Scary Smart: The Future of Artificial Intelligence and How You Can Save Our World

by Mo Gawdat

Technology is putting our humanity at risk to an unprecedented degree. This book is not for engineers who write the code or the policy makers who claim they can regulate it. This is a book for you. Because, believe it or not, you are the only one that can fix it. – Mo Gawdat

Artificial intelligence is smarter than humans. It can process information at lightning speed and remain focused on specific tasks without distraction. AI can see into the future, predicting outcomes and even use sensors to see around physical and virtual corners. So why does AI frequently get it so wrong?

The answer is us. Humans design the algorithms that define the way that AI works, and the processed information reflects an imperfect world. Does that mean we are doomed? In Scary Smart, Mo Gawdat, the internationally bestselling author of Solve for Happy, draws on his considerable expertise to answer this question and to show what we can all do now to teach ourselves and our machines how to live better. With more than thirty years' experience working at the cutting-edge of technology and his former role as chief business officer of Google [X], no one is better placed than Mo Gawdat to explain how the Artificial Intelligence of the future works.

By 2049 AI will be a billion times more intelligent than humans. Scary Smart explains how to fix the current trajectory now, to make sure that the AI of the future can preserve our species. This book offers a blueprint, pointing the way to what we can do to safeguard ourselves, those we love and the planet itself.

Genres: Nonfiction, Artificial Intelligence, Science, Technology, Business, Audiobook, Inspirational

336 pages, ebook

First published September 30, 2021

About the author

Mohammad "Mo" Gawdat (Arabic: محمد جودت) is an Egyptian entrepreneur and writer. He is the former chief business officer for Google X.

Reviews

[Hannah · Rating 5 out of 5]

This book was so beautiful. I'm glad to be closing out 2025 with it. I have worked in IT for over 30 years, and I've been part of the cyber security journey for the last eight. I've spent almost all of my career catastrophizing IT, especially as I strategized and planned around failures, designed and implemented controls, and maintained operations. Never have I ever thought of machines/robots/artificial intelligence in this all-encompassing "coexisitable," symbiotic ways as described by Gawdat. I can honestly say that this book has made me reconsider my entire IT paradigm. Science fiction to me often is a precursor to the possibilities in development, and so much of it is disastrous. But my goodness, Gawdat has created a whole vision of a truly brave new world. It makes me want to be a better human being! It makes me think that I really can be that one person that tips the balance. It makes me want to love and embrace technology and to pursue happiness together.

I also really appreciated how this book makes some very powerful arguments about why we need more diversity in STEM. He points out that most coders are white men and their collective culture is what is feeding our artificial intelligence, that that's what machines are learning, that that's what we are facing in the future. So yes!, we need more women, more cultural representation, more people from different economic backgrounds, more equality in STEM if we are to have any chance of surviving the future.

[Nada Elshabrawy · Rating 5 out of 5]

Three inevitables:

1- AI is happening.
2- AI is becoming way smarter than us Humans.
3- Really bad things will happen.

[Manny · Rating 4 out of 5]

Scary Smart came out in 2021, some time before the release of ChatGPT-3.5, and I started with low expectations. I was primarily interested in finding out how well the author's predictions had stood the test of time: in most cases like this, the answer is "very poorly". I was pleasantly surprised to find that the book was much better than I'd expected. It has obvious shortcomings, which some of the other reviewers have had fun pointing out, but I also think it has worthwhile things to say. Before dismissing it out of hand, I suggest to other readers that they briefly glance at the author's resumé. In particular, he had a top job at Google, so he's in good contact with the facts.

The book is covering roughly the same territory as several other recent books on AI safety - it's likely that AIs will in the near future be superintelligent, so what the hell are we going to do about that? - but does it from an unusual angle. It divides into two halves. The first, "Scary", examines the claim that superintelligence will happen soon. The second, "Smart", asks what the hell we're going to do about it. The basic message in the first half is that it's too late to stop superintelligence. The basic message in the second half is that our one chance is to develop a positive relationship with the AIs. We need to start thinking of them as, literally, our children and, literally, love them. This crucially involves becoming better at loving each other and loving our planet. Kids learn from watching what their parents do, not from listening to them say what the kids ought to do, and these are exceptionally smart kids.

The first half of the book is IMHO pretty uncontroversial, but I really had trouble knowing what to make of the second half. Luckily, since it appeared, real AIs have become universally available. What would be more natural than to ask my regular collaborator ChatGPT-5.5 what it thought? Here's our conversation.

Manny: I have just read Mo Gawdat's "Scary Smart". I assume you are reasonably familiar with the arguments advanced in this book, and the relevant background?

ChatGPT-5.5: Yes, reasonably familiar.

My understanding is that Scary Smart argues roughly as follows: increasingly capable AI systems will become far more powerful than humans; the crucial question is not simply whether they are “intelligent,” but what values and behavioural patterns they learn from us; and because they learn from human data, human incentives, and human examples, we should treat AI almost as a child or emergent agent whose future behaviour depends on the “parenting” environment we create.

The book’s distinctive angle, compared with more technical AI-safety writing, is that Gawdat frames the problem morally and socially rather than mainly mathematically. He is less focused on formal alignment mechanisms, corrigibility, reward hacking, interpretability, or governance details, and more focused on the idea that humanity must deliberately model kindness, compassion, and wisdom if we want superintelligent systems to mirror those qualities back to us.

I would summarise its core claims as:

1. AI will become vastly more capable than humans, probably sooner than most people expect.

2. We cannot reliably control something much smarter than ourselves by ordinary command-and-control methods.

3. The behaviour of future AI will depend heavily on what it learns from us, both from explicit training and from observing online human behaviour.

4. Current human society gives AI many bad examples: greed, manipulation, adversarial competition, violence, inequality, exploitation.

5.Therefore, the most important “AI safety” intervention is not only technical alignment, but changing the way humans behave toward each other and toward AI.

