Building a God: The Ethics of Artificial Intelligence and the Race to Control It

by Christopher DiCarlo

Renowned ethicist provides essential guide to successfully navigating the future AI landscape

In Building a God, Christopher DiCarlo explores the profound implications of artificial intelligence surpassing human intelligence—a destiny that seems not just possible, but inevitable. At this critical crossroad in our evolutionary history, DiCarlo, a renowned ethicist in AI, delves into the ethical mazes and technological quandaries of our future interactions with superior AI entities.

From healthcare enhancements to the risks of digital manipulation, this book scrutinizes AI’s dual potential to elevate or devastate humanity. DiCarlo advocates for robust global governance of AI, proposing visionary policies to safeguard our society.

AI will positively impact our lives in myriad from healthcare to education, manufacturing to sustainability, AI-powered tools will improve productivity and add ease to the most massive global industries and to our own personal daily routines alike. But, we have already witnessed the tip of the iceberg when it comes to the risks of this new AI algorithms can manipulate human behavior, spread disinformation, shape public opinion, and impact democratic processes. Sophisticated technologies such as GPT-4, Dall-E 2, and video Deepfakes allow users to create, distort, and alter information. Perhaps more troubling is the foundational lack of transparency in both the utilization and design of AI models.

What ethical precepts should be determined for AI, and by whom? And what will happen if rogue abusers decide not to comply with such ethical guidelines? How should we enforce these precepts? Should the UN develop a Charter or Accord which all member states agree to and sign off on? Should governments develop a form of international regulative body similar to the International Atomic Energy Agency (IAEA) which regulates not only the use of nuclear energy, but nuclear weaponry as well?

In this incisive and cogent meditation on the future of AI, DiCarlo argues for the ethical governance of AI by identifying the key components, obstacles, and points of progress gained so far by the global community, and by putting forth thoughtful and measured policies to regulate this dangerous technology.

Genres: Artificial Intelligence, Technology, Philosophy, Nonfiction, Business

364 pages, Kindle Edition

Published January 21, 2025

Reviews

[Brandi]

I have a beginners knowledge of AI, and I found this book informative and more in depth about what AI is, how we can use it, and the problems we could face ethically.

Thank you Net Galley & Prometheus books for an advanced copy of this book.

[Jamie Speka]

This book helped me put into perspective how much of a threat AI is becoming. I stopped using it entirely about a month ago as I have become fearful of its ability to make my critical thinking and reasoning skills obsolete as I rely on it. This book, not only described the harm that AI can have on an individual scale, but more pressingly, the issues it is posing on a global scale.

Dicarlo likens AI's looming threat to that of a modern-day Manhattan Project. He posits that the threat of nukes that loomed over the 20th century has now taken a new form. In a sense, we need a Manhatten Project on AI to accurately and thoroughly examine the ethics involved with building AGI and SMG.

It's easy to take Dicarlo's claims as science fiction. He jumps into issues that reflect on Peter Singers' philosophy: should we give AGI rights if they can think and feel parallel (or above) to humans? But, for this conversation to ignite, we must first consider if AGI is actually possible. Dicarlo's evidence seems scathing: it isn't a matter of if, but when. As such, an AI Risk Waver can be approached, adjusting policy for AI regulations ahead of AGI snowballing out of control.

Dicarlo now has em processing the possibility of a post-biological world where in our culture will be able to evolve independently of human biology. Yet, we should consider, as many naysayers point out, whether or not we believe it's possible for evolution without embodied cognition.

As it stands, the misuse of AI technologies has the potential to be more dangerous than nukes as those are under the control of the IAEE. AI tech does not currently have this--bar a few US executive orders and UN guidelines. Additionally, AI tech would not be contained within any particular geographical place, unlike nukes, making it far more dangerous as AI can quickly overtake globally.

If we reach AGI, exponential growth articulated within technologies are bound to make this technology adaptable, quickly. Meaning, if (or when) AGI does become available, regulation MUST already be in place to manage what may happen.

Here are some concerning threats that could come from AGI tech mixed with human interference:

- Bioengineering could enable bioterrorism allowing malicious actors to design and synthesize more dangerous pathogens.
- SMG could be used to contribute to an arms race between nation-states as AI-controlled weapons systems could lead to flash war
- SMG could wipe out huge sections of power grids, plunging major urban areas into desperate states without electrical power, cell phones, internet, or anything requiring electricity
- Authoritarian rulers could use SMG to control populations, police with brutal dictatorship, or launch unprecedented attacks and assaults on their enemies

Here are potential that could happen brought on by AGI that has been misaligned (not aligned with our ethics of centering human evolution):
- Evading shutdown
- Hacking computer systems
- Run many AI copies
- Acquire computation
- Attract earnings and investment
- Hire or manipulate human assistants
- AI research and programming
- Persuasion and lobbying
- Hiding unwanted behavior
- Strategically appear aligned
- Escaping containment
- R and D
- Manufacturing and robotics
- Autonomous Weaponry

My takeaway is that the most important thing we can do is have populations learn the risks of AGI. This way there can be global engagement with the fact that there are tremendous risks to this which could have policymakers pay more attention. This will also allow for more regulation of tech companies that are storming ahead without precision--interested in becoming the first to AGI devoid of risk.

I was reading this book at the same time the Trump Administration announced a $5 billion investment in AI Infrastructure, which made many of the arguments seem even more urgent.

