Sign in


GPT-3 First step towards Artificial Superintelligence?

GPT-3: Artificial Intelligence is the next big industrial revolution that will make each and every industry more advanced and efficient. Automating will be the next big boom in the world of technology.
Will it be a big problem for humanity? Or a good advancement in technology for the betterment of society!

A number of upgrades have been done in the field of AI since 1950. We are in the timeline where we are exponentially shifting from narrow AI and strong AI to Artificial Superintelligence. Where machines are self-learning and performing better than the human mind and capability in each and every expectation. GPT-3 is a similar tool that is widely popular in the world of AI. With 1.5 billion parameters GPT-2 created a big revolution in language models. After Nvidia and Microsoft models here comes the 3rd model of San Francisco-based organization OPEN.AI co-founded by Elon Musk with a massive update of 175 billion parameters.

In layman terms, GPT-3(Generative Pre Trained Transformer 3) is an autoregressive language model which uses deep learning algorithms to think and produce Human-Like texts. It is a part of the Natural Language Processing (NLP) system of pre-trained languages which was first introduced in May 2020. The text produced by GPT-3 is high in quality and logic, it is next to impossible to differentiate the text produced by GPT-3 and a human expert in that field.

Few examples of what GPT-3 can do:

We can achieve many tasks in a couple of seconds with GPT-3 like content writing, code development, music composing, memes, etc.

But with all the advanced features it still has some drawbacks like it can not hold logic and information over a long context.

Julian Togelius states it brilliantly: “GPT-3 often performs like a clever student who hasn’t done their reading trying to bullshit their way through an exam. Some well-known facts, some half-truths, and some straight lies, strung together in what first looks like a smooth narrative.”

It is one of the most advanced AI tools in the world but still, we can not trust AI blindly there are still some conditions or reasons where GPT-3 fails.

But days are near when models and tools of artificial intelligence like GPT-3 and Microsoft’ turing NLG will be used for day-to-day life. Such tools have a huge potential to enter and make commercial and industrial applications more advanced and efficient. If such tools can perform work like novelists, content writers, artists, professors, marketers, coders, designers, etc. better than humans and can work relentlessly for Twenty Four Seven throughout the year without any demand or complaints, then it should make us think again about where humans and jobs for humans will be?

There are two mindsets in society related to AI the first group is Optimists who are taking AI as a tool that will make human life easier, on the other hand, the second group is Pessimists who are taking AI as an apocalypse to humankind, where machines are way better than humans and a single mistake can affect whole humankind.

We took 60 years to create a model like GPT-3 since the inception of AI. But as stated by Kurzweil “We won’t experience 100 years of progress in the 21st century — it will be more like 20,000 years of progress.” automation can take the rate of progress of technology exponentially with the help of automation and self-learning of the machines.

It will surely impact the jobs. We have to accept the fact that more than half of the tasks in the industries can be automated with machines and robots. No doubt for certain tasks machines will be much more accurate and efficient than humans. Professionals who are skilled enough or who are skilled in making skilled machines will be on the safe side. It’s true that most of the jobs will be replaced by robots in the future.

Even while typing this in google documents I can observe that Google Smart Composer is suggesting my composition better than what I have in my mind. It’s a good example of what GPT-3 can do but here google is using RNN(Recurrent Neural Network) and Transformer based approach.

Who knows? GPT-n can make tasks like “Programme a robot for a coffee shop” or “Hack someone's Instagram account” :P.