[Article Review] The Next Word - Where will predictive text take us by John Seabrook

Summary

The author¹ starts by mentioning his experience with GMail’s Smart Compose feature and says that it is heartening to see the machine completing the sentence for him. He further talks about OpenAI’s GPT-2 ma- chine, which is an AI Writer, a supercharged version of Smart Compose. The arti- clequotesDarioAmodei,DirectorResearchatOpenAI,

  Until now, if you saw a piece of writing, it was like a certificate that a human  was involved in it. Now it is no longer a certificate that an actual human is involved.

Seabrook then ventures into language evolution and talks about neuroscience and psychology of spoken language. He says, that humans are evolved in terms of spoken language but writing is an acquired and learned trait. Hence, it is possible for an AI agent to write much better than speak. The author shares his fair that machine will soon be taking over humans as far as writing is concerned.

He mentions that there are two ways to make a machine intelligent - Knowlege Based, where the machine is imparted knowledge and the machine that can learn itself. He says comparitively humans integrate both types of intelligence so seamlessly that we hardly distinguish between them. The machine will require lot of compute power, which he says is now easily available as compared to few years ago.

He ends the article by optimism thinking of the iteration of GPT far more powerful could be hybridized with a procedural system, so that it would be able to write causally and distinguish truth from fiction and at the same time draw from its well of deep learning.

Evaluation

The article portrays the AI in a positive light and accurately describes how an AI agent can be created by talking about Knowledge Based AIs and Self- learning AIs. The author fairly covers the details of an intelligent machine and it’s development. It is a good research based mass-market article, which is expected from The New Yorker.

References