The Uniqueness of Humans — Cont.
Part 2: Man vs. Machine
Since the middle of the 20th century a different but related question has become more prominent: How is human intelligence different then that of computers? The pioneers of artificial intelligence (AI) in the 1960s hoped to develop computers that will be able to perform “intelligent tasks” as well as or better than humans, and along the way will give us insight into our own intelligence and perhaps even our destiny. The 1st goal was perhaps achieved. The 2nd probably not.
As a teenager I spent many hours playing and studying chess, and became a chess master and teacher. As a CS major at the university I wrote a very simple chess program. In the following years chess programs earned progressively higher rankings, until about 20 years ago IBM’s Deep Blue was able to beat the best human chess player, Garry Kasparov. Disappointingly however, these programs employ “brute force” algorithms and use massive computational power in a way that has very little in common with human thought processes.
As for using tools, autonomous robots nowadays clean our homes and offices, handle merchandise in warehouses, operate machines in manufacturing facilities, and not only do they make our cars but now they can also drive them.
What about language? The advancements in NLP (natural language processing) in the last few years have been staggering. The GPT-3 language model announced recently by OpenAI (if you are a ML aficionado — read the paper here) uses a monstrously huge neural network with 175 billion parameters, and training it costs about $250,000. The model has achieved state-of-the-art results on a variety of language related tasks and benchmarks, including translation, question-answering and filling in the missing word in a sentence (the Cloze task). GPT-3 can also generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans.
Similarly, an open source chat bot developed by Facebook a few months ago can carry out a conversation on any topic that feels, well, human. Thus it would be hard to deny that computers can now pass the 1950’s Turing Test.
I will leave aside the philosophical debate on whether such impressive achievements actually prove that machines are, or can be intelligent (i.e. “think”). My point is just that it does not appear that humans poses any skills that other animals, or computers, do not and cannot have.
And thus we are left with a nagging question: is there a truly unique human potential, and how can we explore and develop it?
Acknowledgments
Many ideas and references in this article came from my favorite data scientist, my son Oren Dar. :-)
