Reflections & stories
Ideas, experiences and learnings about education, technology and coexistence.
Showing 12 of 20 articles
What remains when the enthusiasm passes: The artificial intelligence that cannot be seen
After the enthusiasm the maintenance work begins. Initial performance does not guarantee continuity, updating or attention to incidents.
Read article
The price of not looking at how it works: The artificial intelligence that cannot be seen
Not looking at the supply chain comes at a price. The single interface hid models, storage, and services placed under different rules.
Read article
The question that leaves official discourse uncomfortable: The artificial intelligence that cannot be seen
The official precision deserves uncomfortable questions. An aggregate figure can hide serious damage, groups that are less well served, and results that are impossible to appeal.
Read article
A technology that needs more democracy: The artificial intelligence that cannot be seen
Artificial intelligence needs more democratic institutions. Consulting after deployment turns participation into communication, not the ability to influence.
Read article
The future that is being decided today: The artificial intelligence that cannot be seen
The future of AI is decided by current purchases and standards. Today's technical conditions will determine for years what can be audited and changed.
Read article
The detail that transforms news into a public problem: The artificial intelligence that cannot be seen
A public problem appears when the bug is distributed at scale. The same automation that reduces cost multiplies inaccuracy and makes it difficult to detect individual cases.
Read article
The conversation that families are already having: The artificial intelligence that cannot be seen
Families already negotiate uses of AI without common criteria. The conflict focused on the tool and not on learning, privacy, verification and autonomy.
Read article
The difference between innovation and dependency: The artificial intelligence that cannot be seen
Innovating should not mean being locked into one supplier. The integration saved time at the beginning and later raised the cost of negotiating or exiting.
Read article
The myth that makes the debate more difficult: The artificial intelligence that cannot be seen
The myth of a neutral intelligence impoverishes the debate. Every model incorporates decisions about what to learn, what to optimize, and what to consider error.
Read article
A useful idea to understand the present: The artificial intelligence that cannot be seen
Thinking about probabilities improves the conversation about AI. A high average rate does not describe the risk of a specific error or its possibility of repair.
Read article
The signal that we should not ignore: The artificial intelligence that cannot be seen
The change in behavior of the model is a critical signal. Validating a version does not guarantee that data, provider, or configuration will remain stable.
Read article
What we are giving in exchange for speed: The artificial intelligence that cannot be seen
In exchange for speed we can deliver professional knowledge. Repeatedly delegating a task changes the skills available when the system fails.
Read article