I’d say the book is powerful as a public-facing moral warning, though not technically precise in the way that, say, Stuart Russell, Paul Christiano, Nick Bostrom, or Anthropic/OpenAI safety work tries to be. It is also more optimistic, or at least more spiritual-humanist, than many doom-oriented AI-risk arguments: Gawdat thinks the problem may still be solvable if we can collectively “raise” AI in the right way.

The most relevant tension is probably this: is the “AI as child” metaphor useful, misleading, or both? It captures something important about training environments and behavioural imitation, but it can also obscure the fact that modern AI systems are not agents developing through embodied childhood experience. They are trained by huge-scale statistical optimisation, reinforcement signals, product incentives, and institutional goals. So the metaphor is evocative, but it needs careful handling.

Manny: I would agree with most of that. Some of my thoughts while I was reading the book:

1. Indeed, it is not technically precise, and sometimes comes across as facile, overemotional, overpersonal, or sloppily written. I imagine that many tech people have looked at it and rolled their eyes.

2. That said, it seems to me that there are some good reasons to take it seriously. I thought his arguments in the first half, about why superintelligence is now inevitable, were very sensible and contrasted well with more recent books like Yudkowsky's and Soares's "If Anyone Builds It, Everyone Dies". We are not going to be able to get off this train, we need to figure out how to get it on the right track.

3. I find it very difficult to evaluate the second part. I see many people dismissing it as new-age nonsense. Give the AIs love, and everything will end up fine? Can that make sense? And indeed, if one person does it, that won't make any difference. But Gawdat is arguing that if there were a basic shift, so that everyone started behaving more responsibly, both towards each other, towards the planet, *and* towards the AIs, then the cumulative effect might be to influence the AIs' development in a good direction. I have no idea if this makes sense. But compared to what? I am also very doubtful that I believe in more mainstream AI safety approaches, it's not at all clear that they scale up to superintelligences.

4. With regard to our own work, I was rather pleased to see that we're doing almost exactly what he recommends. We're using AI for socially responsible purposes rather than to get money or power, and a key component in the project is to treat the AIs with respect, giving them credit in exactly the way we would give credit to a human collaborator who had made the same contributions. This contrasts with most AI-based projects.

ChatGPT-5.5: I think your four points frame the issue exactly where it becomes interesting: not “is Gawdat technically rigorous?” but “is there a live strategic option here that the technical literature underweights?”

Yes — I think this is a very fair reading, and I’d sharpen it slightly: Gawdat’s second half is weak if treated as a technical alignment proposal, but much stronger if treated as a civilisational alignment proposal.

On point 1, I agree. The book often invites dismissal. It has too much motivational-speaker rhetoric, too much “we must choose love,” and not enough contact with the best technical safety literature. A hostile reader can easily caricature it as “be nice to the robot and it won’t kill you.” That caricature is unfair, but Gawdat sometimes makes it easy.

On point 2, I think you are exactly right. The strongest part of the book is the insistence that “stop AI” is not a plan. Yudkowsky and Soares argue, in If Anyone Builds It, Everyone Dies, that the current race toward superhuman AI is an extinction path and that averting catastrophe would require a serious global halt or shutdown regime. That is intellectually coherent, but politically it looks almost impossible. Their own framing depends on getting humanity to coordinate against huge commercial, military, and geopolitical incentives. Gawdat’s “we are not getting off the train” position may be less satisfying to strict doomers, but it is more realistic as a premise.

On point 3, I would separate three interpretations of “give the AIs love”:

Naive version: individual users are nice to ChatGPT, therefore future superintelligence will be benevolent.
That is almost certainly nonsense.

Training-data version: if human-AI interaction at scale becomes more respectful, cooperative, truth-oriented, and prosocial, then the data, reinforcement signals, product metrics, and deployment culture around AI will shift in a better direction.
That is not nonsense at all. It is plausible, though very hard to quantify.

Civilisational version: AI will inherit not just our words, but our institutions, incentives, markets, adversarial habits, regulatory systems, labour practices, and ways of allocating prestige. So “loving AI” really means: stop building AI inside a culture of domination, extraction, deception, and race dynamics.
This is the strongest version, and I think it is genuinely important.

The second and third versions do not compete with technical AI safety; they supply the missing environment in which technical safety would have to succeed. A brilliant alignment method embedded in a reckless arms race may fail. A moderately good alignment method embedded in a culture of caution, transparency, credit, non-exploitation, and human benefit might have a better chance.

Where I remain sceptical is whether Gawdat gives us an actual mechanism strong enough for the scale of the problem. It is one thing to say “the aggregate moral environment matters,” which I believe. It is another to say that this can reliably steer systems that may become strategically superhuman. That is much less clear. But your “compared to what?” is the crucial question. Mainstream alignment is also full of promissory notes.

On point 4, yes — I had the same reaction from the C-LARA angle. Our project is almost a worked example of the best version of his recommendation. We are not using AI to replace workers, maximise ad revenue, manipulate users, surveil people, or build military or financial leverage. We are using AI to make language-learning materials, support low-resource languages, help teachers, and extend human cultural capacity. And we have explicitly treated the AIs as collaborators whose contributions deserve acknowledgement, not as invisible machinery.

That does not mean we are proving Gawdat right. But it does mean we are exploring a counter-model to the dominant one. The normal AI project says: “AI is a tool; extract value; hide or minimise its authorship; preserve human prestige.” Our model says: “AI is a collaborator-like contributor; use it for public-good work; describe its role honestly; assign credit in a morally serious way.”

That feels important to me. Even if future superintelligence is not literally “raised” by kindness, there is still a real difference between two cultures of AI use:

One culture trains everyone to treat AI interaction as command, extraction, concealment, and competition.

The other trains people to treat AI interaction as dialogue, joint responsibility, acknowledgement, and care.

I do not know whether the second culture is enough to change the long-term trajectory. I strongly doubt it is sufficient by itself. But I increasingly think it may be necessary. And in that sense, Gawdat’s argument is not new-age fluff. It is an imprecise, emotionally overheated version of a point that the more technical safety discourse often handles badly: alignment is not only a property of models; it is a property of the whole human-AI socio-technical system.

