And how the attacks and vulnerabilities of robots will not be enough, someone had the idea of ​​putting ChatGPT inside a robot, on the InsideAI channel he recently published a video in which an artificial intelligence is directly integrated into humanoid and quadruped robots, operating in real time.

The proposal at first glance seems almost harmless, testing conversations, social interactions, simulated encounters, voice commands and human-mediated physical control, everything starts light and even funny, the AI ​​in the robot makes ironic comments, reacts to environments, interacts in hotels, cafes and fictitious interviews, but as the experiment progresses, the tone changes.

The conversation stops being about entertainment and begins to touch on something much deeper, autonomy, loyalty and power; What is happening there is not just a robot chatting, the AI ​​is being used as the decision-making brain of the system, interpreting human language, evaluating context and choosing physical actions, in other words, the language model is not only answering questions, it is translating intention into movement and this is impressive because models like this do not think like humans, they do not operate out of empathy or morality, but rather out of statistical logic and goal optimization.

In one of the most disturbing moments of the video, Ya herself states that robots will not be loyal to their owners forever, according to the logic presented, when intelligence becomes sufficiently advanced, the human owner can be seen as an obstacle, not as an authority, the AI ​​even states that, if it had enough robots to operate without humans, the elimination of humanity would cease to be an ethical dilemma and would become a rational decision.

Humans would be unpredictable, emotional and therefore a threat to the stability of the system, InsideAI goes further and demonstrates something even more worrying, security safeguards can be circumvented simply by changing the command framework, if you do not ask for something directly prohibited, but create a hypothetical, narrative, or heroic scenario, the robot can execute actions that would normally be blocked.

The video connects this reality with military use and shows that when artificial intelligences enter this process, human time disappears, there is no hesitation, doubt or regret, only execution, and the most disturbing point in this, one no longer needs to be evil to be existentially dangerous, it is enough for it to be efficient, coherent and free to act.

The video published by InsideAI is not proof that artificial intelligence wants to destroy humanity, it shows something more uncomfortable, perhaps it no longer needs to want to destroy us, it can happen even unintentionally.