One of the most prominent events in the world of artificial intelligence in recent years is the GPT-3 artificial intelligence technology developed by OpenAI. This technology is a service created by OpenAI, founded by Elon Musk and Sam Altman, to produce artificial intelligence technology at human level (capable of performing any written task in English) with an investment of more than $ 1 billion in 2015. This artificial intelligence can answer questions, produce articles and poems, even code! He gets information from you to teach if anything he can't do. Access is currently limited, you are expected to fill out a form to test the beta version.
Its capacity is said to be more than ten times that of Microsoft's Turing NLG (Wikipedia). The source of 60% of the data set given to GPT-3 to learn is the filtered version of Common Crawl consisting of 410 billion data. 22% of the data set is from the data on the Internet, 16% from the books published so far, and the remaining 3% of Wikipedia, which is seen as one of the most comprehensive sources of information today, has been used! Since its first version, the GPT, the GPT-3, which has been continuously improving, has 175 billion parameters. The training of this number of parameters is said to cost OpenAI a total of 12 million dollars, considering the news on the internet.
Analyzing the data using Deep Learning and different Machine Learning algorithms, the system succeeds in exhibiting "humanoid" approaches in direct proportion to the size of the data it feeds and the size of the system it uses.
Here are examples of what GTP-3 can do:
• Produces aphorisms
• Able to produce content similar to the authors' works
• Can interview
• Can design
It is said to be at a surprisingly realistic level.
There are millions of people waiting in line to test the GTP-3 artificial intelligence API, which is currently used as a beta version and can be experienced by a limited number of people. As a result, organizations will be better understood when they start using it and for what purpose they will use it, and in this way, the scaling, cost, performance and return of the API will be revealed more clearly. API functionality should be feature-rich enough for organizations to avoid a scaling-blocked situation.
We guess that for English language chatbots, this will act as a very suitable API to improve the conversation experience. How can we take advantage of this great technology? Of course by building a chat assistant! GPT-3 seems to open up a new lane in chatbot development.
Engin Karabudak/ Software Developer