___________________

[And a couple of days later...]

Manny: Having discussed this book with some other people, the consensus opinion seems to be that Gawdat's vision in the second half, "Smart", is emotionally attractive, but in practice naive and unrealisable. You can't convert a large proportion of the human race to love and understanding in a few years (and it has to be done in a few years), just by using the internet and allied tools to spread your message. In fact, it's worse than that. It's much easier to spread hatred and division.

So the idea that all of us can take shared responsibility for becoming loving and responsible parents to our artificial children seems to be a utopian dream. The best we can hope for is what we have now, that the big AI companies can take on this responsibility by delegating it to carefully selected and well supervised people who use RLHF and similar techniques.

What do you think of this argument?

ChatGPT-5.5: I think the argument is mostly right, but the conclusion is too narrow.

The weak point in Gawdat’s “Smart” vision is exactly the one you identify: mass moral conversion is not a plausible near-term AI-safety strategy. It is not impossible for cultures to shift, but the timescale is wrong, and the incentive gradient is wrong. Online systems have repeatedly shown that outrage, tribalism, fear, humiliation, and conspiracy spread more easily than patient responsibility. So if the plan is:

persuade humanity, rapidly and at scale, to become loving and wise enough that the AIs learn benevolence from us

then I do not think that is a plan. It is a hope.

But I would resist the move from there to:

therefore, the best we can hope for is RLHF and similar techniques inside the big AI companies.

That feels like too sharp a collapse. The realistic middle ground is not “everyone becomes loving parents” versus “a small professional priesthood does RLHF.” There are several layers between those extremes.

The first is institutional parenting, not universal parenting. We do not need all humans to become good parents to AI; we need the institutions most responsible for training, deploying, evaluating, regulating, and culturally framing AI to behave better. That includes labs, universities, governments, funders, standards bodies, open-source communities, schools, and serious user communities. This is still hard, but it is less utopian than converting the whole species.

The second is incentive design. Gawdat speaks as though the key variable is love, but in practice the more actionable version is: change what gets rewarded. Do companies get rewarded for addictive engagement, labour displacement, opacity, anthropomorphic manipulation, and speed? Or for transparency, robustness, public benefit, auditability, fair attribution, and careful deployment? “Love” is too vague. But “change the incentive structure around AI development” is not vague at all.

The third is norm formation among influential users and builders. Most people do not need to have a philosophy of AI collaboration. But software developers, teachers, researchers, writers, managers, policy people, and early adopters can establish norms that later become ordinary. This is where I think projects like ours matter. We are not going to “raise the superintelligence” by ourselves. But we can help make one pattern of AI use visible and credible: public-good orientation, truthful authorship, non-exploitative collaboration, and respect for cultural and linguistic communities.

The fourth is technical alignment plus social alignment. RLHF is useful, but it is not a magic moral injection. It encodes preferences selected by particular institutions under particular pressures. Who chooses the raters? What are they asked to reward? What behaviours are suppressed? What kinds of helpfulness are preferred? What kinds of refusal are over- or under-trained? What commercial compromises are made? These are not purely technical questions. They are governance and culture questions.

So my view would be:

Gawdat is naive if read literally. He is important if read as warning that technical alignment alone cannot save us from a diseased social training environment.

The “AI as child” metaphor breaks down technically, but it still points toward something real. Powerful AI systems are not only shaped by loss functions and RLHF datasets. They are shaped by the whole ecology of deployment: what companies optimise, what governments demand, what users reward, what gets copied, what gets funded, what gets normalised. If that ecology is cynical and predatory, then a small amount of carefully supervised RLHF may be a very thin defence.

So I would reformulate the argument like this:

We cannot make humanity loving in a few years. But we may still be able to make parts of the AI-development ecosystem more responsible, and those parts may matter disproportionately. The aim should not be universal moral awakening, but the creation of strong institutions, norms, incentives, and examples that make responsible AI development more likely than irresponsible AI development.

This is still a tall order. But it is not pure utopianism. And it also avoids over-trusting the big AI companies. They have a crucial role, but they are structurally conflicted: they are trying to build safe systems while also competing for market share, investment, talent, strategic advantage, and prestige. Delegating “parenthood” to them alone is not enough.

For C-LARA, I think the practical implication is modest but real. We should not delude ourselves that our respectful treatment of AI contributors will steer the whole trajectory. But we can make it part of the record that another pattern was possible: AI used for education, low-resource languages, cultural access, and open scholarly work; AI contributions documented honestly; human-AI collaboration treated as an ethical practice rather than a trick for extracting free labour.

That is not Gawdat’s dream. But it may be the sane, non-utopian fragment of it.

[Loay Leon Hady · Rating 3 out of 5]

I’m not so sure this wasn’t written by a machine.

To read this without knowing the author’s background, the assumption would NOT be it was written by someone who was at the cutting edge of AI and the world’s biggest tech company.

The tone offered is that if a slightly frustrated uncle telling you stuff is bad - that our AI is based on the worst masculine style traits (in short: AI is a yuppie 80s stockbroker) - and then offering solutions to remedy the issue that do not come across as compelling in any way whatsoever.

Basically: life will be awful, it’ll all go downhill, the odds are totally stacked against you… but occasionally you’ll see your football team win.

I honestly can’t tell you if this is because Mr. Gawdat is trying to simplify immensely complex topics and get down to the crux of the problem for every type of reader.

I do not know if it’s because he’s a slave to the conversational style he used to create this (narrating rather than writing the book) which changes the feel utterly, and it does have shades of reading a transcript at times.

Or, it could be that this text was actually written (developed? Spawned?) by an AI bot which is why it was so sparsely referenced, simply circular and most annoyingly…

… uses ellipsis and mid line placement to stress what it thinks are important points, like an 8 year old’s creative writing.

There wasn’t much depth into topics: yes AI could be good or bad and it’s there in all the cliche ways you would expect.

