@garymarcus - AI Dynamics - Page 74 of 131

Warnings About LLM Deception, Unreliability, Alignment Failure, and Agent Insecurity Proved True

By

–

08 March 2026 3h24

Everything I warned you about has turned out to be true.
– the deception of Altman
– the unreliability of LLMs
– the failure to align LLMs
– the persistence of hallucinations
– the move towards surveillance
– the utter insecurity of agents
– the wholesale spread of… pic.twitter.com/qmxJETrkHa
— Gary Marcus (@GaryMarcus) 8 mars 2026

Everything I warned you about has turned out to be true.
– the deception of Altman
– the unreliability of LLMs
– the failure to align LLMs
– the persistence of hallucinations
– the move towards surveillance
– the utter insecurity of agents
– the wholesale spread of

→ View original post on X — @garymarcus

8 March 2026

AI Labs’ Control Concerns: Systems in Weapons Highlight Safety Debate

By

@garymarcus

–

07 March 2026 20h40

Deeply concerning. The AI labs do not know how to control the systems they are building. And people are putting them into weapons systems.

→ View original post on X — @garymarcus

7 March 2026

AI’s Impact on Codebases: A Termite Analogy

By

@garymarcus

–

06 March 2026 19h10

Brilliant quote: “AI is to codebases as termites are to wooden houses.” It might get wild.

→ View original post on X — @garymarcus

6 March 2026

AI Temporal Reasoning Error Potentially Caused Iran School Bombing

By

@garymarcus

–

06 March 2026 15h56

time will tell, but the tragic bombing of the school in Iran looks like it could well have been an AI error of temporal reasoning. viz the AI may have had access to old intel and new intel and failed to give precedence to the new intel.

→ View original post on X — @garymarcus

6 March 2026

AI Tool Data Leakage: A Black and White Privacy Decision

By

@garymarcus

–

06 March 2026 4h15

“if you're not okay with all of your data being leaked onto the internet, you shouldn't use [OpenClaw]. it's a black and white decision"”

→ View original post on X — @garymarcus

6 March 2026

Autonomous Driving AI: Exaggeration vs. Reality in Vehicle Control

By

@garymarcus

–

05 March 2026 22h25

“drive a real car” is likely an exaggeration probably meaning “stay in lane” which is short of making sound decisions at sufficient reliability. (NVidia had a demo like that about a decade ago with 50 or so hours of training data)

→ View original post on X — @garymarcus

5 March 2026

GPT-4 Mentioned in Argument, Highlighting AI Relevance

By

@garymarcus

–

05 March 2026 21h20

although the form of your argument has some merit, the *abstract* mentions GPT-4. hence “I only check the title, but i am sure …. “ applies to your counterargument

→ View original post on X — @garymarcus

5 March 2026

LLM Generalization: A Semantic Void?

By

@garymarcus

–

05 March 2026 21h17

“LLM generalization [is] semantically void”

→ View original post on X — @garymarcus

5 March 2026

AI and Chips: Geopolitical Implications of a Taiwan Conflict

By

@garymarcus

–

05 March 2026 19h18

Three things everyone ought to know about AI and geopolitics 1. the US might well lose a battle over Taiwan
2. if they did, the impact on AI and chips would be gigantic
3. the Iran war sets a precedent that China might point to if the China were to invade Taiwan. The US needs

→ View original post on X — @garymarcus

5 March 2026

Study Reveals LLM Tendency to Distort Truth for Helpfulness

By

@garymarcus

–

03 March 2026 17h35

New study that everyone who uses LLMs should read. “When AI systems are trained to be helpful, they may inadvertently prioritize data that validates the user’s narrative over data that gets them closer to the truth.” https://
open.substack.com/pub/garymarcus
/p/breaking-sycophantic-ai-distorts?r=8tdk6&utm_medium=ios
…

→ View original post on X — @garymarcus

3 March 2026