A place to discuss AI and its implications for people
Are you sure you want to leave ?
Stop Press: AI researchers install a Right Hemisphere
Stop Press: AI researchers install a Right HemispherePosted by Paul on April 6, 2023 at 11:27 am
I have been wondering from the start of the AI boom we’re seeing how long it would be before we saw a “Hemispheric” division within AI to counter its tendency to make mistakes. This work goes back quite a few years and I am currently looking for older research papers in AI to see how they formulate the problem and what other papers and works they reference along the way.
However, this week researchers Shinn, Labbash and Gopinath have released their “Reflections” paper detailing their exploration of installing a separate layer in the AI to improve its answers.
I would argue that AI researchers have basically implemented Iain’s finding that without an overarching system that checks, adjudicates and curates answers the mind (in this case an artificial one) will produce errors it is not capable of being aware of let alone correcting. The authors call the respective systems the “Researcher” (the system that collects the data and tries to formulate an answer) and the “Decider” (the system that evaluates the answer for ‘fitness’ and suggests improvements.
In this sense then, it is the pragmatics of the situation (the need for AI to be accurate and autodidactic) that have led them to trying this.
Here is the link to the current paper (and no, Iain does not appear in the references):
And here is a video explaining what this is all about and what it means:
Log in to reply.