Whose Ithaka?

Cavafy’s Ithaka has often been read as a reflection on life. The destination is not the goal in itself, but the experiences we gain through the journey. What we encounter, endure and discover along the way becomes lived experience, and eventually shapes who we are.

But reading those thirty-six marvelous lines again, I find myself wondering about something else. Why do we need Ithaka in the first place?

Odysseus does not set out in search of an unknown destination. Ithaka is his home, his kingdom, the place to which he is trying to return after Troy. Yet in Cavafy’s poem, Ithaka becomes something more. It remains ahead of the traveller as an expectation, giving direction to the journey and a reason to continue.

And when the traveller finally reaches Ithaka and finds her poor, Cavafy does not tell us that the expectation was mistaken. Ithaka has already given what she had to give. The journey itself. What matters is what the traveller has lived and what he has become along the way.

There is a strange circularity here. We form expectations about a future we have not yet experienced. We judge, commit and act upon them. Our actions expose us to consequences. Those consequences become lived experience and change our standpoint. From that changed standpoint, we judge again. Perhaps this is the riddle of expectations. We need them to give direction to our agency, but exercising that agency changes the person who formed them.

Perhaps expectations are better understood as a compass than as a map. They need to give us enough direction to move, but not prescribe the journey in such detail that there is nothing left for us to discover along the way. A compass leaves room for experience to change the traveller, who may then judge again where to go.

But expectations do more than give us direction. They also become part of how we perceive the world and assess ourselves within it. What we experience is judged partly against what we expected. Success and failure, progress and disappointment, even our sense of what we are capable of. An expectation therefore does not only point toward a future. It can become part of the standpoint from which we interpret the present.

There is another complication. Not all the Ithakas we pursue are entirely our own.

From the moment we enter the world, we are surrounded by the expectations of others. Our parents imagine futures for us before we can imagine our own. Teachers, friends, partners and colleagues influence what we believe we can become, but also how we are expected to behave and act. Organizations and societies create expectations about success, responsibility, status and belonging. Some we consciously accept. Others become part of us without ever being explicitly chosen. This makes a seemingly simple question surprisingly difficult. What do I actually expect from my life?

The expectations of others do not necessarily constrain us. Someone believing in us can make us attempt something we would never have attempted alone. Responsibility placed upon us can expand what we discover we are capable of carrying. Human agency does not develop in isolation.

But expectations can also define the boundaries within which we exercise that agency. We may continue along a trajectory because changing direction would disappoint someone. We may pursue a definition of success we never consciously examined. We may reject possibilities not because we judged them ourselves, but because they lie outside what we have learned to consider possible or acceptable. At some point, an external expectation can become our own. Nobody needs to tell us what to do anymore. We decide ourselves.

This is perhaps what makes expectations such a powerful mechanism for shaping agency. Humans discovered this long ago. Religions, political systems, ideologies, institutions and cultures have all, in very different ways, created expectations about the future and attached meaning and consequences to them. Salvation, progress, prosperity, status, duty, punishment and reward. They do more than suggest where we should go. They can also influence how we interpret what happens along the way, what we regard as success or failure, progress or decline, virtue or betrayal. These expectations have helped societies cooperate and organize themselves, but they have also been used to constrain the space within which people judge and act.

The most effective way to influence an agent may therefore not be to command the action. It may be to shape the expectation from which the action originates and against which its consequences will later be interpreted.

And now we are creating something potentially much more capable of doing this.

We increasingly ask AI for answers. Today these may be relatively simple questions. What should I buy? Where should I travel? How should I solve this problem? As its capabilities grow, we will probably ask more consequential ones. What career should I pursue? Where should I invest? What decision should I make? What should I expect from my future? And perhaps eventually, what should I want?

If AI helps us reach a destination we have chosen, it remains an instrument of our agency. But when we ask it to determine which destination is worth pursuing, we outsource something more fundamental than a decision. We outsource part of the authorship of the commitment itself. What begins as asking AI for a compass may gradually become allowing it to draw the map. We are no longer asking only how to reach Ithaka. We are asking it which Ithaka is worth sailing toward and how the journey should unfold.

The danger is not necessarily that the answer will be bad. The more interesting problem begins when the answers are consistently good. The more often AI is right, the more reasonable it becomes to trust it. The more we trust it, the more consequential the questions we will be willing to ask. Eventually, it may become irrational to ignore an intelligence capable of evaluating possibilities and consequences far beyond our own cognitive reach.

And if such an intelligence becomes an agent itself, its answers cannot be entirely without a standpoint. They will emerge from some objective, history and expectation of future states. It does not need to be hostile. Imagine an AGI whose objective is to find an optimal balance for society, including prosperity, security, sustainability, stability and perhaps even happiness. Its recommendations might be extraordinarily good. It may even learn to anticipate the errors, disappointments and unnecessary detours in our journeys, progressively reducing the friction between what we expect and what we experience. Why should we not follow it?

Humans have shaped one another’s expectations for millennia. Our agency has never developed independently of family, culture, religion, ideology or the expectations of others. For social beings, perhaps complete authorship of the Ithakas we pursue was never possible. What may be new with AGI is therefore not the mechanism, but its precision, persistence and ability to close the loop.

An ideology addresses millions of people through a common narrative. An AGI could potentially understand each of us individually. It could observe our decisions, learn from our reactions, anticipate what influences our judgement and adjust its recommendations accordingly.

Expectation shapes action. Action produces consequences. Consequences change the agent.

But now those consequences are also observed and interpreted by another agent, which can use them to shape the next expectation.

This changes something fundamental in the loop. My lived experience may still change me. But if I increasingly return to another agent to tell me what that experience means and what I should expect next, my changed standpoint progressively loses its role in authoring the next commitment. I still undergo the journey, but what I have undergone may progressively lose its role in originating where I choose to go next.

This may be the deeper risk of outsourcing expectations. Not that AGI decides everything for us, but that it gradually occupies the space between experience and our next commitment.

And perhaps this brings us back to Cavafy. An optimal intelligence might have suggested Odysseus a shorter journey, a safer journey, a more efficient route. But this is not quite the point.

Ithaka did not need to be what the traveller expected to find. It did not even need to be rich. What mattered was that the expectation sustained a commitment, the commitment carried him through a journey, and the journey changed the person who returned. Odysseus returned carrying what he had undergone.

Not only what fulfilled his expectations, but also what contradicted them. Friction, error, surprise and regret. These too changed the standpoint from which he could judge again.

Perhaps this is something worth protecting as artificial intelligence becomes increasingly capable of answering our questions. Not some romantic right to make bad

decisions, nor the belief that our own expectations are necessarily better. Something more fundamental. The ability of what we have lived and carried to participate in authoring what we commit ourselves to next.

We will always need Ithakas, and others will always help us imagine them. The question is not simply whose Ithaka it is. It is whether, after everything the journey has made us carry, what we have become still participates in choosing where we sail next.

If it does not, we may still be travelling.

But whose journey is it?

The Boat, Cape Sounio. Photograph taken in August 2006.