5. Findings
5.4 Ready-Made-AI Adjusted
5.4.1 IDA
In this project, Ready-Made-AI was primarily used to establish what the technical grounds for the project should be. Both Martinsen and Bach Keldsen agreed on the focus for their collaboration and this focus was chosen relatively quickly as a result of the workshop (Appendix 3, l. 158-166, Appendix 4, l. 568-569). They do, however, differ in how they believe the focus was chosen, in Bach Keldsen’s opinion the focus for the chatbot was chosen due to budgetary constraints:
“
So we were very clear in what we had ((in terms of budget)), or what we wanted, and then they scoped it”(Appendix 4, l. 181)
.
Bach Keldsen clearly had an exploratory approach to the project as she, and by extension IDA, had limited knowledge about the possibilities within AI and were therefore not willing to commit vast financial sums into the project. This was perceived somewhat differently by Martinsen, who recognised the project as a proof of concept for IDA, but the project could have been far more reaching than it was. This became evident when Martinsen considered the budgetary constraints of the project:
“
Yeah, especially when we talk about AI and because it's still a black box for many customers a little bit afraid to say ‘ok so we're gonna put in half a million’ for example for this project.” (Appendix 3, l. 176-177).Martinsen was speaking generally, however, we believe that IDA in Martinsen’s view, is one of the companies, which sees AI as a black box. IDA are therefore hesitant to commit vast financial resources to an AI project - at least for the moment. Ultimately, the size of the budget limited the scope of what was possible to execute for IDA. It started with a general focus on chatbots, but was later limited to a chatbot for car insurance (Appendix 3, l. 176-180). In sum up, Bach Keldsen and Martinsen share the understanding that the project was scoped by its budget. Martinsen believes that this constraint was imposed - not due to a lack of financial means - but as evidence of a certain reluctance regarding AI on the part of IDA, whereas Bach Keldsen saw it as due diligence.
Nonetheless, Ready-Made-AI has, in this case, been used as forum for scoping the project with regards to the budget available rather than a forum for discussing all possible uses of AI by IDA regardless of budget. In other words, Ready-Made-AI has not been used as a way to create a shared understanding of the two parties’ aspiration for the technology - rather it was a forum in which Bluefragments presented what was possible only within AI’s current technical limits.
Bach Keldsen did not immediately recall what Ready-Made-AI was, but recognized the name
- so the concept was not front and centre in the way she remembered IDA’s collaboration with
76
Bluefragments (Appendix 4, l. 193-194). It seems that the concept of Ready-made-AI might not have been a centrepiece because the collaboration was neither perceived to be a conceptual discussion about what exactly an AI is. In fairness, neither Bach Keldsen nor Martinsen mentioned the need for such a discussion. The workshop was primarily a practical discussion about IDA’s technical needs for an AI. An example of such a practical need could be the use of the Azure platform. Therefore, Martinsen saw the project as a practical proof of concept, meaning the chatbot was a way to show the value of AI:
“
We can prove some cases, you can prove the business case and based on that we can actually expand the project ((with IDA)).” (Appendix 3, l. 84-85).
As evident in our findings, there was a strategic element to IDA’s adherence to the Azure platform.
However, Martinsen did not see Ready-Made-AI as something that would accommodate IDA’s strategic need, as he did not see the strategic considerations IDA had with regards to how the AI should be developed. Nonetheless, we have found no indication that Ready-Made-AI was used as a forum for a discussion of the strategic value ascribed to the Azure platform by IDA.
What was also evident in our comparison in chapter 5.3, is that the two parties perceive general AI differently. It is not clear whether the use of Ready-Made-AI has done anything to
accommodate these differences either. In fact, Bach Keldsen was not even aware whether there was any distinction between her and Martinsen’s definition of AI:
“
I don't think ((there was a distinction)), I don't know. Maybe there was, I didn't pick up on it, if there was.”(Appendix 4, l. 177)
.
Martinsen did point to the fact that there was a difference in opinion between them:
“
I think Lisbeth ((Bach Keldsen)) has a perspective on AI where (.) well I think she is a typical user of AI. And I believe she's pretty open-minded around the AI. And hence, I think we share some of the common ideas about AI. But I'm not sure we have the same perspective on things.” (Appendix 3, l. 93-95).A typical user in this context refers to someone who needs to first adopt and adapt a technology
before fully integrating its use. Martinsen gave an example of how Facebook was first adopted by
young people before eventually being adopted, adapted and integrated by older generation who
now uses as much, if not even more, than the younger generation (Appendix 3, l. 105-114).
77
When Martinsen says that Bach Keldsen fits the description of a typical user, it would seem that he has not imposed a set definition of how and for what an AI should be used for.
Martinsen’s use of Ready-Made-AI, in IDA’s case, has focused on the development of a practical solution that would accommodate IDA’s needs and therefore a conceptual discussion has not been necessary. Nonetheless, as evident in our findings, there are differences in Martinsen’s and Bach Keldsen’s definitions of AI. However, due to the practical nature of the discussion these differences were never brought up, hence Bach Keldsen was not even aware that there was a difference of perception with regards to their definitions of AI, as mentioned in 5.3.
Martinsen has consciously tried to establish terminology with IDA, however, he is not sure that he and Bach Keldsen share the same terminology:
“
Well that's hard to say ((whether a successful terminology was established)), I really hope so, but I’m not sure.” (Appendix 3, l. 136).
Again, it is evident that the workshop conducted, as part of Ready-Made-AI, has created a shared understanding with regards to the technical requirements of the project. However, a more fundamental discussion of what AI actually is has been neglected, which Martinsen also acknowledges, as he is unsure that a common terminology has been established.
Despite their different perceptions, both Bach Keldsen and Martinsen believe they have had a successful collaboration, Bach Keldsen presented Bluefragments quite positively:
“
Very professional. We had a really good period of time together. It was a very interesting way of working.”(Appendix 4, l. 554-555)
.
Martinsen agreed that the collaboration had been successful between them:
“
Yeah I think so, we had some challenges along the way, but I think they got a pretty good result out of it”(Appendix 3, l. 140-141).
Clearly, the project is seen as a success by both parties as they could mutually agree on a technical
solution. Despite that their definitions of AI are divergent, the project has created a common
understanding between the parties. However, this understanding is centred around a common
technical understanding of IDA’s requirements for the AI rather than a shared understanding of AI as
a concept. Therefore, Ready-Made-AI was not used to discuss the strategic value IDA ascribes to AI.
78
In document
Perceptions of Artificial Intelligence
(Sider 77-80)