OpenAI is the Elon Musk-founded artificial intelligence research lab. So, last month, it announced the arrival of the newest version of an artificial intelligence system it had been working on that could mimic human language, a model called GPT-3.
Furthermore, people got the chance to play with the program, in the weeks that followed. The hype is real. It has its shortcomings. Nevertheless, do not make a mistake: GPT-3 is representing a tremendous leap for artificial intelligence.
I, a year ago, sat down to play with precursor dubbed of GPT-3. It was GPT-2, of course. It was reasonably good at that time. When given a prompt (sentence or frame), GPT-2 could write a decent news article, making up imaginary organizations and sources and referencing them across a couple of paragraphs. By no means was it intelligent. It did not understand the world. Nevertheless, it was still an uncanny glimpse of what it may be like to interact with a computer.
A year later, GPT-3 came out. It is a lot smarter. OpenAI took the same basic approach it had taken for GP2-. Moreover, it spent more time training it with a bigger dataset. Thus, the result is a program that is much better at passing various tests of language ability that machine learning researchers developed to compare out computer programs.
Nevertheless, that description understates what GPT-3 does and is.
Arram Sabeti is an inventor with early access to GPT-3. He has published hundreds of examples of results from the program. Sabeti said that GPT-3 surprises him continuously. He said that a turn of phrase, a witty analogy – the repeated experience he has is There’s no way it just wrote that’. It is exhibiting things that feel very much like general intelligence.
Nevertheless, not everyone agrees with that. Gwern Branwen is a researcher. He wrote in his article concerning GPT-3 that Artificial Intelligence programs lack self-awareness and consciousness. Moreover, he added that they would never be able to have a sense of humor. They will never be able to appreciate love, or art, or beauty.
Moreover, they will never feel lonely. They will never have empathy for the environment, for other people, and animals. They will never cry at the drop of a hat, will never fall in love, or enjoy music.
It is not Brawnen’s writing. The author of those words is GPT-3. Brawnen just expressed a few words of skepticism about artificial intelligence. GPT-3 came up with a convincing and long rant concerning how computers will not ever be really intelligent.
Brawnen said that he was taken aback by GPT-3’s capabilities. They get steadily better at predicting the next world, as GPT-style programs scale. Nevertheless, Brawnen noted to the point that improved prediction makes a little more accurate mimic: a bit better at trivia questions, and a bit better at English grammar. GPT-3 suggests to suggested to Brawnen that past a certain point, the starts coming from reasoning and logic and what looks entirely too much like thinking. In some ways, GPT-3 is an elementary program.