Just Stop Slop!
The word you didn't know you were waiting for, and other AI news.
Dear Reader,
Welcome to BN Edition: concise analysis on the stories that offer us hints at our unfolding future. Fresh from the desks of the Brilliant Noise team.
Each edition takes a handful of stories from recent weeks and asks three things:
What? The story in a few sentences.
So what? Why do I need to know?
What next? What do I need to do or watch out for?
This week, find out more about:
Apple and OpenAI get into bed
AI can pick stocks better than you
Just before we dive in. A few weeks ago our CEO gave a talk for the Chartered Institute of Marketing on the topic of “How to think with and about generative AI”. You can watch it on YouTube here.
“Slop”
What?
Slop is word that has been repurposed for a modern curse: carelessly created AI content.
Appropriated to describe low-quality, AI-generated content made to generate ad revenue and grub a little SEO advantange, “slop” can also be applied to generic, poorly made images (the new clip-art), and to overly-long emails that obviously weren’t written by a human.
So what?
If “spammy” is mass emailed nonsense that wastes millions of people’s time to get a few sales leads, “sloppy” feels like the perfect term to describe LinkedIn posts, communications and mailers.
You don’t want to be tainted by accusations of slop, either personally or as a company, so use this catchy term to open up discussion about what and why it is important to avoid.
It’s not that AI can’t be very useful in communications and creative, but if you’re outsourcing everything to it and not bothering to edit and check something, then you’ll be setting yourself up for trouble. In the same way that an email that starts “Hi, I was wondering if you were aware of the advantages of outbound sales for your company” makes you delete it in a split-second, so American spelling and idiom (if you’re not American), and some phrases are a dead giveaway that the author wasn’t the sender.
What next?
Let’s talk about it. There’s an ad that says “I just turned a document into a presentation in five minutes”. It should have a health warning.
Joking aside, “slop” is a useful word because it gives us a label for an emerging poor behaviour with AI. The line where slop stops and useful begins is one that will need to be clearly drawn.
Source: Simon Willison, via New York Times
Apple is seeing OpenAI - but like, not long term?
What?
Apple and OpenAI have announced a partnership to integrate AI technology into iPhones starting in September. This collaboration will bring generative AI capabilities to a massive global audience through the Apple ecosystem.
So what?
What this says is: AI is essential. It’s another inflection point, another acceleration point.
Apple’s reputation for safety and user experience means the partnership is a PR dream for OpenAI to address safety concerns about AI usage.
This partnership could force competitors like Google and Microsoft to catch up in terms of usability and integration quality. Additionally, it reflects Apple's cautious but strategic approach, ensuring the technology works seamlessly within their existing products.
What next?
As of September, the world’s most powerful AI will be in a device used by the wealthiest people on the planet. We can expect a phased rollout of AI features in upcoming iOS updates, likely starting with newer devices running on Apple Silicon processors.
This is how the majority of the world will experience and understand generative AI – as smart features on their phone. It will raise the bar for competitors like Google and Microsoft.
However, even though Apple has partnered with OpenAI, it’s not committing to them. It’s very probable they’ll have their own LLM in the making, but for now, they want to use the best available, so they’re betting on OpenAI. This approach is the opposite to what Microsoft are doing which is trying to make Gemini fit into all their existing products. Instead, Apple is using the best partner for now.
Apple is showing what is possible within the current bounds of AI capability.
AI can picks stocks better than you
What?
A study by University of Chicago researchers found that ChatGPT could generate earnings forecasts from financial statements with greater accuracy than human analysts.
The researchers used ChatGPT to analyse thousands and thousands of balance sheets and income statements, stripped of dates and company names, from a database spanning from 1968 to 2021 and covering more than 15,000 companies.
After asking for standard financial analyses, like “What has changed in the accounts from last year?” and “What is the gross margin?” they asked the model to predict whether each company’s earnings in the next year would be up or down; whether the change would be small, medium-sized, or large; and how sure it was of this prediction.
Source: Chicago Booth Business School
So what?
According to the FT’s Lex newsletter, human stock price predictions are correct 57% of the time, which is slightly better than flipping a coin.
After researchers used ‘chain-of-thought’ prompting, which is getting an AI model to work through a task step at a time rather than giving it one huge task to analyse at once, the model got a 60% accuracy score.
What next?
Obviously, this study has plenty of implications and raises a lot of issues, which is why researchers expressed that this is preliminary data, and only a proof of concept rather than a new stock picking product.
But what’s intriguing is the fact that the accuracy of the AI was improved by getting it to ‘think’ more like a human.
Only after conducting basic analysis of the businesses was it able to more accurately predict the future stock price.
“What is perhaps a bit more surprising is that an out-of-the-box LLM was able to outperform humans pretty significantly with quite basic prompts (the model also outperformed basic statistical regression and performed about as well as specialised “neural net” programs trained specifically to forecast earnings.”
Banks and investment firms have spent a lot of money building bespoke predictive models. But those expensive systems are now being beaten by out-of-the box generative AI models like ChatGPT.
Generative AI differs from those old predictive systems because it ‘thinks’ more like a human than a computer – thinking through the implications of the financial statements appeared to be the key to unlocking greater forecast accuracy.
Thank you for reading.
The Brilliant Noise team