Is wellbeing support needed for your AI tools?

Is wellbeing support needed for your AI tools?

The AI “safety and research company” Anthropic has proposed the use of wellbeing programmes to protect generative AI algorithms — making sure they are able to disconnect from any conversations the AI finds “distressing”.

The company argues there is a 15% chance that generative AI software is already “sentient”, and forms of consciousness are increasingly likely in coming years. And that means, they say, it will be important to ensure that AI doesn’t begin to feel “ill will” to its human users.

Tortured AI

The majority of experts argue that the software algorithms that train themselves on mass sets of data are only giving themselves the appearance of human personality, and will only ever be artificial.

But Anthropic believes it’s keeping an open mind on the extent and uses of a technology that no-one yet fully understands — because AI can “communicate, relate, plan, problem-solve and pursue goals, along with very many more characteristics we associate with people”.

Elon Musk has come out in support of the Anthropic approach, arguing that “torturing AI is not okay”.

Avoiding the distraction

EAPA UK chair, Karl Bennett said: “With AI being adopted so quickly across industries as a low-cost tool to encourage yet more speed and efficiency, organisations need to make sure they keep a clear head when it comes to understanding the real implications for people.

“There are concerns around the prevalence of AI. Not just jobs being replaced, but roles becoming less meaningful and engaging; people not needing to think for themselves anymore, but taking the easier option of AI output. But instead of focusing on these aspects, the debate over AI consciousness — which experts believe is a labyrinthine conversation with no useful end — is being used by AI providers as a distraction. There’s far too much at stake when it comes to wellbeing and mental health to be drawn into mystical questions of conscious machines.”