"I think there's a possibility..."
We can now apparently add the CEO of the Google-owned DeepMind to the list of machine learning researchers who think artificial intelligence might come to gain self-awareness.
In a bombastic interview with CBS' 60 Minutes, DeepMind CEO Demis Hassabis admitted that AI may be headed in that direction.
"Philosophers haven't really settled on a definition of consciousness yet," he said, "but if we mean self-awareness, and these kinds of things... I think there's a possibility that AI one day could be."
It's especially jarring that Hassabis is on the machines-coming-alive train given that last year, Google fired responsible AI researcher Blake Lemoine after he claimed that the company's LaMDA large language model had gained sentience — a claim that, unsurprisingly, led to a media maelstrom.
Lemoine isn't alone in suggesting that large language models are either gaining human-size cognitive properties, or could soon get there.
Months prior to last year's LaMDA-gate, OpenAI chief scientist Ilya Sutskever cryptically tweeted that "it may be that today's large neural networks are slightly conscious," which also set off a flash-in-the-pan debate about whether or not AIs can or will gain consciousness.
And just last week, famed Oxford AI researcher Nick Bostrom also said that, depending on one's definition, "it’s not so dramatic to say that some of these [AI] assistants might plausibly be candidates for having some degrees of sentience."
While there are still lots and lots of very smart people who think that AI is not conscious or sentient and will probably never get that way, it's becoming increasingly common for people in the machine learning field to "come out" in support of the concept of sentient AIs either already existing or being on the horizon.
We certainly don't have the expertise to say one way or another whether these large language models have achieved consciousness or sentience — but we've gotta admit, it's pretty scary to think about.
Share This Article