Illustration by Tag Hartman-Simkins/Futurism. Source: Getty Images
AI models are everywhere now hospital To churches.
The surprising thing is that even AI experts still do not know exactly what is happening inside them. black box models, even if they are being deployed in the highest-stakes settings imaginable. The latest strategy to find out: studying them like biological systems.
For example, mit tech review reportsScientists at Anthropic have developed tools that let them explore what’s happening inside models while they perform a task, a type of study called mechanistic interpretation — which looks like How do doctors use MRI? To study brain activity, another type of intelligence that we do not yet understand properly.
“This is very much a biological type of analysis,” said Josh Batson, a research scientist at Anthropic. technical review. “It’s not like math or physics.”
In another experiment that is similar to that of biologists organoidsWhich are miniature versions of human organs, the journal reports that Anthropic has developed a special neural network called a sparse autoencoder, whose inner workings are easier to understand and analyze than regular large language models (LLMs).
Another technique is chain-of-thought monitoring, in which models explain their reasoning behind their behavior and actions – much like listening to the internal monologue of a real person. This has helped scientists detect wrong behavior.
“It’s been quite successful in terms of being able to find models that are actually doing poorly,” said Bowen Baker, a research scientist at OpenAI. MIT.
An imminent danger is that future models will become so complex – especially if they are themselves designed by AI – that we will effectively have no idea how they work. Even now, with the current tools and technologies we have, unexpected behaviors still emerge that do not align with human objectives of truth and security.
We see strong evidence of this in the news, which is littered with reports of people harming themselves because AI told them to do so – which makes it all the more troubling that we still don’t understand how they work.
More on AI: Indie developer removing entire game from Steam due to embarrassment over using AI
