Roberto A. Santiago · Applied AI Research · Technology Innovation · Management Strategy
The most valuable thing about AI is everything it can't do.
The judgment, the relationships, the mastery people count on, and the mission they trust you to fulfill: that's what AI brings into focus. No machine can replace that, ever. I've spent thirty-five years mastering the creation of intelligent systems and delivering technology into the hands of everyone. What started as a vision of intelligent technology serving human potential is now here. Let me help bring the best of you and your team to the world.
The Work
Three ways I serve one human mission.
A human-first approach to AI is a mission big enough to need all of us, and the moment is urgent. It takes work on more than one front, so I serve where I can make the biggest difference: fostering the new leadership this moment demands, advancing AI toward forms we can reason with and trust, and doing the practical, ground-truth work of strategy, management, and engineering that moves an organization forward.
We make a difference by building the things that change everyday life.
Strategy
Navigate the shift to maximizing human value, the work AI can't do, and weigh the real risks and rewards that come with it.
Prototyping
Prove an idea fast with grounded, working prototypes, before anyone bets the roadmap on it.
Planning
Bring new and unfamiliar capabilities into your organization in a way that carries your people and your infrastructure along.
Engineering
Always available to step in and help with any engineering need you have, to help your organization move forward.
AI must continue to evolve in order to serve real human needs.
-
Stateful convolutional networks
Built upon Paul Werbos's foundational work in recurrence and reinforcement learning, I've created the SCN architecture, a novel neural substrate inherently focused first on reasoning over pattern matching.
-
Generative reasoning
Just as generativity transforms how AI handles language, reasoning can work the same way: a system learns about reasoning in one situation and carries it into the next.
-
Knowledge graphs and ontology
Knowledge graphs anchor what AI says in checkable truth and guard against hallucination. They can be trained, with reinforcement learning, to make it rarer still. Combined with other methods, they form a vast, structured store of human knowledge, which is ultimately what makes AI useful to human purpose.
Leadership is the key difference now, more than ever.
- Speaking. Talks and interactive engagements that bring people into the urgent need for leadership right now. In Dialogue with AI is one example, gathering students, community leaders, and people of all ages to understand what AI really is and how to meet it human-first.
- Partnering. I partner with public and private organizations, from new ventures to mission-driven nonprofits, bringing years of work across both AI and technology to help build the kind of organizations this change calls for.
- Advising. Even seasoned leaders are navigating hard, shifting ground. Someone who has lived in technology and AI this long makes it easier to see where to build and where not to, so you stay focused on people instead of drowning in the details.
The thinking
Roles were never the truth about people.
They were a compression scheme, a shortcut we invented because no one could hold the full reality of who can actually do what. That limit is lifting. The most interesting thing about this moment isn't that AI can do our work; it's that AI, used well, could finally let us stop boxing one another.
The catch is trust and alignment, and that work is ours, not the machine's. It is built first by sharing our stories and breaking bread, and now, urgently, by refining AI to rely on reasoning rather than just patterns of words, kept grounded in knowledge that comes from us.
The love the world needs can never come from AI, only from us. After decades in this field, the one thing I most want you to know is simple: do not be afraid.
Public lecture series
In Dialogue with AI
A six-part series that uses the defining technology of our moment to re-examine our relationship to one another. Not AI as a tool to master or a threat to repel, but a sustained question: who are we in dialogue with when we use AI, and what is it teaching us about ourselves?
Demystifying AI.
Seeing plainly what AI is, and learning to look at it without projection.The AI Culture.
How AI quietly became the medium between us, and what that costs.The Promise of AI.
The limits it can lift, and its promise of attention for every person.The AI Dilemma.
The real costs, from resources to the distortion of human relationships.AI and Humanity.
A someone, a something, or a chorus of someones? The mirror it holds up.The Myth of AI.
Past savior and demon, back to our oldest questions about ourselves.Forthcoming. Inquiries about hosting the series are welcome.
Thirty-five years
It looks like many roles. It's really one road.
I learned reinforcement learning and recurrent architectures directly from Paul Werbos, the inventor of backpropagation through time, and have spent every year since finding the places where machine intelligence meets the work people actually do.
Reasoning that stays grounded in us
Building neural substrates that reason first and speak after, grounded in trustworthy, curated knowledge.
Knowledge graphs and graph networks, in production
From petabyte-scale retrieval to a graph network that helped discover new materials.
Where brain met machine
Unifying spike-timing plasticity with reinforcement learning; the first work on context discernment.
Modeling knowledge so machines could share it
Semantic models of data, information, and knowledge: an ancestor of today's ontology-driven integration.
Large-scale ML in production, before the field existed
Neural-network forecasting and optimization for airline revenue management at USAirways and IcelandAir.
Learning the foundations
Mentored by Paul Werbos in adaptive dynamic programming, neurocontrol, and the roots of reinforcement learning.
Let's talk
Tell me your story.
Whether you're wrestling with grounded, trustworthy AI, thinking about what comes next, or just want to talk through a hard problem (one of my favorite things to do), I'd like to hear from you.