My first few weeks into the worlds most powerful AI language generating model — OpenAI’s GPT-3

I had been waiting a long long time to get API access to GPT-3. Finally, I did get access, and oh-boy, the last few weeks have been a real revelation to me to see the supreme power of the latest and greatest neural network for unrestricted natural language generation.

When I got beta-access, my initial questions were — how human is GPT-3? How close is it to a human like a conversation? Is this a step towards general artificial intelligence that would allow a machine to reason broadly like humans without having to train for every task it encounters?

Basic Question — what is GPT-3?

Answer — GPT-3 is a model which is trained to auto-complete sentences, trained on a huge amount of uncategorized data from the internet. It can also complete coding lines in programming languages.

And yes, it had also learned a lot of interesting stuff too.

GPT-3 was developed by Elon Musk-owned OpenAI and is currently the largest artificial language model.

Question — How does GPT-3 work?

Answer — GPT-3 is based on neural network architecture as part of machine learning. Machine learning works by memorizing patterns.

Historically, each model has been able to learn one set of patterns. For instance, we can learn a model to tell us whether a FB post is positive or negative. We do this by showing the model examples of positive and negative FB posts, and teaching it “FB posts that look like this are positive, FB posts that look like this are negative”.

This is what GPT-3 can do. It hasn’t been learned just to play one piece. It’s learned how to learn to play new pieces quickly.

GPT-3 allows AI without data :

In AI, entering facts into databases is often called “knowledge graph construction”, and it’s time-consuming and difficult to automate. Google has been working on their Knowledge Graph since 2012 — it’s the thing that powers those helpful info boxes that appear above Google results — but GPT-3 appears to have replicated much of the same content in just a few months of training, with no explicit effort.

GPT-3 just bypasses the problem of “how should I structure my database and how do I get all of my data into it”.

If we can convincingly solve the problem of knowledge update, then I think GPT-3 powered knowledge graph could be incredibly helpful.

I was looking at Gartner Magic Quadrant for Cloud AI, it states that the urgent need to deploy artificial intelligence puts businesses at risk of making bad choices. Hopefully, further developments into GPT-3 in key areas of language, vision, and automated machine learning, can help developers around the world enhance their applications.

GPT-3 model can be accessed via an API. It's a tool and it's up to us to do something incredibly interesting with it.

For sure, GPT-3 shows sparks of cleverness that are sure to accelerate the march of automation and the possibilities of intelligent computer systems.

While results are useful and certainly impressive, at times they could be not completely reliable. This is with 175 billion parameters and 450 gigabytes of input data.

I see a huge amount of possibilities in the field of entity-based chatbots, news article generation, fiction writing, code generation, and powerful search engines.

While, it's just been few weeks of me playing around with the GPT-3 model, my use case involving human-like-machine intelligence in image and voice recognition systems is already showing promise. I am building a platform that will leverage GPT-3 into our everyday life, powering a smart fridge, a self-driving vehicle, to a Holographic virtual assistant.

With OpenAI’s collaboration, I would be publishing my pilot soon. The Force is really strong with GPT-3.

I am incredibly certain that the results will be even more amazing.