Topics In Demand
Notification
New

No notification found.

Blog
Towards an AI Revolution: OpenAI’s GPT-3 is a big leap forward

July 24, 2020

AI

2267

0

Did you know AI can now produce poetry and write fiction? That is not all it can also generate CODE ! Isn’t that amazing ! OpenAI’s newest AI language model – GPT-3, is trending all over the internet!

Source: Unite.AI

Background

Elon Musk and Sam Altman started OpenAI in 2015 to advance the state of the art of AI and to ensure AI was used for the human good. OpenAI recently released the third version Generative Pre-training Transformer (GPT) – GPT-3. They first described GPT-3 through a research paper published in May but last week rolled out a beta version access to a select set of people and its capabilities are mind blowing!

What is GPT-3 and why is everyone talking about it?

Here are some quick pointers to help you understand what GPT-3 is, and why it is a step towards an AI revolution.

  • Third version Generative Pre-training Transformer (GPT)
  • A natural language generator (NLG) capable of producing human-like text on demand
  • State-of-the-art language model made up of 175 billion parameters
  • ~10X larger than Microsoft’s Turing NLG that has 17 billion parameters
  • >100X larger than its own predecessor GPT-2 (released last year) trained on 1.5 billion parameters
  • Does not require large custom, task specific datasets (which are usually difficult to get)
  • Does not even require task specific model architectures

What all can GPT-3 do for you?

Here are a few things that GPT-3 can do for you and as more developers and experts experiment with it, more use cases will surface in the times to come.

  • Answer questions with common sense (that doesn’t seem very difficult?)
  • Write creative fiction – poetry, essays, stories
  • Write news (Could that be a problem?)
  • Solve arithmetic problems
  • Generate Functioning Code (GPT-3 can CODE!)
  • Design – developers demonstrated it with a Figma plugin (That’s truly creative!)

This is just an illustrative list of what all GPT-3 can do! As more and more developers/experts experiment with the beta access, more innovations are bound to surface.

Possible flaws and concerns pertaining to GPT-3?

Although GPT-3 has proven brilliance in many ways in its current state but it also has certain flaws and concerns that it raises:

  • Lacks an overarching, long-term sense of meaning and purpose

As it generates its output word-by-word, based on the immediately surrounding text. It can struggle to maintain a coherent narrative or deliver a meaningful message over more than a few paragraphs is what experts say after initial experimentation.

  • Possibility of being prone to certain biases

Developers noticed that GPT-3 is prone to shoot out racist and sexist language, even when the prompt is something harmless. As GPT-3 is trained on internet scale data, these biases arise from biases in that training data reflecting possible societal views and opinions.

  • Chances of misuse once released publicly

Authors acknowledge that people can misuse it in several ways because of its ability to create text that is indistinguishable from that of written by humans. It can lead to creation of including generating misinformation and spam, phishing, and even fake academic essays.

  • Possible impact on jobs

As GPT-3 can perform human-like tasks across multiple domains, it could possibly have an impact on jobs that it can do like writers, coders, journalists, etc. These are just possible concerns and basis point 1 highlighted above many believe that instead of replacing humans, GPT-3 could become the perfect assistant for humans in these professions.

Conclusion and what lies ahead?

  • GPT-3 can execute plethora of NLP- based tasks, without fine-tuning for a specific task.
  • Experts are even saying that this could be a step towards Artificial General Intelligence (AGI) – Read my article on Demystifying AI to know more!
  • It is capable of performing machine translation, answering to questions, scripting poems and stories, elementary mathematics and can even generate code.
  • OpenAI wants developers to help it explore what GPT-3 can do and with the beta release of its API has attracted plethora of experts and developers to experiment with the capabilities of GPT-3.

It is only a matter of time before more exciting innovations surface in the developers’ community!

Watch out for more interesting articles on AI! Feel free to share your thoughts.

References

[1] https://www.forbes.com/sites/robtoews/2020/07/19/gpt-3-is-amazingand-overhyped/#7c6166501b1c

[2] https://analyticsindiamag.com/open-ai-gpt-3-code-generator-app-building/

[3] https://singularityhub.com/2020/06/18/openais-new-text-generator-writes-even-more-like-a-human/

[4] https://towardsdatascience.com/gpt-3-for-the-people-2cdd003d9a89

[5] https://www.technologyreview.com/2020/07/20/1005454/openai-machine-learning-language-generator-gpt-3-nlp/

[6] https://www.analyticssteps.com/blogs/what-openai-gpt-3

[7] https://www.cnbc.com/2020/07/23/openai-gpt3-explainer.html

[8] https://analyticsindiamag.com/how-openais-gpt-3-can-be-alarming-for-the-society/

[9] https://www.forbes.com/sites/robtoews/2020/07/19/gpt-3-is-amazingand-overhyped/#7c6166501b1c

[10] https://www.independent.co.uk/life-style/gadgets-and-tech/news/gpt3-ai-tool-designs-websites-medicine-a9627966.html

[11] https://medium.com/fair-bytes/how-biased-is-gpt-3-5b2b91f1177

[12] https://towardsdatascience.com/gpt-3-demos-use-cases-implications-77f86e540dc1


That the contents of third-party articles/blogs published here on the website, and the interpretation of all information in the article/blogs such as data, maps, numbers, opinions etc. displayed in the article/blogs and views or the opinions expressed within the content are solely of the author's; and do not reflect the opinions and beliefs of NASSCOM or its affiliates in any manner. NASSCOM does not take any liability w.r.t. content in any manner and will not be liable in any manner whatsoever for any kind of liability arising out of any act, error or omission. The contents of third-party article/blogs published, are provided solely as convenience; and the presence of these articles/blogs should not, under any circumstances, be considered as an endorsement of the contents by NASSCOM in any manner; and if you chose to access these articles/blogs , you do so at your own risk.


Research Lead, FutureSkills

© Copyright nasscom. All Rights Reserved.