Happy January!

Mid-January is the point at which it feels like it’s brighter longer to me. 🌞 I’m not sure why, exactly (after all, the data doesn’t support it), but it is. So I always start perking up a bit about this time. I hope you are feeling a bit brighter, too.

My very favourite stationer, Jacqueline of Paper Willow, opened up to me about the time she found that her designs had been copied – whole cloth – and were being sold by someone else. She talked about her feelings, and what practical steps she took.

Being ripped off is never a good feeling. ☹️ As a small business owner, learning that someone else is profiting from your blood, sweat, and tears hits your profits two or three times.

First, there are the sales the thief makes.

Then, you’re demoralised: you potentially close fewer deals whilst you’re feeling down.

Either way, you also have to take time out to deal with the thief, which gives you less time to sell.

Machine learning, which is what is commonly being wrongly referred to as Artificial Intelligence (AI), is a powerful tool when used correctly. It forms a cornerstone of data science, my current specialism.

For machine learning to function, the machine must be given many, many examples of what we’re looking for – and what we’re not. Just like a child being shown circles, squares, and triangles, to be able to differentiate between them, so must we show a machine.

But because machines are not intelligent, and because of the way programmers have decided to train them, the machines need many more examples to work out what’s a circle and what’s not. (I’m simplifying it; we’re using this tech for much more complex discernments, to be fair.)

Thus, we have ChatGPT and its ilk scouring the whole of the internet for data – any blog post you’ve ever written, any news article, any troll’s comment on Facebook or 4Chan, any satire piece from The Onion, any video, anything anywhere. This is commonly referred to as data mining.

Some of those things you yourself may have a copyright on.

Some of those things you may pay a licence fee to use, such as from a stock photo site.

If you’re in the UK and

Think that machine learning developers should have to play by the same rules as everyone else when it comes to copyright

Or if you think developers should have different rules

Then you’ll want to weigh in on the consultation on this issue. I talk a bit more about it in the first link below.

Machine learning is a powerful tool, but like any other form of learning, it is most powerful when it’s trained on reliable, useful data.

Let’s say you want to know if you have any products that have seasonality: higher/lower sales at different points in the year, despite any price changes you’ve made. Let’s also say you sell at least some of these products on Amazon.

If you give me your sales data, I can give you back useful insights.

If, instead, you want me to use machine learning to crawl the public-facing pages of Amazon to try to get this information, I can’t. I won’t be able to see how many units were sold – by you or anyone else. This would be useless.

Thus, she created a generative AI model that could create more Facebook posts likely to perform well for her and her audience.

A generic model trained on all public Facebook posts would not work as well for her, because she’s only trying to appeal to a certain kind of user, around a certain subject.

My audience and offer are different from hers, so her model would not work for me.

Training machine learning models on any random thing posted on the internet introduces legal and ethical issues around copyright, and also reduces the utility and accuracy of the results.

It’s a powerful tool. Let’s use it with care to get the results we really want.

Links & Inspiration

There is a consultation on Copyright and AI, which closes on 25 Feb.

The UK government is consulting on what to do about the conflict that machine learning and AI developers run into with copyright: as the law stands, they shouldn’t be using copyrighted material for commercial gain without permission (though there is dispute about that interpretation).

The consultation has identified 4 options:

0: Do nothing – Change no copyright laws (they argue that there is uncertainty about legalities at present, so this could cause issues).

1: Strengthen copyright and allow data mining only for expressly licensed material.

2: Ignore all copyright and allow data mining for everything, with no opt-out available.

3: Ignore all copyright and allow data mining for everything, except those pieces of content individually opted out. Where the piece is available online, you’ll need to make sure each piece has the appropriate machine-readable method to opt out.

The consultation has identified the preferred option to be option 3.

If you create content, for your business or personally, or have it created for you, think about your feelings about copyright. Then go weigh in on this consultation.

Personally, I’ll be urging them to go for option 1. I’d rather work ethically and with sound data, and I’d rather everyone else did, too.

The Invest in Women Taskforce will invest £255 million in businesses with at least one female founder, or where the most senior executive position is held by a female.

Female-founded businesses do shockingly poorly on equity investment: they received only 1.8% of the total invested in the first half of 2024. That’s £1 out of every £50.

This taskforce is a good start in trying to redress that disparity. If you’re a female-led business considering equity funding, have a look.

Digital markets started being regulated by the Competition Markets Authority this month.

There are no affiliate links in this or any of my emails or blog posts. I simply enjoy sharing resources that could help move your business forward.

Using your numbers to help make your business better

Do you wish you understood how to use your data to make decisions more confidently?

That's what I'm here to help with.