In this issue

• Brace for AI’s Big Bang

• New study finds AI boosts work quality by 40%

• Google’s new chatbot ‘5x better’ than ChatGPT

• ChatGPT gets ears, eyes and a voice

• Scientists find telling AI to chill out works wonders

• Why we’ve created The AI Writer

LATEST DEVELOPMENTS

Brace for millions of AI reports, emails and presentations

We’ve quietly entered the next big phase of the workplace AI revolution. And bots that write your reports and email replies are just the half of it.

It may have been the quietest of Big Bangs, but it was as big as they come. Because we’ve just entered the next stage of the AI revolution.

Last month saw a surge in AI developments that will change how we work forever. And we’re not talking about the distant future either. One of the biggest changes of all is imminent.

ChatGPT’s decline in user numbers now seems to have been just a seasonal blip. Students were on summer break and no longer needed it for their homework. Now, like the rest of us, they’re back at their laptops, with work to do – work they know AI can help with.

Nor is it just ChatGPT itself. Every app from card-payments giant Stripe to Zoom and Duolingo are using the tech behind the chatbot, GPT-4, to enhance their platforms. Little wonder that OpenAI, its parent company, now expects to generate $1 billion in revenue over the next year.

AI writing in weeks

But it was another launch, by Microsoft (OpenAI’s biggest investor), that should have created the biggest stir. After months of speculation, the company has announced that, next month, it will finally switch on Copilot, the AI sidekick for its Microsoft 365 suite of work apps.

That means that, within weeks, more than 200 million people who use apps like Word and PowerPoint will start to get access to a bot that can reply to emails, write reports and create entire presentations for them.

I saw Copilot demonstrated in an internal meeting of leaders at a large oil company in London recently. There were audible gasps around the table when attendees saw it generate whole paragraphs and entire slides from just a few words.

Microsoft has now even added a ‘sound like me’ feature, which will analyse your own writing style and (so the company claims) reply in your written voice.

Will it help?

It’s hard to say how much this will help those of us who rely on the software. Writing is hard, and writing matters. So anything that makes it easier is to be welcomed. But the potential for miscommunication at scale must be huge.

Email, for instance, is already one of the biggest causes of workplace arguments. Will letting a bot type your replies make that better, or worse?

Perhaps an AI writing assistant will be better at reading between the lines and replying accordingly. We’ll see. The AI Writer is on the list to get early press access. So we hope to test its features for ourselves before it goes live and report back.

Copilot isn’t free – it will cost up to $30 per user, per month. But that’s probably a price that many companies will be willing to pay. (At least 100 big firms have already forked out $100k each just for the privilege of testing it, as we reported in June.)

But with bots churning out documents and emails at a blistering pace, one thing’s for sure: information overload isn’t going to get better anytime soon.

AI AT WORK

Research: AI boosts work quality by 40%

Extensive research finds that generative AI most benefits workers who usually underperform.

A major study has revealed AI to be a powerful skill leveller in the workplace.

The collaboration between social scientists from several prestigious business schools found that management consultants who used GPT-4 massively outperformed colleagues who didn’t.

The generative AI tool improved work quality by a staggering 40% on average. Users were also able to finish their work 25% more quickly and get 12% more done in the same time.

‘Consultants using ChatGPT-4 outperformed those who did not, by a lot,’ commented Ethan Mollick, one of the lead researchers. ‘On every dimension. Every way we measured performance.’

AI bridges skill gap

But the randomised study of more than 750 consultants from Boston Consulting Group (BCG) found the benefits of AI weren't uniform across the board.

Researchers from Harvard, Warwick and Wharton business schools and MIT’s Sloan School of Management found it was the consultants who did worst in tests without AI who got the biggest boost when using it. This suggests that AI can be a great equalizer, bridging the gap between varying skill levels.

GPT-4 is the language model behind the subscription-only version of ChatGPT. But it’s also available free of charge in the Bing chatbot built into Windows 11 and the Edge browser, and now as the secure, private Bing Chat Enterprise feature recently added to the corporate version of Microsoft 365.

Companies can also build their own apps that run on GPT-4, as BCG did for this experiment.

