GPT3 by Open AI, is a state of the art language model. There is a lot of hype around it and is only growing as recent applications using GPT3 have started to surface. Since GPT3 can do tasks without any training (meta learning) from a single example, it may take away some of the human jobs in the near future.

What’s Open AI?

It’s a research laboratory of the corporation Open AI LP which is a for-profit corporation and Open AI Inc which is a non-profit. It has big name investors such as Elon Musk, Sam Altman and Microsoft. Its main objective is to direct the development of AI for the general good. Musk himself considers AI as humanity’s biggest existential threat. If used wrongly, that is.

How GPT3 Works

GPT3 is a text generating neural network released in June 2020. With 175 billion parameters, it surpasses its competitor Microsoft Turing NLG which uses only 17 bil parameters. These parameters are network calculations which are applied with particular weights to different aspects of the data. Since it has trained on a huge amount of data, it is even capable of doing tasks without any training. All it needs is a single example.

Once a request is made, GPT3 analyses it and provides the answer which it deems to be the most probable. The surprising fact is that the text predictor can be said to process all of the text existing on the Internet for providing the best match. This has only been possible in recent times due to the rapid development of AI and decrease in cost of computer hardware with relation to the processing power.

Developers have built an impressively diverse range of applications using the GPT-3 API, including an all purpose Excel function, a recipe generator, a layout generator (translates natural language to JSX), a search engine and several others. Here are some examples.

What GPT3 is capable of

Since it’s a relatively recent product, it has been used with only a limited scope. With a single example GPT3 can write stories or books, articles for blogs. Copilot is a recent example of its use case where, with co-operation with GitHub, Open AI has created a tool able to assist coders. It can autocomplete a code that it deems to be the most probable.

With this, code with boilerplate won’t require to be typed out as it can be autocompleted on languages such as Python, Typescript, JavaScript, Ruby and Go. The output produced is very similar to that of a human, mostly indistinguishable if seen without a context.

Which Jobs can it take?

With provided examples, it is clear that IT and software development jobs might be at risk. If the technology continues growing, the low skilled developer jobs will be replaced by AI systems. This will increase the skill requirement for such jobs, not replace it completely (we hope). High level programming languages might be automated and the programmers might not need to pay attention to every minute details as the assistant will take over.

Journalists and writers might need to have a look at the example texts available throughout. As the outputs get more intuitive and human-like, more jobs in the writing field might be replaced by AI. Unlike humans who base their writing on experience, taste or the articles they have read, AI systems can scan through millions of blogs, articles, books and provide a conclusive summary. It clearly has more resources to pull from. Summaries can be created in an instant, whole plays can be created with a small provided context on how the book should be.

Conclusion

Machines drove farmers to produce larger quantities of food of better quality. Artists will find inspiration to create better content by taking references. So it might drive us forward by eliminating some of the less skilled tasks taking up our precious time and effort. We’ll be more productive as a result in this age of nano-technology, cloud computing, Internet of Things and Artifical Intelligence.

Sources:

https://arxiv.org/abs/2005.14165

https://en.wikipedia.org/wiki/GPT-3

https://www.skynettoday.com/briefs/gpt3

Image Credit: possessedphotography

Author

Write A Comment