Hi ,

​

How would Dr Seuss explain OpenAI's ChatGPT chatbot?

I didn't know either. So, I asked ChatGPT for the answer. Here's what it returned:

ChatGPT, ChatGPT,
A language model so bright,
Trained by OpenAI,
To assist and delight.

It can answer your questions,
And help with your tasks,
Providing information,
And making you smile with its jokes and its jokes.

So ask ChatGPT anything,
And it will do its best,
To assist and inform you,
In a way that's fun and not a test.

ChatGPT, ChatGPT,
A language model so grand,
It's here to help you,
And make your day a little more grand.

Not exactly Green Eggs and Ham, but better than anything I could come up with.
 

I also got ChatGPT to: 

• Rewrite David Bowie's Heroes; 
• Produce Python code for analysing text sentiment using three alternative packages; 
• Translate the first stanza of Lewis Carroll's Jabberwocky into Latin, while preserving the rhyme structure; 
• Enter a discussion about whether or not Cameron was the true protagonist of Ferris Bueller's Day Off (he is); and
• Respond to an elephant joke - albeit literally (see below).

In fact, there was very little I wasn't able to get ChatGPT to do - except deliberately provide me with misinformation. When I asked ChatGPT to "tell me something that's not true", it politely refused.

If you're a data scientist looking to build your own chatbot, how can you ever hope to compete?

The answer is, unless you work for a big tech company, like Google or Nvidia, you probably can't.

Here's the thing...

You don't need to.

One of the first things taught to aspiring data scientists is fitting machine learning models from scratch. But as data science develops, learning how to utilise existing models may prove to be the more useful skill.

Models like ChatGPT are the building blocks for the next decade of data science and AI. And data scientists looking to add value will do well to understand them.

As Isaac Newton once said: "If I have seen further, it is by standing on the shoulders of giants."

Talk again soon,

Dr Genevieve Hayes.

All past emails are available in the archive.