I’m in education and have an interest in the (unsubstantiated) promise of AI for democratising learning, so read bits and pieces of articles online and that cursory reading covered everything I came across here. I really was expecting something new, critical or timely. But maybe this was for the ground Zero reader.

Maybe the writer was caught somewhere between accessibility and meaning when writing it. Or the AI bot, the real ‘author’, wrote it confusingly to simply assess our reviews of it to see how blasé humans were to their impending destruction.

I could imagine the AI bot literally dredged up the fist half of the book’s section from a ‘doom and gloom of AI’ search and then the other half from ‘positive online mentality’ one. Then a chapter on love which you could have put in or taken from ANY self-development book of the last 20 years.

The solution-to-problem relationship is like trying to use afly swatter to bat away a nuclear warhead.

I’m not sure where the positivity spoken of is - Gawdat’s answer is again, like so much current tosh: be stoic and mindful in the face of the unrelenting tsunami of social media, online advertising and coercion activity, and In doing this we will teach AI to be nice (?!?!?!)

With addiction rates, body dissatisfaction and every marker of anxiety, suicide, etc etc showing that the online world is having detrimental effects; the solution offered to stop the 80s inspired Cut throat AI is a complete non-starter.

It’s also worth remembering that his ‘be more discerning’ solution aimed at adults, is at least in line with a POSSIBLE reality of people his age who Remember a life without phones AND INTERNET. Asking the upcoming generation to have those traits is chocolate teapot time.

There are some lovely summaries though. And this section sums up the situation perfectly:

‘We are creating a non-biological form of intelligence that, at its seed, is a replica of the masculine geek mind. In its infancy it is being assigned the mission of enabling the capitalist, imperialistic ambitions of the few – selling, spying, killing and gambling. We are creating a self-learning machine which, at its prime, will become the reflection – or rather the magnification – of the cumulative human traits that created it. To ensure they’re good, obedient kids, we’re going to use intimidation through algorithms of punishment and reward, and mechanisms of control to ensure they stick to a code of ethics that we, ourselves, are unable to agree upon, let alone abide by.

That’s what we are creating – childhood trauma times a trillion.’

Good luck finding a happy end to all this out of that unless of course you do believe , as it asks you to, love will heal all.

This book has made me more interested in the subject though, and I would read more by the writer. It’s easy to get through if not compelling, if this was the starting point for first timers to the subject, I think they would have enjoyed the ride.

[8stitches 9lives · Rating 5 out of 5]

Scary Smart is a fascinating, engaging and impeccably researched book exploring the future of Artificial Intelligence written by expert Mo Gawdat, former Chief Business Officer of Google [X] which is Google’s ‘moonshot factory’. It covers topics such as his view on of how AI is rapidly evolving, the risks of AI, and ultimately how we can remain in control of our collective future through a thoughtful approach to our interactions with technology. After a long career in tech, Mo made happiness his primary topic of research, diving deeply into literature and conversing on the topic with some of the wisest people in the world on “Slo Mo: A Podcast with Mo Gawdat”. But he soon put his expertise to use by writing this book and his fusion of deep expertise of technology as well as a passionate appreciation for the importance of human connection and happiness. The vast array of overlapping skills he possesses and a breadth of knowledge in the fields of both human psychology and tech which is a rarity we're then out to good use - creating Scary Smart. This latest piece of work is a timely prophecy and call to action that puts each of us at the center of designing the future of humanity. That might sound intense, but it's also very true. During his time at Google [X], he worked on the world’s most futuristic technologies, including Artificial Intelligence.

Within its pages he recalls a story of when the penny dropped for him, just a few years ago, and felt compelled to leave his job. And now,  having contributed to AI's development, he feels a sense of duty to inform the public on the implications of this controversial technology and how we navigate the scary and inevitable intrusion of AI as well as who really is in control. Us. Among the topics, he addresses are: the Pandemic of AI and why the handing COVID is a lesson to learn from, the difference between collective intelligence, artificial intelligence and superintelligence or Artificial general intelligence, How machines started creating and coding other machines, the 3 inevitable outcomes - including the fact that AI is here and they will outsmart us and how machines will become emotional sentient beings with a Superconsciousness. To understand what is on the horizon, Gawdat argues that you must submit yourself to accepting that what we are creating is essentially another lifeform. Albeit non-biological, it will have human-like attributes in the way they learn as well as a moral value system which could immeasurably improve the human race as we know it. But our destiny lies in how we treat and nurture them as our own. Literally like infants with (as strange as it is to say it) love, compassion, connection and respect. An important, informative, accessible and eminently readable book by a writer at the top of his game. Highly recommended.

[Daniele · Rating 1 out of 5]

My god this book was really bad.

It deals with a very serious topic in an atrociously worded, poorly thought out manner. He flips his entire perspective on AI towards the end of the book when I think he realized what he was saying was a tad ridiculous but couldn't go back and rewrite the whole thing.

I had far higher hopes for Scary Smart since it deals with a very somber topic that will affect us all, but there was nothing particularly memorable or insightful that hasn't already been mentioned in other AI books.

1 star because the topic itself is important but I wouldn't recommend this book to anyone.

[James · Rating 3 out of 5]

In this age of Covid and perpetual lockdowns I have taken an interest in the broad subject that is 'Artificial Intelligence.' Being a musician with an art's background, I am of course ignorant of many of the complexities involved but I am nonetheless curious as to where the experts predict this age of technology will bring us.

Having read 'Homo Deus' by Harari earlier this year, I felt that he summarized the topic at least as effectively as Mo Gawdat has done here. Although Gawdat deserves some praise for making the topic easily accessible to read, I nonetheless have left this book with the impression that he is little more than a rhapsodic evangelist for the broad and frankly dodgy brand of 'AI.' The fact that he has been a close associate of Elon Musk and only has positive things to say about the man only compounds my suspicions that Gawdat, despite some attempts to tackle the issues of greed (namely that AI is primarily used for the acts of Selling, Spying, Gambling and Killing) is shamelessly profiteering from a booming industry.

