[Music] what do you want to be a creator of ai or a user of ai that's the key question we want to talk about in this video we want to think about firm boundaries and make or buy decisions and strategic positioning when it comes to ai technologies throughout we will make a distinction between downstream firms the ones for which ai is more of a tool than a product and upstream firms the ones for which ai is more of a product than a tool we will see strategies for each type of firm and characterize the circumstances
under which each of them should be chosen well this all sounds quite general are there any specific issues that arise when we talk about ai or machine learning as a tool or product the answer lies in the complex nature of data and its interactions with algorithms and models and its interactions with computing hardware the machine learning value chain that i discussed in another video the key takeaway is that the value chain is not linear and there are complex interactions for example it is less than clear whether data and algorithms are complements or substitutes if this
sounds new to you go back and watch that video first let's start upstream the big question here is whether you want to integrate and be active in downstream markets or whether you want to license your technology to other firms that are already active in those downstream markets let me elaborate the red firm is integrated whereas the green and blue firms are active in an intermediate market for technology and license their technology to firms in downstream markets two firms have decided to produce only the upstream technology and sell it in an intermediate market the other firm
has taken the decision to produce upstream technology and supply it directly on downstream markets this strategy is to specialize in generality your technology is so general that not only you but multiple firms further downstream find it valuable when would you want to specialize in generality can think at least of two conditions say you don't have a lot of prior assets in a downstream market then you would need to invest first before you can enter this should make it more interesting to specialize in generality and similarly if there's broad intermediate demand then you would also want
to specialize and sell licenses rather than entering a potentially already quite competitive downstream market now let's switch perspectives and think about strategies of downstream firms here the question is really whether you want to make or buy machine learning and this question about how far you want or don't want to reach beyond your organization's boundaries does not only relate to algorithms but also to data the computing power and so on the entire machine learning value chain one reason to buy ai and some or all of its components from a market is that specialization in upstream markets
can allow downstream firms to access intermediary inputs at lower cost or at higher quality than if the corresponding technology would have been developed or produced in-house another reason to buy or license from upstream markets exists if the value of machine learning is coming from adding firm specific assets such as proprietary datasets hence you bring something rare and non-imitable to the value chain that can lead to improved processes and products and therefore create a competitive advantage on the flip side what are reasons to do it yourself what are reasons to integrate into parts or the entire
machine learning value chain well the first reason comes from unfavorable economics on the upstream market there might be technological complementarities or market power that could create switching costs that force downstream firms to choose the same suppliers in multiple stages or limit their ability to use complementary inputs from other suppliers another reason to integrate is if the upstream market is concentrated and many downstream firms receive identical upstream inputs if everybody has access to the same resources they are no longer rare and difficult to imitate then it is better to specialize in some or all parts of
the machine earning value chain thanks for bearing with me now you have a better way of thinking about whether you want to be a creator of ai or a user of ai and so often it depends on where you are upstream or downstream and what your external environment looks like but most importantly taking decisions of specialization and make or buy have additional complexity because ai and machine learning in particular is not a singular technology but consists of interrelated components
