Tag: ai

Supremacy: AI, ChatGPT and the race that will change the world by Parmy Olson

fiendfullyreading

Supremacy is a book charting the race between OpenAI and Deep Mind to bring out their AI products, focusing on their founders, Sam Altman and Demis Hassabis. Journalist Parmy Olson explores these two founders’ starts in the tech industry, interest in AGI (artificial general intelligence) and ultimately the race between their two companies (and the tech giants behind them) to have the best AI product on the market in this generative AI age.

As someone who reads and teaches people about AI, I was interested to see how this quite recent race would be turned into a book. The narrative starts early and quite broad, looking at Altman and Hassabis’ initial failures and interest in AI, as well as early tech industry contacts, and I found this part of the book was a bit too obsessed with the ‘tech genius’ idea, not just from them but around people like Elon Musk as well. Thankfully, as the book goes on, Olson moves away from this idea and looks more broadly at the big picture, including the struggles around both ideas of AI safety and of AI ethics, and the need for the AI world to be dominated by existing tech giants like Microsoft and Google.

The part of the book describing the invention of the ‘transformer’ and the impact of this on work to build better AI models was a highlight as it was an approachable explanation of why this breakthrough was so important, helping people to understand why these huge AI products seemed to come out of nowhere a few years later. I found the section that explores how effective altruism ended up connected to some of the movements in AI also interesting, showcasing how it is often the ideas of billionaires that have a massive impact on world-changing technologies. There was plenty I learnt from the book, even as someone who has a fair interest in the topic, and knowing where these companies came from is a useful part of critiquing and evaluating generative AI.

I did find that sometimes the book was so focused on tech billionaires and companies that it dragged, sometimes accepting at face value what these people say and argue for. Especially by the end, there was decent discussion of many of the issues surrounding AI, but I did think some were notably absent, particularly the climate impact of the GPU power needed and the human cost of data labelling for training data. The climate angle in particular I felt was needed, given that these companies often try and hide their negative climate cost, and it links back to other technologies like cryptocurrency that are mentioned in passing in the book. I do think that the way the book clearly distinguishes between AI safety and AI ethics, and how these can even be in conflict with each other, was very useful, especially for raising awareness of these to a general audience reading the book to learn more about the world of AI as it has become.

Overall, Supremacy is a detailed account of these two AI juggernauts over the past fifteen years and the road to get to tools like ChatGPT that have become household names, and it is a good place for people who want to know where AI has come from recently to start. For me, I found it did lean too heavily on ideas of the solitary tech genius billionaire and I wasn’t interested in that much detail about conversations between them, but the book didn’t go entirely in for the AI hype and did address a lot of the issues and controversies around AI at the moment.

Code Dependent: Living in the Shadow of AI by Madhumita Murgia

fiendfullyreading

Code Dependent is an investigation into the human side of AI: the ordinary, non-Silicon Valley people affected by and involved in areas relating to artificial intelligence. Journalist Madhumita Murgia tells the stories of people and communities impacted by AI from people labelling AI training data to people whose lives are changed by the decisions of AI systems or having deep fake videos made of them. Not everything is negative: there’s also healthcare benefits, if only these technologies can be made freely available and in places that most need them. And as the book moves towards the ending, Murgia argues that these stories give us principles we should consider going forward to ensure AI works for ordinary people, not the other way around.

Notably, this book focuses on the human side of technology, rather than the technological side, and foregrounds the experiences of people and the complexity of AI’s role. Even for areas that are often discussed in other books, such as predictive policing, this book offers examples I’ve not seen before and direct interviews with people affected, not something all technology books have. At the same time, it does provide an accessible description of a lot of AI-related technologies; for example, it’s the first time I’ve seen—as someone who reads a lot about AI—a simple explanation of what a ‘transformer’ is and why it has been so important for generative AI. This combination makes Code Dependent useful both for people who do read tech books, but are interested in human stories rather than the same talking points, and people who are newer to the topic and would like a way in that focuses on people.

Sometimes I found the framing or phrasing a bit simplistic or lacking nuance and complexity, but generally, it was an accessible book about AI that tells stories rather than just facts, and takes areas we might have heard or read plenty about and shows specific people’s lives in relation to these topics. The parting message about religions coming together to discuss AI was not where I expected the book to go and I’m not quite sure how I feel about that being the conclusion (given that high up people in a religion aren’t really ‘ordinary people’ necessarily), but I do appreciate that this was a book about AI that had a lot of things I’d not read about before, or at least not in this form.

Given the current hype and fear around AI, Code Dependent is likely to become a much-talked-about book, offering people a different way in to reading and thinking about artificial intelligence and what it means for our lives.