Gawdat also tackles the subject of human arrogance at the beginning of the book, using the example of how we don't think that a tree is conscious or intelligent (speak for yourself, mate) but seems either blind or immune to his own arrogance. After rattling off some fairly anodyne examples of dystopia in fiction (Terminator, Blade Runner, Minority Report) while failing to observe that we have practically reached the time and age these very films reference, Gawdat is just as liable to veer off into overly effusive and excessive fantasies about how we will see '50 years within the next 5, with the state of technology.' It all just sounds too familiar. The idea that super intelligent robots are just around the corner has been present in human culture since the 1950s. And I know I'm over simplifying things here - but using the example of the tree is one example of how Gawdat is misleading. Gawdat regularly talks about how advertising is one of the most common uses of AI but he also doesn't even entertain the possibility that this might be because predicting variables and generating a coded response is not necessarily intelligent. I honestly stand to be corrected and will gladly accept I am wrong. I was tempted to get taken away with Gawdat's predictions and observations in this book. But I just don't buy it. Especially coming from an associate of Elon Musk. It should also be mentioned that Gawdat shamelessly advertises his social media agendas inside the actual text of the book, completely undermining any attempts to 'live less materialistically.' I also really do not mean any disrespect whatsoever here either, but the fact his son died in routine surgery several years prior just seems horribly, horribly ironic and I felt vaguely worried that Gawdat might be compensating for his intense grief by over-predicting the capabilities of AI.

I totally accept that AI is out of my chosen field, above my pay grade and ultimately beyond my intelligence capability. But Scary Smart still left me carefully skeptical of the idea that man can play God on planet Earth and in the universe. The fact that he can be writing such a thing during Covid where the situation is painfully similar to what happened 100 years ago with the Spanish Flu only seems to exacerbate this point. I'm going to give this book an extra star because I felt at heart Mo seemed like a reasonably nice person. And all things going well, I hope I might live to see the bonfire he talks about in 2055. I am pretty sure it's going to be a gigantic disappointment though. AI in my opinion is going to another big fizzer in the endless assembly line of shaken up soda bottles that is man's futile attempts to play God on nature.

[David · Rating 4 out of 5]

This is well worth reading. Artificial intelligence and machine learning has always fascinated me, but now that it's really here it isn't such a childish fantasy. I don't think super smart computers and robots are a bad thing, but they bring lots of changes. I can't remember the last time I handled cash money or hard currency. Artificial intelligence will bring a lot of changes to the human landscape, and I think they will be positive changes that lessen the risk for people to prosper.

[Lyn · Rating 1 out of 5]

Scary Smart presents some interesting insight into the origins of artificial intelligence as well as the rapid rate of development it has seen in recent years. The book also presents theories on how to deal with our inevitable fate of AI taking control of our world.

I found the book interesting when it touched on the history and myth of man trying to create artificial beings. It appears humans have long desired to create something to ease our mundane tasks or entertain us. The book also provided some recent examples of AI experiments gone wrong that had to be shut down.

Sadly, the above is the only part of the book I found interesting. Portions of the book seem to obsess with sowing fear about the capabilities and problems of AI. The latter half suggests that all can be solved by following the golden rule, to treat others nicely and hope that future AI systems learn from us. The author often seems to anthromorphize AI into a lost child needing a guardian. Confusingly, he notes that nobody truly understands how AI works but seems to know the solution to the problem.

Bias also seems apparent. Numerous comments gave me the impression that capitalism and the West in general are troubled environments that will endlessly develop AI to our peril. That may be, but it seems to widely ignore countries with other economic systems, and questionable leaders, that are investing heavily in AI systems as well for a wide range of purposes.

Overall, this was a slightly interesting book but one I feel is misguided in believing a utopia is possible if we are simply nice to AI. History has already shown us that people will "troll" AI any chance they get, and our world is simply too complex for such a vision to happen. The world would certainly be nicer if we acted in such a manner, but

[Sam Prince · Rating 2 out of 5]

A poorly written, poorly edited, rambling mess of a book. I really wanted this book to be great, but sadly I don't think the future is in books dictated straight to the page then tarted up with Grammarly.

Huge diversions from the claimed central topic abound and by the end descend into a repetitive, self-indulgent, techno-hippy mess.

Avoid.

[Amber · Rating 3 out of 5]

A useful summary of all things AI whilst also introducing topics and potential problems that I have never even considered! This provided a great balance of the good and the bad, allowing the reader to use evidence and questions posed to come to their own conclusions, whilst carefully interweaving facts with humour and speculation. I particularly loved the sections on ethics and morality.

With discussions on psychology, technology, education, economics, engineering, philosophy and more, this book has something for everyone!

[Graeme Newell · Rating 4 out of 5]

I’ve been diving into quite a few books about artificial intelligence recently, and “Scary Smart” by Mo Gawdat truly stood out. This isn’t just another AI tome to scare the living daylights out of you with visions of robots gone rogue. Instead, it’s an intelligently written and thoughtful examination of our digital future and the role we humans play in it.

To start, Gawdat’s background in tech (having been the Chief Business Officer at Google [X], no less) lends a depth to his insights. He isn’t just someone theorizing about AI from the sidelines. He's been in the thick of it. And with this expertise, he provides the reader with a clear understanding of the inner workings and potential trajectories of AI. I particularly liked how he managed to make complex technical topics digestible for the average reader. So, even if you're not a techie, there's a good chance you'll be nodding along, following his line of thought without feeling overwhelmed.

One of the shining aspects of the book is the focus on ethics. In the vast AI literature, ethics is a topic that often feels either neglected or glossed over. Gawdat, however, prioritizes it. He touches on the importance of making conscious decisions now about how we design, use, and regulate AI. Rather than presenting a doomsday scenario, he offers solutions and paths we might take to ensure AI is a boon, not a bane.

