What I found to be incredible was in 2020, 5 years prior to AI passing the Turing test, there was a test performed at MIT. I read about this experiment in a very good book called Scary Smart by Mo Gawdat. Among his many accomplishments, he was the Chief Business Officer at Google. One of the experiments he wrote about in that book occurred at MIT back in 2020 which lead to a new antibiotic called Halicin.
The amazing part of this experiment was how AI was used, resulting in the scientists there to admit it would have taken them decades, if ever, to discover Halicin. The AI found it within hours.
General Overview
The inferences the AI was able to make were astounding, not something a human would normally take. We all have our own knowledge, thoughts and experiences. In general, there is learning by association and learning by inference.
Learning by association is distinct events and behaviors that our brain links together. After the event and behavior occurs many times, this link becomes stronger, shaping how we view future events.
Leaning by inference is basically a logical conclusion. A judgement call, educated guess based on the evidence available and prior knowledge.
When the AI in the MIT experiment made its own inferences, it was far different than humans because the human brain couldn’t make the jump given it’s current knowledge and past experiences. It assumed a given approach wouldn’t work based on this learnt association. The AI didn’t make such an association and instead proceeded purely on the information or evidence provided.
Final Thoughts
I found that to be fascinating but upon further review, I also found it, well, obvious. The AI addressed the problem from a pure logic point of view. That’s what a computer would do, right? Where-as the scientists followed their logical approach as well. However, given the nature of humans and how we learn, the resulting find the AI made of Halicin would not have been discovered (so quickly) because, the path the AI took would have been deemed, illogical. Sounds a bit confusing, but it shows how our brains makes certain associations that could or does actually prohibit us.
To close, we initially talked about the Turing test, then we make a jump to this MIT experiment. WOW, the rate of growth I find to be incredible and it’s only going to improve.
