ITAI 1370 — AI History, Theory & Platforms

End-of-Year Portfolio — Cesar Zaldivar | May 2026

Reflections & Conclusion

ITAI 1370 — End-of-Semester Reflection

Carl Sagan once warned that when almost no one understands science and technology, we create “a combustible mixture of ignorance and power” — a prescription for disaster. That warning feels more urgent than ever in the age of artificial intelligence.

That quote opened this portfolio’s homepage, and it accurately captures what this semester meant to me. ITAI 1370 was not simply a course about technology — it was a course about becoming a more informed participant in a world increasingly shaped by AI.

The Journey Through the Semester

Where We Started: Foundations

The semester began with the big picture. Dr. Anandkumar’s Trinity of AI framework — Data, Algorithms, Compute — gave me a lens for understanding why AI advanced so rapidly in recent decades. The Dartmouth Conference of 1956 then provided the historical anchor: the problems McCarthy, Newell, Simon, and Solomonoff identified in a single summer — language, neural networks, self-improvement, creativity — are the exact same problems researchers are still working to solve today.

The Middle: Architecture and Scale

The middle modules took a deep dive into the architectures that define modern AI. AlphaGo Zero showed that removing human bias from training data can produce superhuman performance. GPT-3 demonstrated that scale itself is a form of capability — 175 billion parameters unlocked few-shot learning that no previous model could achieve. T5 offered an elegant counterpoint: instead of brute-force scale, a unified text-to-text framework can make one architecture serve dozens of different tasks. Studying these systems side by side made the trade-offs between approaches genuinely clear.

Hands-On Learning: Experiments That Stuck

The most memorable learning moments were the hands-on experiments. Building a neural network in TensorFlow and comparing ascending vs. descending architectures challenged my assumptions about which designs perform better. Training a computer vision classifier on Google Teachable Machine made abstract concepts like epochs, overfitting, and loss functions tangible. The GAN storytelling exercise was unexpectedly profound — playing the role of a discriminator taught me more about what makes AI outputs feel “real” than any textbook definition could.

The Bigger Picture: Platforms and Society

The final modules broadened the lens from models to systems. The AI assistants comparison revealed that technical capability and genuine conversational intelligence are not the same thing. The IBM Watson lab reinforced that even powerful platforms fail when their user experience does. The predictive AI essay brought everything together: AI is not magic or consciousness — it is pattern recognition at massive scale, with enormous implications for privacy, autonomy, and accountability.

Module Summary Table

Module Topic Key Takeaway
1Trinity of AIData + Algorithms + Compute drive every AI advance
2Dartmouth ProposalAI’s goals in 1956 are still AI’s goals today
3AlphaGo ZeroRemoving human bias can produce superhuman results
4Raytracing + QR CodesGPU advances democratize photorealistic rendering; QR codes enable AR
5GPT-3 + Visual AuthenticityScale unlocks capability; AI blurs the line between real and generated
6T5Elegant unified frameworks can outperform brute-force scale
7Neural Networks + Deep TrafficEmpirical results often contradict theoretical expectations
8GANsCompetition between generator and discriminator drives iterative improvement
9Computer VisionPerfect training accuracy can mean overfitting, not generalization
10IBM WatsonUX quality determines whether a platform demonstrates its capability
11AI AssistantsDesign purpose directly shapes and limits conversational ability
12Predictive AIAI calculates statistical likelihood — it does not truly “understand”

Looking Ahead

This course gave me the historical context, technical vocabulary, and critical thinking tools to engage meaningfully with AI — as a user, a student, and eventually a professional. The field is moving fast, but the foundational questions raised at Dartmouth in 1956 remain the same: Can machines learn? Can they reason? Can they create? And perhaps most importantly: Should they, and on whose terms?

Carl Sagan was right. Understanding this technology is not optional. It is a civic responsibility. I am grateful for a course that treated it that way.

Acknowledgments

Thank you to Professor Maryam Esmali for designing a course that balanced history, theory, and hands-on practice in equal measure. The breadth of assignments — from video reviews to neural network experiments to creative storytelling — made for a genuinely engaging semester.

Cesar Zaldivar — ITAI 1370, Houston Community College — May 2026