Chuck Brooks (@ChuckDBrooks) header
Chuck Brooks (@ChuckDBrooks) profile picture

Thought Leader: FORBES Contributor, Georgetown Prof, Named Top Tech Person to Follow by LinkedIn. Voted Top Cybersecurity Person of Year, #cybersecurity, #tech

Washington, DC
Joined August 2013
Photo by Chuck Brooks (@ChuckDBrooks)Photo by Chuck Brooks (@ChuckDBrooks)Photo by Chuck Brooks (@ChuckDBrooks)Photo by Chuck Brooks (@ChuckDBrooks)Photo by Chuck Brooks (@ChuckDBrooks)Photo by Chuck Brooks (@ChuckDBrooks)Photo by Chuck Brooks (@ChuckDBrooks)Photo by Chuck Brooks (@ChuckDBrooks)Photo by Chuck Brooks (@ChuckDBrooks)Photo by Chuck Brooks (@ChuckDBrooks)
Chuck Brooks retweeted
🧠 New Research Tests Whether AI Models Know Their Own Thoughts 📍By injecting “thoughts” directly into large language models, researchers are trying to map out AI’s capacity for introspection. Whether that amounts to introspection is still an open question. That’s the central finding of Ben Brubaker’s November 4 article in @QuantaMagazine 🤖 The experiment: Anthropic researcher Jack Lindsey amplified activity patterns inside Claude associated with concepts such as denim or vegetables, without naming the concept in the prompt. The strongest Claude version tested correctly flagged the change and identified the concept in about 20% of trials. In that experiment, the unmodified controls never reported an injection 📍The follow-up is what caught my attention. Researchers found that models could still flag a disturbance even when their ability to identify the concept deteriorated. The injected signal itself may have nudged the model to mention the concept. Detecting that something had changed seemed to be a separate ability. 🧠 The neuroscience question. Detecting a disturbance could be closer to a reflex. A stronger test would ask whether the model forms a representation of its own thinking, something like I’m thinking about apples. Drawing that distinction is difficult even in humans, and these experiments leave subjective experience untested. Note: If a model flags an unusual internal event, its explanation of the event needs independent verification. #NeuroAI #Interpretability #CognitiveScience #LLMs quantamagazine.org/new-resea… @Khulood_Almani @Shi4Tech @jblefevre60 @dinisguarda @Eli_Krumova @ipfconline1 @HaroldSinnott @ThEiLe0 @Mack1510019 @drsharwood @EvanKirstel @ahier @RLDI_Lamy @Timothy_Hughes @Ym78200 @Nicochan33 @mvollmer1 @jeancayeux @rvp @SusanFourtane @amalmerzouk @bamitav @Fabriziobustama @FrRonconii @domingonarvaez1 @c4trends @TerenceLeungSF @aure79lien @EstelaMandela @NevilleGaunt @stanleywaite1 @pchamard @corixpartners #Consciousness #ML #NeuralNetworks Credit: Kristina Armitage Quanta Magazine
2
21
2
31
897
Chuck Brooks retweeted
🧠 Google’s Quantum Computer Completed a Calculation in Five Minutes That Could Take Supercomputers 10 Septillion Years. Imagine a computer solving a problem in five minutes that would take the world's most powerful classical supercomputers an estimated 10 septillion years. That's what Google's Willow quantum chip achieved in a specialized experiment. With 105 superconducting qubits, Willow demonstrated extraordinary computational power using quantum physics. Even more importantly, researchers showed that increasing the number of qubits used for error correction could actually reduce errors, bringing reliable quantum computing one step closer. But there's an even stranger question. Could quantum computers be revealing something about parallel universes? Google's quantum research leader Hartmut Neven raised this possibility, connecting the achievement to the many-worlds interpretation of quantum mechanics. However, this remains a philosophical interpretation, not scientific proof that parallel universes exist. One thing is clear: quantum computing is pushing the boundaries of what we thought computers could do. Sources:
Google Quantum AI. (2024). Meet Willow, our state-of-the-art quantum chip. Acharya, R., et al. (2025). Quantum error correction below the surface code threshold. Nature, 638, 920–926.
11
19
4
77
2,133