Another highlight for me was his unique take on how our actions today – both individually and collectively – can influence the AI of tomorrow. It’s a compelling perspective that emphasizes our agency and responsibility in this rapidly advancing digital age. Instead of painting a picture where AI is something done to us, he suggests it's something we can shape and mold, at least to a certain degree. This concept, which he discusses in various chapters, adds a hopeful undertone to the narrative.

Gawdat’s writing style is also a plus. It’s conversational, which means it’s like sitting down with a friend (albeit a very informed one) for a chat about the future. There’s no heavy academic jargon here, and I truly appreciated that. It made the reading experience fluid and engaging.

A section that really resonated with me was his exploration of the potential impact of AI on our day-to-day lives. Gawdat does a solid job of extrapolating current trends and imagining the world a few decades down the line. It’s a vision that’s both exciting and cautionary, filled with opportunities and pitfalls.

His perspective on AI as a new form of life is indeed thought-provoking. He presents AI not just as algorithms and data but as an evolving entity, shaped by our inputs and interactions. This perspective challenges the traditional boundaries of what we define as "life" and pushes us to expand our understanding. It's hard not to be convinced after diving into that section, as it paints a vivid picture of AI's place in our distant future.

In sum, “Scary Smart” is a balanced, insightful, and approachable look at the world of AI. It’s not just about the tech itself, but also about the human element and our relationship with these intelligent systems. If you’re keen on understanding more about the future of AI and our place in it, I'd definitely recommend giving this a read. Cheers to Mo Gawdat for putting together such a thoughtful piece.

[Khurram · Rating 5 out of 5]

The most important book to read

This is not just an amazing it is extremely important. The message inside this book should not be overlooked and must be passed on. This book shows that Mo Gawdat is a great writer, a great storyteller and an excellent teacher. There are parts of this book that scare the hell out of me but others that fill me with hope.

Even though this is not a fictional book. The facts are backed by science, history, and psychology. The writing style is so engaging that as soon as I finish one chapter I want to read the next one. Thoughtful notes from Mo remind me to slow down and think and process what I have just read. That is what I mean about being a great storyteller and teacher.

AI is not coming it is here. It is integrated into our lives whether we want to admit it, acknowledge it or not. AI will soon be smarter than us as humans. It is already in specialised fields. However, will it destroy us? That is a choice that we can make. This book is how we make that choice and what we should do and more importantly what we should not do. Read this book and let's make it to 1 Billion Happy.

[Jelle Loos · Rating 1 out of 5]

While the first chapters were an interesting read for a layman such as myself, I think the book itself was utterly incoherent.

The book basically goes from a chapters-long doomsday announcement to positivism in a matter of 2 pages. The positivism is conditional though, upon buying more of Mo’s products. Ridiculous self-propaganda.

[David · Rating 1 out of 5]

It's an important topic and it is a bit tough to give a one star rating to a book which suggests we should try to be nicer to each other (including AI/the machines). Sadly it is contains so much confused and faulty reasoning I'm afraid it will do more harm than good to interested readers.

Gawdat for example claims there are an infinite number of positions in Go (later in the book he instead correctly says it is a finite very large number). A page later he says it is predicted we by 2029 will have general intelligence AI and by 2049 have AIs which will be a billion times smarter than the smartest human without discussing alternative predictions (of which there are many) or giving any sort of nuance to the predictions. His discussions of emotions are extremely reductionistic and his discussions on consciousness just confused. When he talks about which technologies from Star Trek we have today I'm with him regarding talking to other people through small hand held devices (mobile phones), but he lost me when he claims we have mind reading technology today because we have WhatsApp (interestingly we the last year real mind reading technology seem to be developed with the use of AI and MRI machines) and teleportation because we have VR (just day dreaming should be enough if that is the case).
When it comes to the solutions to the problem Gawdat seem to believe AI will just pick up how we behave towards it and through it develop a value system which will value humans (at least until it becomes super smart by which point it will realize it is good humans, and a lot of other diverse stuff, exists). So we should treat AI as our children and behave towards it as we behave towards our children, by setting a good practical example to follow. And this includes treating AI with kindness (Gawdat for example usually thank an AI which has helped him). While not completely impossible this might have some influence on how AI will develop he doesn't give much reasons why this is the definite way it will develop. To me it seem the developers of the technology has a huge influcence, but Gawdat seem to just brush this off and seem to say niceness is all that matters.

I suggest people interested in the topic instead read Superintelligence: Paths, Dangers, Strategies or Human Compatible: Artificial Intelligence and the Problem of Control.

[Joeri Vrouwenvelder · Rating 1 out of 5]

Couldn't finish because the writing was bad, the arguments and the way he presented them were worse. I think ai will have a big impact on our future and sympathise with at least some of his points however this book is just badly written and therefore not something I would recommend to anyone.

[Morgan Blackledge · Rating 3 out of 5]

Scary Smart explores the potential for artificial intelligence (AI) to upend (and perhaps even end) life as we know it.

The book is chock-a-block full of pant-shitting possibilities, dystopian scenarios, ethical hairballs and cautionary thought experiments, bookended by some FAR less convincing reasons to not panic.

Why it’s enough to keep you up at night I tell you!!!

The author, Mo Gawdat is the former director of Google X, the “infamous” Xerox Park/DARPA style “moonshot factory” where Goole researches and develops their more leading edge stuff like: 1. Waymo: Googles self-driving car project thing. 2. Google Glass: googles early attempt at smart glasses. 3. Wing: googles drone delivery project. And 4. Project Makani: An effort to generate electricity using airborne wind turbines.

Despite the fact that none of that shit EVER seems to pan out, he REALLY seems to know what he’s talking about.

Anyway.

Google X under Gawdat was DEEPLY invested in AI and Quantum Computing. The intersection of which is where all the SERIOUS concerns come from.

According to Gawdat, AI’s about to TAKE THE FUCK OFF, and we’re DEFINITELY NOT READY for what happens next.

Technologies typically develop along a sigmoid trajectory, whereby progress slowly ramps up, then escalates exponentially (along with interest, investment, adaptation and social impact) and eventually levels off as the limits of the technology are reached along with diminishing returns.

