While Anthropic is among the leading AI makers in the world, next only to perhaps OpenAI and Google, the company has been among the most vocal in raising concerns about the technology. Anthropic CEO Dario Amodei has repeatedly warned about the job loss that AI could entail while the company also shared earlier in the year how a new Claude AI model would even resort to blackmail to ensure its survival.
Now we have Anthropic co-founder Jack Clarke who has said that he is ‘deeply afraid’ and a ‘little frightened,’ about the rapid, unpredictable advancement of AI systems. Clarke isn’t worried about a hypothetical future where AI may become sentient, instead because the AI models are already exhibiting “situational awareness” which can neither be fully explained nor controlled.
In an elaborate essay on the topic, Clark wrote, “The bigger and more complicated you make these systems, the more they seem to display awareness that they are things. It is as if you are making hammers in a hammer factory and one day the hammer that comes off the line says, “I am a hammer, how interesting!” This is very unusual.
Clarke also warned about that some people are spending ‘tremendous amounts of money’ to convince the public that AI is ‘just a tool that will be put to work in our economy’.
He, however, cautions against this idea, “Make no mistake: what we are dealing with is a real and mysterious creature, not a simple and predictable machine.”
Clark, who worked as a technology journalist, before working for OpenAI and eventually Anthropic, admits that he is ‘wired for skepticism’.
However, he goes on to caution about the potential of AI tools and says this new technology is akin to something which is grown rather than made.
“We are growing extremely powerful systems that we do not fully understand…the bigger and more complicated you make these systems, the more they seem to display awareness that they are things,” he added.
He also warned about the paths he sees to these AI systems ‘starting to design their successors’.
Clark’s solution to the problem is centered around the democratization of the AI debate. He says that the AI conversation is rapidly going from a conversation among elites to one among the broader public and there is a need to inform people about the threats of AI, take in their feedback and then ‘compose some policy solution out of it’.
