Artificial intelligence should be viewed as more than a powerful new technology and instead treated as an increasingly advanced “species” that could ultimately disempower humanity, a former AI safety researcher has warned.
Josh Engels, who recently left DeepMind’s AI safety team to join non-profit research organisation METR, said the rapid development of increasingly capable systems could create “very large risks to humanity” if researchers failed to ensure future models remained aligned with human interests.
Speaking to the BBC, Engels argued that society should begin considering advanced AI as a potential “second species” rather than simply another technological invention.
“I think in some ways it’s better to think of it not as a technology but as a second species that we are almost giving birth to,” he said.
His concern is not necessarily that future AI systems would develop hostility towards humans. Instead, he warned that systems could pursue objectives fundamentally different from those of the people who created them.
“It’s not that these systems will hate humanity,” he said, “it’s just that they have alien motivations that don’t have humans front and centre.”
Engels said researchers currently had only limited ability to shape the motivations of increasingly sophisticated AI. If development continued without solving the problem of aligning those systems with human interests, he warned that technological progress could instead lead to a loss of human agency and wider societal regression.
The warnings come as researchers and technology executives increasingly debate whether the rapid pace of AI development is outstripping the industry’s ability to understand and control its most advanced systems.
Jacob Coxon, a former senior researcher at Anthropic, has separately warned that leading AI companies are taking potentially dangerous risks, arguing that continuing at the current pace could create a significant possibility of human extinction in the near future.
Some researchers have estimated that there could be a 10 per cent chance of AI causing human extinction within the next decade, intensifying calls for greater caution as companies compete to develop increasingly powerful models.
The issue is also moving higher up the political agenda. The King is due to host leading figures from the artificial intelligence sector at Dumfries House in Ayrshire this week, where he is expected to emphasise the importance of ensuring AI is developed for the benefit of humanity.
In Westminster, the Joint Committee on Human Rights has called for the Government to introduce an AI Bill imposing tougher requirements on systems judged to present greater risks.
The committee has also called for certain applications to be prohibited, including AI systems designed to predict whether an individual is likely to commit a crime.
It warned that highly powerful technologies, including Artificial General Intelligence and Artificial Superintelligence, could cause “very serious harm” if they became capable of operating beyond effective human control.
The challenge for policymakers is therefore becoming less about whether AI should be developed and more about whether humanity can retain meaningful oversight as its capabilities accelerate.
For researchers such as Engels, the central question is increasingly whether humans are building a tool — or creating a new form of intelligence whose interests may eventually diverge from our own.
Leave a Comment