Google Gemini:
Turing:Excerpts provided by Gemini, an AI model by Google (2026).
The Turing Test was actually a test of human perception, not machine consciousness.
Now that the “mystery” of conversation is fading, the fear is shifting toward the “black box” of AI reasoning
Crazy Smart book experiment:
Author: Mo Gawdat
Google Gemini Summary:
The experiment you’re referring to from Mo Gawdat’s book Scary Smart is the landmark 2020 MIT study that led to the discovery of a powerful new antibiotic named Halicin.
Gawdat uses this example to illustrate “the edge”—the point where AI moves from being a tool to making leaps of logic that humans simply cannot replicate. Here is the breakdown of that experiment and why it shocked the scientific community.
The Experiment: Finding the “Unfindable”
For decades, the discovery of new antibiotics had stalled. Traditional methods involved humans looking for molecules that “looked” like existing antibiotics. The MIT team, led by James Collins and Regina Barzilay, decided to change the approach:
The Training: They trained a deep-learning algorithm on about 2,500 molecules, teaching it which ones were effective against E. coli and which were non-toxic to humans.
The Search: They then set the AI loose on a library of over 6,000 compounds.
The Discovery: Within hours, the AI identified a molecule that was structurally completely different from any known antibiotic. The researchers named it Halicin (after HAL 9000 from 2001: A Space Odyssey).
Why the Scientists Were Stunned
What Mo Gawdat highlights in Scary Smart is the “super-human” nature of this inference:
Non-Obvious Logic: Humans would never have tested Halicin as an antibiotic because it was originally being researched as a drug for diabetes. Its chemical structure didn’t “look” like a killer of bacteria to the human eye.
Decades of Work in Hours: The scientists admitted that using traditional laboratory methods, it would have taken decades (or they might never have found it at all) to stumble upon this specific molecule among the millions of possibilities.
Unique Mechanism: Halicin doesn’t just attack bacteria the way old drugs do; it disrupts the bacteria’s ability to maintain an electrochemical gradient (basically short-circuiting their battery). The AI figured out this “rule” for killing bacteria without being explicitly told to look for it.
The “Scary Smart” Takeaway
Gawdat points to this as a “Black Swan” event for AI. The AI didn’t just “calculate faster”; it showed insight. It looked at the data and found a relationship between atoms and survival that thousands of the world’s best biological minds had missed for sixty years.
As Gawdat notes, once the AI was “taught” what a killer looks like, it surpassed its teachers instantly, identifying a weapon against “superbugs” (like Acinetobacter baumannii) that are currently resistant to all known human-designed drugs.