In Part II of the Review Barry Phillips looks at what has changed both behind the curtain as well as front of house.

Transcript:

Hello Humans!

My name is Barry Phillips and welcome to the weekly podcast that aims to summarise an important AI development in the world of HR in five minutes or less.

Last week we took a moment to reflect on just how far we’ve come in terms of access to knowledge with the appearance of ChatGPT5— not just in speed, but also in quality. This week, I want to look at what has improved in ChatGPT-5 under the Settings menu by going “behind the curtain” and seeing what’s now possible that wasn’t before.

Before I do that, let’s take a quick look at what has generally improved “front of house”:

First – speed. It’s much faster; answers are now pretty much instantaneous. I can’t help but remember when I was at university and my friend showed me the first spell-checker on his word processor. Whilst we waited for it to check his 1,500-word essay, we went for a game of pool. When we returned, it was just finishing off the last paragraph.

Second – ChatGPT-5 is available for free to everyone. Some commentators worried it wouldn’t be, but OpenAI, in making this free to all, has given the world a tool substantially better than ChatGPT-4. It’s rare we thank companies relentlessly driving for mega-profits, but maybe at this juncture we should.

Third – OpenAI claims this latest iteration of the model doesn’t hallucinate as much as its older siblings. How much better it really is, we don’t yet know — especially when OpenAI guards so closely so much of its data relating to model performance.

Fourth – the context window has doubled overnight. This means ChatGPT can now hold — or remember — a lot more data in each conversation. Previously, it could store an amount equivalent to Wuthering Heights. Now, it’s Ulysses.

Fifth – it’s much better at actually working out what you want it to do from your prompt. Soon, it may understand it better than you do. Death to the “prompt engineer”? Possibly.

Finally – it has simplified its model offering. You no longer have to decide whether to use 4o, 4.5, or 03-mini; it chooses the most suitable model for your query.

But what about the Settings function? What’s better there that means you no longer have to keep tweaking your prompts every time?

Well, you can now do the following:

Connectors – There’s a new section that allows you to connect ChatGPT directly to Gmail, Google Drive, and Google Calendar. Great for the user, though I can see the data-protection people turning grey over this one.

Language of operation – There are now numerous languages to choose from, including Canadian French, Finnish, and Chinese. Be careful here: getting back from Chinese is not an easy task, as I discovered when experimenting earlier today.

Voice mode – There are now ten voice options available for any voice work you have planned.

Customise ChatGPT – You can now choose from different personality types, such as Cynic, Robot-like, Listener, or Nerd. OpenAI assures us it’s also less sycophantic  thn before.

Next week we’ll be back to HR use cases for ChatGPT5.

Until then bye for now.