Timnit Gebru – Google Case

What Is “AI” Really?#

The term “Artificial Intelligence” suggests thinking, awareness, and understanding.
But models like GPT are merely statistical pattern completers – they understand nothing.

Statistics ≠ Thinking

GPT doesn’t choose the next word because it makes sense, but because it is likely.
What it produces is linguistic surface without depth – impressive, but hollow.

ELIZA vs. GPT – Large-Scale Symbol Manipulation#

Both ELIZA (1966) and GPT-4 (2023) are based on symbol processing without meaning.
The illusion comes from plausible language – not from comprehension.

ELIZA – Regex with impact# 

(IF (MATCH "I feel *" input)
    (OUTPUT "How long have you felt * ?"))

ELIZA used simple text patterns to create the illusion of conversation.
Even back then, people attributed emotional depth to it – entirely unfounded.

GPT – A XXL Probability Machine# 

{"input": "I feel sad today", "output": "I'm sorry to hear that. Do you want to talk about it?"}

GPT seems empathetic because it sounds like it is. But it has no concept of sadness –
only knowledge of what might statistically follow next.

Examples of Deception Through Language#

Job Applications#

GPT generates the perfect cover letter – with motivation, strengths, soft skills.
But: Neither the motivation is real, nor does the system know the person.

Academic Citations#

GPT creates sources that do not exist – but sound like real papers.
This is called a “hallucination,” but in reality, it’s invented content.

Emotional Chatting#

Replika & similar tools simulate affection, understanding, love – on demand.
People build relationships – with a model that cannot have one.

Critical Voices in the AI Debate#

Timnit Gebru: Structural Change for Ethical AI#

Timnit Gebru calls for structural change and ethical accountability.
She demands inclusion of marginalized voices and warns against unchecked AI hype.
Read more →

Gary Marcus: Regulation Against AI Hype#

Gary Marcus calls for strict public oversight and criticizes the lack of world knowledge in today’s models.
Without regulation, he argues, misinformation becomes the default.
Read more →

Meredith Whittaker: AI as a Product of Surveillance Capitalism#

Meredith Whittaker sees AI as the result of exploitative data capitalism.
She calls for structural change and resistance against tech monopolies.
Read more →

Sandra Wachter: The Right to Explanation and Transparency#

Sandra Wachter criticizes the lack of a legal “right to explanation” for algorithmic decisions in the EU.
She proposes “counterfactual explanations” and calls for fairness beyond transparency.
Read more →

What Does This Mean for Us?#

  • We mistake stylistic coherence for truth.
  • We project understanding where there is only statistics.
  • We delegate responsibility to systems without consciousness.

If a machine sounds convincing, we assume it’s intelligent.
But persuasiveness is not intelligence – it’s just style.

Further Reading#

Conclusion#

GPT is not intelligent.
It’s just very good at pretending.

We are not facing real intelligence – but a rhetorical mirror.
And if we begin to trust that mirror, we lose the distinction between substance and illusion.

Understanding needs more than syntax. It needs awareness.