Video summary

#251 – IA e Semiótica, com Lúcia Santaella

Main summary

Key takeaways

News and Commentary

Episode Overview

In this episode, host Marcos Carvalho Lopes discusses with Brazilian semiotics scholar Lúcia Santaella whether artificial intelligence (AI) is “truly” intelligent. Drawing on philosophical semiotics, especially Charles Sanders Peirce, they rethink intelligence, language, responsibility, and education.

1) Intelligence isn’t a single human property—and AI shouldn’t be judged by a human-only yardstick

  • Santaella argues that people often assume there is one model of “intelligence,” typically equated with human intelligence.
  • In Peircean terms, intelligence and thought can be understood as forms of semiosis (sign-processes) that do not need to be restricted to the human brain.
  • She criticizes earlier human-centric claims (e.g., “AI is dumb,” “as dumb as a typewriter”) for failing to ask what “intelligence” means in the first place—or for treating human intelligence as the universal standard.
  • Using examples from biology and even non-animal life (e.g., fungi), she suggests that proto-intelligence—intelligence-like sign processes—may exist beyond conventional human cognition.

2) Anthropomorphizing AI is dangerous: it fuels false demands for human-like moral agency

  • A key warning is that projecting human traits onto machines—especially when chatbots “simulate” meaning—creates expectations AI cannot meet.
  • Santaella emphasizes that AI producing language does not imply human-like attributes such as consciousness or moral responsibility.
  • The appropriate response is to evaluate AI outputs critically, rather than treating AI as a moral agent.

3) Education must incorporate AI while training “care” and “suspicion”

Santaella argues society needs two complementary intellectual attitudes toward AI:

  • Art of care: use AI responsibly, support learning, and avoid harm.
  • Art of suspicion: interrogate errors, hallucinations, omissions, and the reliability of outputs.

She ties this to education as a major responsibility:

  • Educators and institutions influence how knowledge is produced.
  • Therefore, they should develop AI literacy rather than ignoring or uncritically adopting AI tools.
  • AI literacy for faculty is described as necessary to guide students without creating “cognitive loss” or undermining learning.

4) From semiosis to truth: AI systems operate via language, statistics, and human data biases

  • Santaella frames AI language behavior as statistical/semiotic patterning (“statistical language”) that can look like understanding without being the same as human interpretive agency.
  • AI models are trained on human-generated data, so biases embedded in language and society cannot be fully removed through “ethical design” alone.
  • She highlights how language “tricks”—hidden structures, assumptions, and distortions—help bias enter AI systems.

5) Abduction vs. deduction: AI doesn’t do human-like hypothesis formation

Using Peirce, Santaella distinguishes reasoning modes:

  • Deduction: logic moving from premises to conclusions.
  • Abduction: hypothesis-making—an instinctive/logical hybrid connected to real-world discovery and survival.

She suggests AI may generate logical hypotheses or answers, but does not truly perform abduction in the way humans—and other living beings—do, because AI lacks the instincts and experiential grounding that make abductive leaps meaningful.

6) Technology, language, and power: language “externalization” and data ownership

Santaella expands the discussion beyond algorithms:

  • Language evolved from bodies into external supports (writing, technologies).
  • Over time, the means of language production became captured by capitalism and large corporations.
  • For her, contemporary AI problems involve not just technology, but also ownership and control of data, shaping regulation and governance.

7) Generative AI, fake news, and responsibility remain human obligations

  • She links semiotics to misinformation and disinformation: AI can generate persuasive outputs, but responsibility remains with humans who set goals, deploy systems, and interpret results.
  • She pushes back against claims that AI is racist/sexist as an agent, arguing instead that these biases reflect biased human data and development choices.

8) Closing philosophical positions and recommendations

  • Santaella reiterates her preference for Peirce as a “triune logic” thinker:
    • abduction / induction / deduction
  • She contrasts Peirce with William James:
    • Peirce’s semiotics is framed as a general philosophical science.
    • James is framed as more psychological and specialized.

Book and author recommendations

  • Tatiana Roque, A máquina dentro de nós (Portuguese)
  • Yuk Hui, Cosmotechnics / Cosmotechnical art (not translated into Portuguese, but recommended)
  • She also mentions her interest in Tonzé (Tom Zé) and semiotic interpretations of tropicalist culture.

Presenters / Contributors

  • Marcos Carvalho Lopes — host
  • Lúcia Santaella — guest, semiotics scholar

Original video