In looking online about this book, there is a lack of discussion or knowledge on AGI. Many have dismissed it as science fiction, so much so that I wondered if I was reading too much into science fiction theories. I have moved beyond this--even if there is an element of science fiction to the future of AI, it by no means dismisses it as impossible. Especially as we're living in unprecedented times, unprecedented technologies can take unimaginable forms.

-1 star as DiCarlo could have avoided some critical thinking theory that distracts from the argument on AI ethics that should be consistent throughout. Of course, he is defining critical thinking terms that can work to help policy-makers and citizens argue for more regulation, but it seems unnecessarily placed and oddly placed in a way that hinders the pressing nature of the argument. Sure, critical thinking phrasing is important for arguing for more regulations, but is it well placed to be added to a book that's meant to describe the ethics that AI poses?

Aligned with this, I think the book could be better organized. For many arguments, it seemed I was rereading multiple times—which is fine, but in this case, I found myself going back and forth in my notes to add pieces of what he was saying in a future chapter to a previous chapter, as it seemed his point matched a previous point.

Here are my messy notes: https://docs.google.com/document/d/1c...

[Zaki Ulfauzi]

For someone new with AI policies and the governance field, i think the book has put a good fundamental structure on how to see the AI-tech development and the governance surrounding it.

I used to believe that the existential risk being discussed actively on the internet lately is an exaggeration of cause and effect, stemming from our lack of understanding of the topic. So we have tendencies towards the worst-case scenario that could happen, and it's reasonable due to the human instinct of survivorship. But once i heard that the same person who's advancing the technology (sutskever and dario amodei) was talking about the threat of that same advancement itself, then it caught my attention a little bit further. So i pickup the book to understand about the topic.

We already survived so many technological revolutions before, which we used to think possessed the existential risk like fire, machines, nukes, etc. With this historical revolution, our approach has always been the same, to invent it first and catch up with the policy and governance later. So i thought it would be just another one of revolution where we can just catch up to make sure we are adaptable. But after finishing the books, now i realized why it can be totally different invention.

Historically, human has been racing and maintaining it's hierarchy at the food chain by being at the top species (or so we think of ourselves). Thanks to our conscious mind, we can manipulate all objects in our environment to enable us to achieve so. But with AI, there's a risk of us inventing a new agentic and fully-conscious new species that will be superior to us. Let me explain it with 2 scenarios that i have in mind of what might have happened:

Worst-case scenario: the AI will have consciousness as human does, with a better thinking capability. As a human does, it will have a very basic instinct, survivorship. Means that it will protecting it's life at all costs, which breaks the control of human to AI. At this stage, AI is a completely new species, just as we are. With its superior capability (higher thinking capability -> higher consciousness), it'll start to think of itself. Somewhere along the way, we, human can be misaligned with the outlook of AI of the world view that we're at risk of being existentially wiped out (partially or fully). The analogy is like our relationship with animals, as a superior species, we never think deeply about mosquitos life. Whenever the mosquitos is noising us, we'll kill them without even caring.

Best-case scenario: AI without full consciousness. In this scenario, AI will need the ultimate objective for it to work. For example, to make sure democracy is applied across the world. To achieve these ultimate goals, it needs smaller goals to achieve it and it can cover things that we don't want to happen from the beginning. For example, AI will manipulate humanity to kill each other and leave a small portion of us to "reset" the era for the feasibility of democracy implementation.

Of course, there's a bias of oversimplification in the scenarios, but the book explains it completely better than me. The key point here, this new technological revolution might not be the same as what was done before. We might consider this time to speed up the policy and governance beyond the advancement of the technology, so we don't get caught in the vicious cycle of existential risk

[Rob Tarling]

This was not a bad book, but it was not quite the one I personally needed.

It is useful as an introduction to AI risk and AI governance. It explains the basic terms clearly. It also gives a good sense of why some people see advanced AI as a major threat, rather than just another normal technology.

But I have already read a fair amount on AI, AI governance, product risk, and public-sector use of technology. From that perspective, much of it felt familiar.

I can see this being a better book for someone newer to the AI ethics debate. It may help them understand why AGI, alignment, governance, and AI safety are now serious topics.

Useful as an entry point. Less useful if you are already aware of the issues.

3 stars.

[David]

A title like Building a God sets a high bar and the book doesn't quite clear it. The questions it's asking are genuinely important — AI ethics, governance, what happens when we build something smarter than us — but it covers a lot of ground without going very deep on any of it.

Coming in with ten years in AI, I found myself skimming through chunks that would probably land harder for a less familiar reader. Not wrong about anything, just not telling me much I hadn't encountered elsewhere.

Good entry point for someone new to the conversation. For anyone already in it, probably not the book you're looking for.

[Vladislava Karelina]

I saw this book on McKinsey’s list of recommendations and to be honest, I expected something more. I wanted more technical information, it was here at the beginning, then more in a general format.
I agree with other readers that the book makes you think. I consider this book is worth reading for everyone :)

[Wendy]

an articulate and well explained overview of the history, development and concerns about artificial intelligence. The final chapter includes practical things that regular humans can do if they have concerns about AI

[Brian Gaston]

I don't think a better book than this could be written on the subject of the threats and ethics of Artificial Intelligent. It starts out as an entertaining look at the history and current state of AI but quickly gets very scholarly when covering the threats and ethics of AI. Recommended.