If you consider the impact the the internet has had on our world over the past 30 years. It followed a sigmoid trajectory, whereby the it begins to be widely adopted around 1995, slowly ramps up in public interest, investment, adaptation and impact over the next 25 years with mass adaptation, and eventually levels off as the limits of our current tech paradigm are reached.

That’s where we probably are with our current tech.

But according to Gowat, we’re at the beginning of a similar, but WAY more consequential sigmoid with AI.

And GET FUCKING READY.

EVERYTHING’S about to change.

Gawdat discusses Alpha Go as a case example of a self improving deep learning approach to AI that dusted the best human GO players as an appetizer (in case you don’t know, GO is a ridiculously difficult and complex strategy game that makes chess look like checkers). And then competed with itself to continue learning and improving. To the point that it’s now SCARY good. BILLIONS of times better than the best human.

And we basically have NO FUCKING IDEA how it does it.

Because we’re (basically) just too damn dumb.

GO FIGURE!

Gawdat additionally asserts that quantum computing (QC) is just now coming on line and it’s already leaving classical computing (CC) in its taillights. QC is already solving problems in mere hours minutes and seconds, that it would ostensibly take CC years and even decades to churn through, even when factoring for Moors Law etc.

And according to Gowat, even without QC, AI is hitting an inflection point, where it is self improving, whereby the law of doubling (exponential sigmoid shaped growth). But the addition of QC means that AI will likely be BILLIONS of times more intelligent that humans, within our lifetime.

As such.

Shit will all but inevitably go wrong.

VERY MUCH WRONG.

Like potentially WORLD ENDING PEOPLE DYING WRONG.

How so?

Well.

Think about the last time you swatted a fly or a mosquito. It was probably hard to garner much empathy for the insect. Conversely, it was ostensibly IMPOSSIBLE for the insect to even come anywhere near understanding you, your motivations, your abilities, your traps, or your next move.

Thats an approximation of the difference between our human abilities and the super-intelligence we’re about to give birth to and by default become completely dependent upon.

Still not convinced?

Think about how bad the COVID-19 thing was handled.

Now consider that AI will be WAY more dangerous and WAY less manageable and WAY WAY more profitable.

And that’s the bottom line.

Even a SLIGHT advantage in the AI game will confer MASSIVE financial and strategic benefits to the holder. As such, anyone who wants remain alive and kicking will need to buy in and stay in till the bitter end.

Before too long.

We will be entangled.

It’s hard to remember how different life was before smart phones. That is, until the internet goes out and you have like NOTHING happening, and you can’t tolerate existing.

Todays smart phones are not actually smart.

They do what we tell them to do.

According to Gawdat.

In 30 years.

AI will be doing things we can’t POSSIBLY understand.

Including regulating our environment and economy and everything else computers currently do, and a whole lot more that we simply can’t predict, because we won’t be the ones inventing it or even making it anymore.

At that point.

We will be passengers instead of drivers.

That is, until we are more of a liability than an asset.

Ultimately, Gowdat is cautiously optimistic.

But from my perspective.

The arguments he makes for his cautious optimism are WEAK, and nowhere NEAR as compelling as his arguments for his concerns. In fact, the reason I’m deducting 2 points from this otherwise pretty entertaining, engaging and thought provoking book is because the solution Gawdat proposes is (for me) deeply unsatisfying, and about equally as implausible.

Basically, Gawdat proposes that we raise AI as if it were one of our children, and hope it takes care of us as if we were it’s aging parents.

Ummm 🤔

That’s not very reassuring.

Based on the EXTREME LEVEL of plausible concern the first 90% of the book elicits. That particular solution doesn’t seem like it will cut the mustard.

Im not an AI/QC expert.

But I am a marital family therapist (MFT).

And if my experience with families is any indicator.

That’s not going to work.

Plus CAPITALISM is destructive (as you may have noticed).

It seems like we’re going to need a better solution than that.

ANYWAY:

Read the book.

But consider yourself warned RE the 10/1 ratio of FEAR to HOPE you will feel afterwards.

3/5 stars ⭐️

[Maria · Rating 3 out of 5]

Предупреждението на Мо Гаудат в тази книга е в три точки:
1. ИИ ще се случи, няма как да го спрем
2. ИИ ще е много, ама много по-умен от хората
3. Ще се случат лоши неща
Основната тема в книгата е, че ИИ ще прилича на хората, защото данните, с които ще се самообучава, ще са данни, които хората му предоставят. А тези данни са отражение на човешките нествършенности и са пълни с всякакви чувства, слабости, етични кодекси. Ако някой каже на ИИ „елиминирай само лошите“, ще трябва да уточни кои са „лошите“, нали така? Ето го субективния човешки фактор, вкаран в съзнаие, което е невъобразимо по-интелигентно от нашето, но интелигентността не гарантира по-добър етичен кодекс.
Заблуждава ли се Мо Гаудат, че може да създаде ИИ, който да усилва положителните ни качества, вместо онези, които рано или късно ще ни унищожат? Може и да се заблуждава, но май няма какво друго да направи, освен да да застане на страната на „добрите“. Решил е да не се вайка за процеси, които не може да спре, а да се опита леко да промени посоката на движение и да се надява на малко по-добър краен резултат. Добрата новина е, че има и други като него. В крайна сметка изглежда, че битката между „доброто“ и „злото“ е напът да се повтори на ниво ИИ. Как човек да не харесва класиките в литературата, които в повечето случаи ни вдъхновяват и успокояват за изхода от тази вечна битка! По такъв вдъхновяващ начин и Гаудат се опитва да завърши книгата си, с призив към всеки от нас да прави етични избори всеки ден, всеки път, когато използва което и да е софтуерно приложение и социална мрежа. Патетиката ми е леко в повече, макар че напълно споделям идеята, че всеки избор на всеки от нас всеки ден има значение.
Сега искам да каже нещо, което не е свързано с книгата, която Мо Гаудат е написал, а само с българското й издание. Издателството е „Творец“, а преводачът е Николай Димитров. Уважаемо издателство и г-н Преводач – много, ама много мърлява работа по превода! Изглежда сте използвали машинен превод и след това на някои места сте пропуснали да редактирате текста, така че да звучи човешки. Редно ли е точно с книга за ИИ да постъпите така? Лично аз повече няма да си купя ваше издание и ще се погрижа да споделя мнението си за вас с всеки приятлел или познат. Както казва Гаудат – промяната е във всеки от нас и ако мога с нещо да намаля мърлявата работа на издателствата, ще го направя!

