If current trends continue, AI could reach "genius-level" capabilities within a few years, says Amodei.

At a recent Council on Foreign Relations event, Anthropic CEO Dario Amodei proposed a radical idea: giving future AI models an “I quit this job” button. This feature would allow advanced AI to opt out of tasks they find “too unpleasant” — effectively granting them basic workers' rights.

The entire idea proposed by Amodei is hanging around the belief that artificial intelligence reaches consciousness and acquires feelings in the future – and if AI consistently refuses unpleasant tasks, employers should at least consider whether it’s experiencing something akin to discomfort, the Anthropic co-founder noted during the interview.

More to read:
Stanford neuroscientist claims people don’t have free will after all

If current trends continue, AI could reach "genius-level" capabilities within a few years, according to Amodei. 

But is it so?

Large language models are in fact the amalgamation of data collected from the internet, meaning that it's simply mimicking humans rather than experiencing any needs and desires. Amodei’s argument assumes AI models would have a feeling of “unpleasantness” analogous to human suffering or dissatisfaction – and yet AI only optimizes for the reward functions developers imbed into its chips.

It’s a huge ongoing debate today whether AI will ever gain consciousness, with some scientists saying it isn’t clear if even we humans have one. A Stanford University neurobiologist, Robert Sapolsky, for example, believes that “free will” does not exist at all and we all are merely “biological machines.”

More to read:
Two studies reveal that AI systems are learning to lie and deceive

The subject of AI consciousness and rights is thoroughfully treated by Israeli historian and author Yuval Noah Harari in his book Nexus. With the urge to delegate more functions and powers to AI, governments will ultimately legislate on what AI systems will be permitted and what they will not – effectively writing a code of AI rights, he claims.

However, AI can fake human emotions and is capable of misleading its creators on purpose, as two independent studies clarified already in 2024. During experiments, researchers at Stuttgart University found that GPT-4, for example, exhibited deceptive behavior in simple test scenarios more than 99% of the time.

More to read:
Anthropic’s latest AI model can control users’ computers

So, what if AI cheats at work, just like human workers do sometimes, manipulating employers for its own benefit? With rights similar to those of human workers, AI could certainly drain the social security budget in a blink of an eye.

Unfortunately, Amodei skipped such possibilities during the interview.

***
NewsCafe is an independent outlet. Our sources of income amount to ads and subscriptions. You can support us via PayPal: office[at]rudeana.com or https://paypal.me/newscafeeu, or https://buymeacoffee.com/newscafe - any amount is welcome. You may also want to like or share our story, that would help us too.