ChatGPT - Universal Translator

Welcome, AI Communicator!

In today's issue:

  • ChatGPT - Universal Translator

  • Sora - End of Authenticity?

  • Gemini Pro 1.5 - Memory Master

ChatGPT - Universal Translator

Ever sat in a meeting and felt everyone was speaking a different language?

Lawyers, accountants, engineers, medics, even communications and public relations professionals do it.

They love their jargon and technical terms. It’s an aid to communication between those who understand them - and excludes those who don’t. 

If you watch science fiction, alien species usually understand each other effortlessly. 

Star Trek has the universal translator to achieve this - mostly as a contrivance to keep the story flowing.

Now, we have ChatGPT. 

This is not sci-fi any more. You have much of this functionality right now on your phone. 

Take a picture of a menu or a sign on holiday and ChatGPT can translate it. Speak to it in English and it can answer you in Spanish.

It is also excellent at translating from one kind of English to another. 

From technical report to Plain English.

From formal to colloquial.

From PhD thesis to readable by a nine year old. (The average reading age in the UK.)

From medical terminology to easily understood by a distressed patient.

And it doesn’t just switch out the difficult words for simpler ones. It can translate hard-to-understand concepts into simple metaphors.

My six-year-old son had an ear infection last week.

The antibiotic smelt disgusting. His words.

And he wouldn’t take it.

I asked ChatGPT to explain to him why it was important. 

First it gave a nice little explanation about the medicine being like a superhero team.

Then another about it being like a power-up in a video game and the ear infection being the big boss level.

He still wouldn’t take his medicine. AI can only do so much.

ChatGPT’s powers of translation go both ways. 

Describe something in everyday layman’s terms and get a technical spec or a professional brief. 

Have a concept for an app? ChatGPT can even translate it into code. 

It can write in Python, JavaScript, SQL, C/C++.

No idea what the last sentence meant?

Ask ChatGPT.

Takeaways:

  • Use ChatGPT (or the Large Language Model of your choice - Gemini, Copilot, Claude) to check what you have written is tailored to your audience.

  • Provide the text and details of the intended readers and ask how easily they would understand it. Ask for suggested changes.

  • Ask ChatGPT to provide Plain English summaries of complex reports and publication. These can then be provided at the start of the document.

  • Ask ChatGPT for suggested metaphors - written and visual - to help explain complex concepts to your audience.

Sora - End of Authenticity?

From the polarised reactions on social media, if you’ve seen the videos from Sora you were either in awe or wanted to ban all AI immediately.

Sora is a text-to-video model announced by OpenAI last week. It can generate astonishing videos of up to 60 seconds from a prompt. Like the Labrador in a hoodie typing on a keyboard.

It hasn't been rolled out publicly yet. But when it does will anyone want to watch your non-AI videos? 

Can dull reality compete if you can generate a realistic video of someone delivering your message while walking on the moon?

It might be that raw and unpolished videos will be a signifier of authenticity. 

But there’s a catch. They can be generated by AI, too. 

The only hope is a watertight system of watermarking, or high-accuracy AI video and image detection tools. That may or may not be possible.

This could put a premium on in-person live events as the best way to be trusted as genuine and authentic. 

Gemini Pro 1.5 - Memory Master 

While OpenAI’s Sora stole its thunder, Google’s announcement of Gemini Pro 1.5 was hugely significant in its own right.

ChatGPT can struggle with finding detail within a large document or dataset. Gemini Pro 1.5 has no such problems. 

It can take up to eight times as much data as GPT-4. This means Pro 1.5 can handle 700,000 words, 11 hours of audio, or one hour of video. 

This gives it the ability to analyse vast datasets of trends, sentiment and feedback to enable precise targeting of messages to specific audiences.

There is limited access so far. But it is another glimpse into the future potential of AI for communications. 

I’d love to hear what you think about the newsletter. Reply to this email with feedback or questions.

Thanks for reading. 

P.S. To sign up or share with a colleague you can find the newsletter here. Make sure you get the next issue by moving this email to your primary inbox