Tech giants square up to put powerful AI in everything, everywhere

Experts say that Google’s upcoming AI chatbot is ‘five times more powerful’ than ChatGPT. So Open AI launches a new version that can now see, hear and speak. And now even LinkedIn is getting in on the AI act.

A meme did the rounds a few years ago. It was a spoof ad for a job vacancy at Google. ‘No need to apply,’ it read. ‘We already have all your information.’

The truth behind the joke explains why expectations are high for the company’s upcoming rival to ChatGPT, Gemini. Unlike OpenAI’s chatbot, Gemini can train on the huge amount of data Google gathers from its other products, including its eponymous search engine.

Google has just given early access to a small group of companies for testing, which suggests that Gemini could go live soon. Some analysts claim it is already five times more powerful than GPT-4, OpenAI’s latest software and that it’s on course to be 20 times more powerful than its rival by the end of next year.

AI everywhere soon

It certainly seems to have OpenAI worried. Less than two weeks later, the company announced that it was rolling out GPT-4 V(vision), which can see, hear and speak.

An arms race between two AI superpowers can only accelerate the pace of change.

Even LinkedIn is getting in on the act, apparently. It’s emerged that it’s adding a whole raft of AI features, including AI-powered learning. A bot even offered to edit my profile recently and it did a fair job.

Unfortunately, it also tried to embellish my bio with a few things that I haven’t actually done. Maybe AI is getting closer to being human than we thought …

Who owns LinkedIn? Microsoft again, which – as we predicted in May – really is now adding AI to everything, everywhere, all at once.

Things are about to get even more interesting.

PROMPT ENGINEERING

It may pay to tell ChatGPT to chill out

Scientists investigating how to phrase requests found that telling an AI to take a breath before answering dramatically improved its answers.

Large language models like those that power ChatGPT may be one of the pinnacles of human achievement. But even the brightest minds behind them still don’t fully understand how they work.

As if to illustrate this, researchers have been studying what kind of language to use for the best results, just as psychologists might conduct experiments to reveal how human brains react to different words and phrases.

Now, scientists at Google’s DeepMind AI team have discovered that the best ways to communicate to humans and to AI models are even more closely linked than first thought.

We’ve said before that writing clearly works best with both bots and people. And the study backed up our previous advice to break down long requests into several shorter ones to improve results.

But the researchers also found that it pays to go even further.

And … breathe …

They set Palm 2 – one of Google’s AI bots – a series of logic problems, varying their phrasing and measuring the accuracy of the results it gave them.

When they just gave it the problem, the bot got one in three tasks right (34%). Adding the instruction ‘Let’s work this out in a step-by-step way to be sure we have the right answer’ increased the success rate to almost two out of three (58.8%).

But when told to ‘Take a deep breath and work on this problem step by step’, it got the answer right four out of five times (80.2%).

Maybe it’s the Brit in me, but I usually feel compelled to be polite in my dealings with AI bots, being careful with my pleases and thank yous. I know they’re not human, but it seems to work better (as does a little encouragement for a good result).

But even I would never have thought to go this far. What’s next? Positive visualisation prompts? Giving ChatGPT a little pep talk before you ask it to do something?

If nothing else, it might earn you a little credit for when the bots finally rise up in the AI Apocalypse.

Why The AI Writer?

When I founded Emphasis, 25 years ago, it was just me, a cat and a kettle. It’s since grown to become the most trusted writing consultancy in the world and helped more than 80,000 professionals from 32 countries.

We’ve worked with tech giants, top 10 law firms and major financial institutions, as well as at the highest levels of government. (We’ve even sent trainers to work with clients in the Himalayas.)

I would not claim that we’re experts in everything, as we’re certainly not. But, after working with the authors of around 100,000 documents, it’s safe to say that business writing is something we do know a thing or two about. And that includes witnessing all the various ways in which organisations get it wrong.

Now, we’re at the forefront of using AI to help them get it right.

Of course, we’ve seen nothing like the AI revolution that’s hitting us. Nobody has. But I do feel a responsibility to use our experience to help people navigate this brave new world of communication bots.

So here’s the deal. You get on with your job, while we immerse ourselves in the latest developments in workplace AI. We’ll worry about keeping up so you don’t have to. Then we’ll tell you exactly what you need to know to stay ahead of the curve.