Human-aware AI systems are AI models that use the capabilities of both technological intelligence, including existing AI models and AI languages, and human intelligence to achieve a better performance than neither of the two could achieve on their own.
In other words, human-aware AI is a sort of AI that is augmented by human intelligence. Some call it human-aware technology, or human-aware augmented intelligence. And it seems, the human-aware AI models could hold the key to incredible scientific breakthroughs, and in time, they could even reach AGI (artificial general intelligence) and ASI (artificial superintelligence), according to the latest research studies.
The two scientists behind the study, Jamshid Sourati, and James Evans, agree that these human-aware AI models are capable of accelerating science with the way they approach the research methodologies. They trained unsupervised AI models on scientific research cognitively accessible only to experts, while also making them aware of human expertise and the results are incredibly promising: AI prediction of future discoveries beyond those already learned improved by 400%.
In other words, human-aware AI models are capable of coming up with their own correct prediction of a future scientific breakthrough. These models succeed by predicting human predictions and the scientists who will make these predictions.
Even more, human-aware AI models are also capable of distancing themselves from existing theories and research methodologies, and they are able to come up with their own ‘alien’ approaches to research, that are unlikely to be pursued by humans now. These approaches generate scientifically promising hypotheses, which can be further explored in research methodologies yet to be discovered or tackled by the human mind. Human-aware AI models accelerate science by tackling blind spots and challenging the current scientific limitations.
This could further lead to the acceleration of AI progress. We’re talking about AGI, and the ASI, which big tech giants, including Microsoft, OpenAI, Meta, and Amazon are currently focusing on. By augmenting human intelligence with artificial intelligence, or vice versa, AGI could be easily reached within this decade. Then it will only be a matter of time until ASI is reached. According to the newest OpenAI project, the Superaligment Project, artificial superintelligence is just within our reach. A human-aware AI model, though, would be the best type of ASI.
source
