Amplifying Human Agency Through Intentional AI Design

Key Quotes

"my daughter is actually prompted me because they they are asking a lot of questions about ai the futures of their future the futures of jobs and education and um and i said like maybe it's time that somebody that works in technology has an attempt to to write a book that you try to demystify some of these concepts to the broader audience"

Marcus Fontoura explains that his daughter's questions about AI and its impact on future jobs and education were a primary motivation for writing his book. Fontoura felt it was important for someone working in technology to attempt to explain these complex concepts to a wider audience.

"i think it gives me um this sense that this is more of an evolution than a revolution if you're not paying attention and then you're just looking into it and then start reading about ai you think oh my god ai is something revolutionary and it's really going to impact our lives if you are tracking how technology has been evolving over the years even um in the early 2000s when you did the first automated machine translation software and then um ibm uh beating uh kasparov like in chess and then jeopardy and then like all the evolution even as text processors right this is this is one thing that i say in the book that like in the beginning when we wrote in in word processor there was not even spell correction and then we evolved it to now we have spelling correction now we have grammar correction and now even we can write full paragraphs for you but this is an evolution"

Marcus Fontoura argues that current advancements in AI should be viewed as an evolution rather than a revolution, drawing parallels to the gradual development of technologies like machine translation and word processing features. Fontoura highlights that by tracking technological progress over time, one can see a continuous progression rather than sudden, disruptive shifts.

"i believe so and and but i also believe that we learn a lot right and one one of the things that when you think about fragility in computer systems is um is a little different right because like if you're trying to build a robust uh bridge or robust house we're thinking about building a solid house or a solid bridge that will um weather like anything any storm and then will be super solid in computer systems when we want to build reliable things um we learned that the approach is to assume that you have unreliable parts and engineer resiliency on top and that's i think was the evolution when we started seeing like better networks uh connecting more and more computers together larger systems more data"

Marcus Fontoura posits that computer systems achieve reliability not by assuming perfect components, but by engineering resiliency on top of inherently unreliable parts. Fontoura explains that this approach, which emerged with the growth of interconnected networks and larger systems, is a distinct method compared to traditional engineering disciplines like building bridges or houses.

"i think one key point that we need to learn together is um what we want these ai systems to do for us and then what and how can we cooperate with ai systems better and how they can augment our agency and they we want ai systems to be designed to amplify our humanity right so like for instance a simple scenario is like um if you had and of course like that is like a um out there but like i'm just wanted it to be thought provoking but like let's say if you had a perfect uh robot tutor that could teach your kids to go to harvard but but you had to delegate the education of your kids to this robot like would you want that or not"

Marcus Fontoura emphasizes the need to define the desired roles of AI systems and how humans can cooperate with them to augment their agency, aiming for AI to amplify humanity. Fontoura uses the thought-provoking scenario of a perfect robot tutor to illustrate the complex decisions humans face regarding delegation and the preservation of inherently human activities.

"yeah it worries me too especially in the current evolution that we are in the ai right i think these systems will will get better and do more but it has been an increasing evolution like since the uh early innings of uh computing we are always looking into this task of like how can we make the job of the programmer easier and easier and so that the programmer can focus on on the business logic or the or the really important concepts that really capture the problem that they are trying to solve and not the all the glue that we have to put together and i think this is a great evolution and i think ai can really automate a lot of of that glue and make the life of the this the expert engineer even better"

Marcus Fontoura expresses concern for junior engineers in the current AI evolution, acknowledging that while AI can automate much of the "glue" code to benefit expert engineers, it may not assist those lacking foundational knowledge as effectively. Fontoura views the ongoing effort to simplify programming tasks as a positive evolution, with AI poised to significantly improve the lives of experienced engineers.

"i would say that um i i probably for azure i probably hold like a very shallow understanding of like uh a very shallow understanding of like the end to end system this meaning that i don't understand the details of many things and there is there is a core part in the the infra that i think i have a deep understanding and i i would i would think that the the architects that put these systems together would be like me right like there'll be our storage architects that know a lot about the storage but then they don't know much about uh how the networking systems are configured because i'm i'm doing this role of like end to end architect architect for azure maybe i'm even more shallow than most people in in the details of each one of the components but i have a pretty good understanding of the end to end flow of of of things"

Marcus Fontoura describes his understanding of Azure's architecture as shallow in terms of component details but deep regarding the end-to-end flow, a perspective he believes is common among architects. Fontoura notes that specialists, like storage architects, often have deep knowledge in their domain but less in others, highlighting the necessity of diverse expertise within teams.

"we want people to be working on the latest and greatest uh technologies so that they they have like great tools so that they feel productive and um and if they feel productive they can achieve more with less but but then i don't feel that like oh because i'm achieving more with less we should have a smaller team and that's not the argument it's more when we achieve more with less we can do more and then and then this flywheel keeps going up and up right so we do more invent new technology and then we are more productive we have more resources to invent yet more technology and we are yet more productive and then i think that that's like a the holy grail for me"

Marcus Fontoura advocates for investing in the latest technologies to boost productivity, arguing that increased efficiency should lead to doing more, not necessarily having a smaller team. Fontoura describes this as a virtuous cycle or flywheel, where greater productivity fuels innovation, which in turn leads to further productivity and resource generation for even more innovation.

"we cannot live without water we cannot live without gravity but you can live without tiktok and then that whole discussion is like uh related to what we were discussing uh talking about before that um which