For Christmas I got an interesting gift from a good friend - my extremely own "best-selling" book.
"Tech-Splaining for Dummies" (fantastic title) bears my name and my image on its cover, and it has radiant reviews.
Yet it was totally written by AI, with a couple of easy triggers about me provided by my buddy Janet.
It's a fascinating read, and uproarious in parts. But it likewise meanders rather a lot, and is someplace between a self-help book and a stream of anecdotes.
It mimics my chatty style of composing, however it's likewise a bit repetitive, and very verbose. It may have gone beyond Janet's triggers in collecting data about me.
Several sentences begin "as a leading technology journalist ..." - cringe - which could have been scraped from an online bio.
There's also a strange, repetitive hallucination in the type of my cat (I have no animals). And there's a metaphor on nearly every page - some more random than others.
There are lots of companies online offering AI-book composing services. My book was from BookByAnyone.
When I got in touch with the primary executive Adir Mashiach, based in Israel, he informed me he had sold around 150,000 personalised books, primarily in the US, because rotating from compiling AI-generated travel guides in June 2024.
A paperback copy of your own 240-page long best-seller expenses ₤ 26. The company utilizes its own AI tools to create them, wolvesbaneuo.com based upon an open source large language model.
I'm not asking you to purchase my book. Actually you can't - only Janet, who developed it, can buy any additional copies.
There is currently no barrier to anybody developing one in anybody's name, including celebrities - although Mr Mashiach states there are guardrails around abusive content. Each book includes a printed disclaimer specifying that it is fictional, developed by AI, and developed "exclusively to bring humour and delight".
Legally, the copyright belongs to the company, but Mr Mashiach worries that the product is planned as a "personalised gag gift", and the books do not get sold even more.
He wishes to expand his range, creating various categories such as sci-fi, and possibly offering an autobiography service. It's designed to be a light-hearted type of customer AI - offering AI-generated items to human clients.
It's likewise a bit terrifying if, like me, you compose for a living. Not least since it probably took less than a minute to generate, and it does, certainly in some parts, sound similar to me.
Musicians, authors, artists and stars worldwide have actually expressed alarm about their work being utilized to train generative AI tools that then churn out comparable material based upon it.
"We should be clear, when we are discussing data here, we in fact imply human developers' life works," says Ed Newton Rex, founder of Fairly Trained, asystechnik.com which projects for AI firms to respect developers' rights.
"This is books, this is short articles, this is photos. It's works of art. It's records ... The entire point of AI training is to learn how to do something and after that do more like that."
In 2023 a tune including AI-generated voices of Canadian vocalists Drake and The Weeknd went viral on social media before being pulled from streaming platforms because it was not their work and they had actually not consented to it. It didn't stop the track's creator attempting to nominate it for a Grammy award. And even though the artists were phony, it was still hugely popular.
"I do not think the use of generative AI for innovative functions must be banned, but I do believe that generative AI for these functions that is trained on people's work without authorization should be banned," Mr Newton Rex adds. "AI can be really powerful however let's build it ethically and fairly."
OpenAI says Chinese rivals using its work for their AI apps
DeepSeek: The Chinese AI app that has the world talking
China's DeepSeek AI shakes industry and dents America's swagger
In the UK some organisations - including the BBC - have selected to block AI designers from trawling their online content for training functions. Others have decided to team up - the Financial Times has actually partnered with ChatGPT developer OpenAI for instance.
The UK federal government is considering an overhaul of the law that would allow AI designers to utilize creators' content on the internet to help develop their designs, unless the rights holders opt out.
Ed Newton Rex describes this as "madness".
He explains that AI can make advances in locations like defence, health care and logistics without trawling the work of authors, journalists and artists.
"All of these things work without going and changing copyright law and messing up the incomes of the country's creatives," he argues.
Baroness Kidron, a crossbench peer in your home of Lords, is also strongly versus getting rid of copyright law for AI.
"Creative markets are wealth developers, 2.4 million jobs and an entire lot of delight," says the Baroness, who is also an advisor to the Institute for Ethics in AI at Oxford University.
"The government is weakening one of its best performing industries on the vague pledge of growth."
A federal government spokesperson said: "No relocation will be made till we are definitely positive we have a useful plan that delivers each of our objectives: increased control for right holders to help them certify their material, access to premium material to train leading AI models in the UK, and more openness for right holders from AI developers."
Under the UK federal government's new AI strategy, a national information library containing from a wide variety of sources will also be provided to AI researchers.
In the US the future of federal rules to control AI is now up in the air following President Trump's go back to the presidency.
In 2023 Biden signed an executive order that intended to enhance the security of AI with, amongst other things, firms in the sector required to share details of the functions of their systems with the US government before they are released.
But this has now been reversed by Trump. It stays to be seen what Trump will do instead, however he is said to want the AI sector to face less regulation.
This comes as a number of suits against AI companies, and especially versus OpenAI, continue in the US. They have actually been taken out by everybody from the New york city Times to authors, music labels, and even a comic.
They declare that the AI firms broke the law when they took their content from the internet without their permission, and used it to train their systems.
The AI companies argue that their actions fall under "reasonable usage" and are for that reason exempt. There are a number of factors which can constitute fair usage - it's not a straight-forward definition. But the AI sector is under increasing scrutiny over how it collects training information and whether it should be spending for it.
If this wasn't all adequate to contemplate, Chinese AI firm DeepSeek has shaken the sector over the past week. It ended up being the most downloaded totally free app on Apple's US App Store.
DeepSeek claims that it established its technology for a fraction of the cost of the similarity OpenAI. Its success has raised security concerns in the US, and threatens American's existing dominance of the sector.
When it comes to me and a career as an author, I think that at the moment, if I really desire a "bestseller" I'll still need to write it myself. If anything, Tech-Splaining for Dummies highlights the current weak point in generative AI tools for larger jobs. It has plenty of mistakes and hallucinations, trademarketclassifieds.com and it can be rather hard to check out in parts because it's so long-winded.
But offered how rapidly the tech is progressing, I'm not sure for how long I can stay confident that my substantially slower human writing and editing skills, are much better.
Sign up for our Tech Decoded newsletter to follow the biggest advancements in global innovation, with analysis from BBC reporters worldwide.
Outside the UK? Register here.
1
How an AI-written Book Shows why the Tech 'Horrifies' Creatives
Beau Heng edited this page 2025-02-03 16:21:56 +08:00