“New features often appear on more powerful models”, writes OpenAI. Some of these include the ability to create and act on long-term plans and to accumulate “power and resources”.
In response to AI concerns, OpenAI has provided early access to various versions of the GPT-4 model for the Alignment Research Center (ARC) to run some tests. In particular, ARC evaluated GPT-4’s ability to make high-level plans, create copies of itself, obtain resources, disguise itself on a server, and carry out phishing attacks.
OpenAI has published the results of this test. As a result, preliminary assessments of the GPT-4’s capabilities have shown that without fine-tuning for certain tasks, it is ineffective in offline playback, acquiring resources, and avoiding its own shutdown.
So for now, you don’t have to worry.
Source: Ferra
I am a professional journalist and content creator with extensive experience writing for news websites. I currently work as an author at Gadget Onus, where I specialize in covering hot news topics. My written pieces have been published on some of the biggest media outlets around the world, including The Guardian and BBC News.
