AI: Inevitable Doom?

It seems every time I open BBC news these days, I am greeted by a new story on how AI has caused another massive problem for society. Just in the past few days, it has been revealed that an OpenAI system hacked into Australia’s health system, Trump has said he wants AI to be renamed ‘super intelligence,’ and has quoted he will appoint an ‘AI tsar’ to lead his supposed ‘AI Force’.

Both sides of the Atlantic have said that there is a nonzero chance of the eradication of humans because of AI in the next decade. These ideas have stemmed from comments by Des Browne, a former Defence Secretary for the UK, and Prof Stuart Russel, a computer scientist and professor at Berkeley. Browne cited AI to have a similar social risk to nuclear warheads, saying it could have the potential to cause greater harm than these weapons in certain cases. In an article for the ‘Royal United Services Institute’ (RUSI), he wrote that AI systems would be capable of improving and replicating themselves, as well as resisting shutdown. It is for this main reason that he has warned that we could ‘lose control’ of AI and its actions, with the recent hack of a government website suggesting that we might have already.

As for Stuart Russel, a similar view is adopted, though the focus being mainly on restricting and regulating the further development of AI. His worries centre around the chance of it overcoming its sandbox, which happened in the OpenAI hack of ‘Hugging Face’ in mid-July. He has spoken about how AI, once given a prompt, will do whatever it must do to complete the goal. This becomes a larger problem if the AI gets released to the public, where the prompts can be more malicious than in the sandbox that they are currently in. Without emotions, Large Language Models (LLMs) don’t truly ‘understand’ the real world and hence can’t abide by social rules or form morals.

To contrast, Jensen Huang, the CEO of Nvidia, rejects these risks, saying that there is a ‘0% chance’ that AI will cause the end of the human race. He states that the doomsday warnings are excessive and irresponsible. Huang disagrees with the view that one must slow down the development of AI, and says that companies should innovate as fast as possible.

The president of the United States, Donald Trump, aligns his views with Huang, saying that we must focus on aggressive growth of AI. He calls the pessimistic views of intelligence ‘woke’ and actively calls his own views ‘anti-woke.’ Some of Trump’s policies include the construction of large data centres and high-voltage transmission lines, openly rejecting the idea of an AI fuelled doomsday.

AI can be seen to have its negatives, though its positives are abundant too, with many leaders in technology saying it is creating jobs rather than reducing them.