to finding information that are less obvious, and perhaps more helpful/impactful.

Thanks for the post! It seems that compared to some other shoe companies (Nike, Adidas) that others have posted about, Brooks is somewhat behind the 3D printing game. It will be interesting to see how a lagging mover in the custom shoe market will be able to maintain its place in the running shoe market. Obviously, Brooks has found that niche as being a great shoe to go for custom soles, but could Nike or Adidas displace them if they develop marketable 3D technology faster than Brooks does? I would see this as a clear competitive threat, and Brooks should focus on developing these technologies, most likely in conjunction with an experienced third party, so that they can protect their brand going into the future. The question is immediacy: does Brooks have to go towards AM now, or is waiting five or ten years for the technology to become more efficient possible?

Great article–I echo Gavriel’s comment on clarity for a complex topic.

The idea of applying machine learning to locate greenfields is fascinating, and the problems associated with it appear to be very complex. To start, the data issue that GoldSpot faces should be their number one priority if they believe that greenfield identification is possible and achievable in the near term. This decision has interesting potential knock-on effects on their current revenue driver, brownfield identification. In the interest of driving toward greenfield identification, I would assume that GoldSpot would be willing to offer their services to almost any large company for almost nothing if they believed that greenfield identification were possible with that additional data; the value of their company would be exponentially higher if greenfield identification were possible. Because GoldSpot has not yet made a “mad dash” for data, I believe that they currently see greenfields as out of reach, an area does not yet deserve the full strategic direction of the company.

There are a lot of intellectual property issues involved, and AM will certainly need to address them going forward. Manufacturers are very reluctant to hand over information on their parts due to IP concerns, let alone hand over CAD drawings for new parts. There will likely have to be a strong escrow process for CAD drawings with clear rules around when and how these files can be accessed and used. This is complicated by the fact that trains use components from many different suppliers, all of which are trying to protect their IP.

Currently, DB is largely producing (through its third party partners) CAD files for parts which do not hit on these IP issues (older out-of-production parts, replacement parts for seats, other parts that do not need to conform to highly technical specifications from manufacturers). When they will need to reverse engineer parts, it’s likely to create replacements for parts that are no longer in production or no longer pose IP risks; reverse engineering a part that has IP concerns is a huge legal risk to take!

Thanks for writing this Gavriel! The competitive dynamics of whether or not to cooperate with other companies regarding machine learning is very interesting to me. I am concerned about cooperating with partners, either with IBM or with other mining companies, unless Rio Tinto believes that it cannot develop a comparative advantage in driving down operating costs or identifying potential new deposits. Currently, if Rio Tinto is the best at deploying data to improve operating costs, I believe that they should hold that information closely so that they could be the low cost provider and generate higher profits. If Rio Tinto begins to share this information, or even the scientists deploying the software behind this information, with comparable companies, they will be unable to use data to beat the competition. In this line of thinking, I would even be concerned about using a company like IBM to help them: it’s likely that they’d be able to use your data to provide competitors with better recommendations than if you hadn’t provided that information in the first place.

In summary, I think Rio Tinto needs to decide whether or not data is a comparative advantage for them (or needs to be such an advantage), regardless of data scientist availability issues, before they decide whether or not to cooperate with other companies.

Great work Tanner! I was curious how the relationship between CircleUp and established players in the CPG data collection space like Nielsen, was unfolding, and I found that CircleUp recently partnered to share data with Nielsen. Because of this, I think CircleUp should double down on being a machine learning company that partners with other data providers to provide actionable insights to its investing arm, and not being the primary source of data and insights. Looking forward, I think that CircleUp will have to choose between being an investor that uses Helios and a company that provides machine learning-based suggestions from Helios to other companies. I see these two business lines as being in conflict with each other: why would a company trust CircleUp to provide it solid recommendations if it is invested in its competitors? I believe that CircleUp could provide co-investment opportunities for existing companies that wish to grow their portfolio of brands, but I’m not sure that they’re best positioned to be giving those companies advice on existing products.

Very interesting, Miguel. To me, a great next step for Fox would be to apply this technology to entire movies, to see what moments should be included in a trailer. By feeding an entire move through Merlin, Fox could potentially build better advertising campaigns. Perhaps they could better choose which moments to include, but from an audio and visual perspective, and build more effective trailers using that information.

To answer your second question, whether machines can predict something so subjective as art, I think that there are a lot of patterns in movies that Fox could take advantage of to understand whether or not a certain movie could potentially be popular. For example, superhero and action movies often follow predictable plot lines–perhaps in these genres a natural language processor that reads scripts could be more effective than in other genres to predict the success of certain scripts. However, because movies are art, I ultimately believe an algorithm like this would have to be used in conjunction with human readers to provide another data point. This data point could help readers reconsider scripts that they did not initially identify with, or it could help readers to make go/no-go decisions about whether or not make films that they are not completely sure about.

Thanks for the article. I see a lot of similarities between the military’s deployment of AM technology and in rail (similar to your comment on my post). It appears that both are experimenting with AM around the low-risk edges of heavy equipment (e.g. on the bumper of an F-35), but are not yet willing to put those parts into safety critical parts of the plane. It will be fascinating to see how AM technology advances to the point where these parts can be integrated into the “core” part of heavy equipment.

Looking into where AM can be deployed most rapidly, areas like 3D printed barracks, and other opportunities to reduce the cost and logistics requirements of rapidly setting up operations in combat zones appear to be the easiest to adopt in the near term. Once the military and other organizations can get used to deploying the technology consistently in non-critical environments, perhaps comfort and expertise will build to the point where they will deploy the technology for critical parts.

One thought on the consolidation of innovation arms of the military: because each section of the military has different operational and logistical requirements and could possible benefit from AM in different ways, does consolidating these departments concern you from an innovation perspective? Could it be better to keep them separate, but focus on coordination and sharing best practices/findings of AM, so that while each organization can be learning AM, they can separately decide how to best apply the technology for the specific military branch/operation they are tied to? I am concerned that consolidation could suppress AM use in innovative applications from this perspective.