Agency

Eric Schmidt has a new TED Talk: The AI Revolution is Underhyped. The Revolution might be underhyped but the current results from investments in AI are far from satisfactory in our view. What work needs to be done and how much more investment is needed to really make AI deliver on the promises made by the technology megalopolis? These are key questions we are being asked by business leaders today.

But before we go there, we take a segway into an important definition. Let us review the concept of ‘agency.’ Why? Because that is where AI must go if we are going to see the societal benefits we have been promised. Agency is action, as in your ability to act and or exert power, according to the Oxford Dictionary. At the divine level, Agency is a God given attribute.

Harvard Business Review had an excellent piece on Agentic AI –the inherent ‘agency’ in Agentic AI is no accident. In the article written by Mark Purdy, he argues that “…Recent advances in agentic AI bring that autonomous future a step closer to reality. With their supercharged reasoning and execution capabilities, agentic AI systems promise to transform many aspects of human-machine collaboration…’ In other words, imagine a future where AI has as primary attribute of agency. It reasons and acts based upon objectives set forth either by external or internal prompts. The risk of course is that this form of AI reasons and acts, producing undesirable solutions, which are without the stated or implied intent of real people. Think about the character HAL in the movie 2001: A Space Odessey – ‘Open the pod bay doors, HAL!’

A simple example of how giving up your agency without giving it too much thought is autonomous transportation. You ‘authorize’ an autonomous taxi to pick you up and then the vehicle and its AI systems can do whatever they deem necessary to get you where you want to go. Paying money for the ride is the least of your concerns. What if the AI decides the best way to get you from point A to point B is to drive on the sidewalk and mow down a few pedestrians in the process. Or it takes you down a street where there is a high probability of dangerous conditions. But wait you say, wouldn’t the programmers have ‘safety’ as a top objective? Maybe. What if profit is the primary objective and safety is woven in there somewhere with the user experience requirements. We often assume a great deal when we give up our agency.

Now you can see pandora’s box, are you going to open it? Should AI be running our governments, organizing our schedules, and paying our bills? As a civilization we will, many without fully realizing it. It starts with the desire for convenience. Still, we should consider how much of our inherent agency we relinquish in the name of ‘progress,’ or ‘productivity,’ or ‘convenience.’ People are inherently lazy, almost all of us anyway, maybe not Eric Schmidt, but certainly includes the tech teams programming AI.

For AI to deliver on the big promises made, we must carefully consider who and or what is building these systems. The intersection of AI and agency should create an opportunity to bring new perspectives into the design process. Linguistics experts who can shed insight into the contextual nuances of language, historians that have a backdrop of the past, and even theologians that can inject moral checks and balances into our planning and eventual AI programming. Before you give up your agency, remember a slave, digital or otherwise, is someone that has no agency.

As we conclude this week, we are reminded that AI runs in Data Centers. Planning your Data Center and infrastructure is just as critical as the software running through it. Additional insight from Mr. Schmidt, who contents that the AI Data Center itself is a strategic asset. According to Schmidt, you can and should expect digital as well as physical attacks on Data Center infrastructure in the future, particularly as we move globally into a war like posture in the coming years. When your agency is at stake, consider carefully who you are partnering with.

Call us, we are your ‘agents’ in AI Data Center design and development.