Antonym: The 200X Productivity Edition
Defying Brevity.
Dear Reader
The impact of generative AI (GenAI) on our lives will be hard to measure, but measure it we must, if we want to persuade others to commit to learning and innovation.
For some time at Brilliant Noise we have relied on the findings of the Harvard Business School study of consultants using GPT-4 (still the leading commonly available GenAI system) compared to peers without at Boston Consulting Group (BCG). In a paper called “The Jagged Frontier”, they described increases in productivity and quality of work from the group using AI. 12-25% productivity (stuff they could get through) on some tasks and 40% increase in quality (how good the work was) are compelling in their own right.
Those numbers felt conservative, especially when you are working with GenAI tools everyday, as we are. But even when a client says they got a week’s work done in half an hour with their team, it sounds almost too incredible to start sharing that in articles and presentations. I tend to mention it as an anecdote, but unless someone has experience working with AI, I’m afraid that it just doesn’t sound realistic. The gains seem too big. But they are real.
This week I got some more numbers from third parties that are credible and that I will share, starting here:
A BCG MD says they are seeing productivity increases of 40-50% when GenAI was integrated into processes. That’s to say, they are getting double or more than double the things done than before when they don’t just use ChatGPT or similar to do a task, but integrate into the whole sequence of how they work.
In an FT article the UK Prime Minister’s chief analyst scientist and data science director said ‘“While there was a general target for a 3.55 times return on investment on AI, […] some of the early tools are more like a 200 times return on the investment”.’ We wrote more about his in the BN Edition newsletter.
Productivity is a trap, as Olive Burkeman said, and it’s certainly not the only benefit from using GenAI. But when then there are numbers like this backing up integrating the tech into everyday work, the question shifts from “should we invest” to “why on earth would we not”. Those incredible returns and boosts from using AI are available right now to anyone who wants to start to learn.
The next question for us all will be how to use that extra time wisely, but in an age plagued by busywork, inefficiency, burnout and too little time for learning and innovation, there’s no shortage of ideas.
Move away from the robot
I propose that this advice from the Associated Press style book become a hard rule for anyone writing about social media.
The similarity of clunkily prompted AI images to accompany presentations, articles and reports makes GenAI the new clip art.
AI is Russian Roulette for our species, says Hinton
Geoffrey Hinton, a prominent AI researcher, warns about the risks of AI, including bad actors using AI for harmful purposes and AI evolving in dangerous ways. He believes there is a 10%-20% chance that AI could wipe out humanity in the next 20 years and suggests increasing the focus on safety issues. Hinton's change in perspective came from realising the potential power of AI models acting as "hive minds" and sharing knowledge with each other.
Given climate change and increasing chances of nuclear war, our species is playing a few games of Russian roulette all at once.
Source: FT
Meanwhile using generative AI is increasing water use
Data centres are consuming huge amounts of water as part of the demand for generative AI servives. Rachel Coldicutt raises the question of how this impact can be shifted fast.
Along with increased mineral extraction and energy use, AI and related technologies are massively extractive and resource hungry, and are expediting the climate emergency. At the very least we need to see a pivot to renewable energy sources for AI; ideally there needs to be a fundamental re-engineering and rethink of AI's material impact.
The problem is not simple, but it is real and pressing. Organisations will need to include Gen AI use in their calculations of energy use and carbon, as they do with all tech use.
What will be harder to manage – but ways will emerge – will be figuring out when 20 minutes of gen AI use replaces a week’s work. Will that efficiency be seen as carbon reduction, or if it simply allows for more works, a accelerator for carbon hungry processes?
Some solutions are simple, and not exclusive to AI on a personal level:
Don’t use cloud storage like an attic – it costs real resources and carbon to keep gigabytes of undeleted files and photos in the cloud.
Use a low carbon energy provider.
Look for reports and policies on data centre use and sustainability on tech providers’ websites.
If you’re a prolific user, think about using smaller AI models for more mundane tasks.
From the Writing For Busy Readers newsletter
I love me a bit of Churchill trivia. The Writing For Busy Readers newsletter (promoting the book of the same name) shared this memo from Churchill asking everyone to cut the nonsense and get to the point in internal memos. Titled “Brevity” it’s aged rather well.
Todd Rogers goes on to talk about studies of communication by government:
I recently read a great paper by a team that included Jessica Lasky Fink, my Writing for Busy Readers coauthor. Across a series of experiments with governments, they found that formal writing was more effective than informal writing.
Two notes: first, this is specifically about government communications. They found that people interpret informal writing as not credibly coming from the government. Second, formality does not mean hard to read—it refers to style and tone.
The Shape of AI
The Shape of AI website fascinating. Emily Campbell, a designer is meticulously tracking the evolving patterns and norms that govern the intersection of imagery and language as we delve into generative AI technologies. At this nascent stage, our approach to leveraging these tools is remarkably experimental. The project also serves as a poignant marker of the infancy of the generative AI revolution, and the the rapid pace at which it's advancing.
This week I’ve been…
Watching
Reality (Pay to stream, or with ads on Prime). An odd an unsettling drama about the FBI arriving at a young woman’s house. The awkwardness and tension is incredible. Based on the true story of Reality Winner. If you don’t know about her, don’t Google it before you watch.
Dune, part 2 (Cinemas)
Everything one could possibly hope for – Denis Villeneuve knocks it out of the park. It’s heavy on stylish visuals and mood, low on irritating plot detail. One scene where desert folk are discussing Timothée Chalamet’s possible “chosen one” status got an inadvertant laugh from the audience as it too closely recalled Life of Brian’s “only the true messiah would deny his divinity” line.
Kin (BBC iPlayer). Finished watching the second series of this brilliant drama this week. Francis Magee ruled it. Great writing and the jarring comic name of the pub “The Furry Bog” where the gangster he plays bases his operations delighted throughout. Apparently it is a real pub, now sadly closed.
Reading.
Burn Book, by Kara Swisher. I’m a fan of Swisher’s work, and have a soft spot for her Pivot podcast with Scott Galloway. She’s covered Silicon Valley and big tech since the late 90s, and is fascinated by the tech and unimpressed by the egos of the players.
Earthlings, by Sayaka Murata. Beautiful and hideously tragic in turns. Childhood innocence and trauma segues into trying to survive as an adult in world that makes no sense.
That’s all for this week…
Thanks for reading and if you liked it, give us a share!
Antony