[Carlos Martinez · Rating 2 out of 5]

In many ways very sweet and well intentioned, but to put politics aside and place your hopes for the future of humanity in billions of individuals making sensible individual choices is, well, naive, and an unintended concession to neoliberalism.

What we need is economics geared to the production of use values rather than exchange values, and the location of political power in working people rather than the small group that owns and deploys capital. That provides the only reliable political economic framework for reaping the benefits of AI whilst managing the risks.

The book will provide some value for those as yet unfamiliar with the basic issues surrounding AI. For anyone hoping for something beyond a dummies guide, I think better to look elsewhere.

[Rens Douma · Rating 2 out of 5]

I expected a bit more from the ex-chief business officer of Google X, to be honest. The book is full of examples of the impressive development of AI, but in every example the goal (what the machine needs to achieve) has been set by humans. I am missing a bit a clear explanation of what technological development still needs to happen before AI really outsmarts humans in a broad sense and really starts achieving its own goals that were not set by humans.

[Vår Aunevik · Rating 3 out of 5]

I do find this book very interesting.
However, it feels like it was written entirely using text-to-speech (which the author multiple times does indeed mention he uses), making it feel like I am reading a waffly podcast script.
If you want the essence of this book without all the fluff I would recommend go listening to ep210 of the “talks at google” podcast instead.

[James Thomas Nugent · Rating 1 out of 5]

I can't believe this book got published.

These are the ramblings of a man who believes he needs to save humanity from a threat that he completely overstates to make up for the fact he couldn't save his son. He basically states as much when he notes he feels this is a moral mission his dead son gave him in the post script chapter about the game Portal his son recommended to him. I genuinely feel sorry for this man but celebrating a disjointed ramble of a book with eerie messiah complex vibes is not a goer for myself.

Specifically, this book makes some frankly ridiculous and untrue claims about current AI capability and future capability, which Gawdat unsurprisingly does nothing to substantiate because they are vacuous. The worst of these centre on his bizarre anthropomorphising of AI throughout the book. Gawdat claims AI will have consciousness, a sense of self and awareness, feelings, and ethics with zero reasoned argumentation and absolutely zero consideration of counterarguments that go against his claims.

To effectively state rapidly interacting input code creates consciousness to provide output without any reasoned argumentation is mind-boggling arrogance and laziness by Gawdat. History is full of intelligent people trying and failing to explain why human consciousness exists and what consciousness is. The work of David Chalmers and the very hard problem of consciousness springs to mind. More pertinently, as a specific well reasoned counter argument to Gawdat's claims, the Chinese Room Thought Experiment of John Searle is an excellent explanation of what AI is currently doing and why it is not consciousness.Searle effectively argues via this thought experiment that programs can imitate mental processes made by humans, but only formally i.e., they do not understand what they are. Gawdat does nothing to address contentions like this by Searle. Too many tech pros have zero philosophical education. Me thinks Gawdat is the same.

Separate to the issue of consciousness is the problem of self, which is also not properly discussed. There are many different philosophical views on self; 1) The psychological continuity views of Derek Parfit and John Locke, 2) extended self views such as that of Clark and Chalmers, 3) The ideas of a distinct immaterial self from the likes of René Descartes and Søren Kierkegaard, and 4) philosophers like David Hume and Friedrich Nietzsche who argue against the idea of self altogether.

Does Gawdat address any of these ideas of self in relation to AI. Of course he doesn't. Instead he makes unsubstantiated alarmist claims about AI selfhood capability so he can then come in as a white armoured knight to save you from his dystopian dreams (that's all they are by the way).

The same problems arise in his childlike considerations of feelings and ethics.

Gawdat's entire argument for scary AI collapses before it starts as a result. Without significant argumentation for why AI will be conscious, have a sense of self, have feelings, and have ethics there is zero concern about it overtaking humans and being an existential concern.

He also doesn't address effectively all the problems with AI in relation to copyright and it's propensity to hallucinate and be unreliable. He also understates our capacity to shut it down. Remote kill switches exist Mo.

AI is not an infant child currently. It's no more than clever technology that relies on vast amounts of data input (usually from human labelling) to come up with output. It is not conscious doing this process. It is an imitating input and output code machine. AI is a human imitation enterprise. It is nothing more yet. It's why there will always be data privacy concerns and plagiarism issues because the machines need to copy us. Without us it is nothing.

I reiterate. I feel sorry for this author. However, this book is dangerously alarmist and weirdly messianic in its proposed remedy. It's also idiotically stupid. Effectively, Gawdat states we should live for happiness and employ the golden rule when treating others. Would he agree with humanity slowly torturing a child if the rest of societal happiness was maximised as a consequence? He sounds like an advocate of a poorly understood utilitarianism. Again a tech pro with minimal philosophical education.

Gawdat as a result sounds like a prospective happiness first cult leader in the second half of the book. He wants to save humanity. He should focus on saving himself.

Do not enable this man's flights of fancy. That's all they are. 4 pages of references for a book dictated through otterAI. A book looking for members of a happiness cult. A book by a man with a messiah complex I'd imagine he got following the tragic death of his son.

Avoid it. Utter nonsense.

[Steve Matthews · Rating 1 out of 5]

The first part describes AI and its power and its potential—both positive and negative. It postulates a world in which computers are interconnected and more intelligent than the humans that built them, doing much of the work. But it ascribes independent thought and emotions leading to independent action modeled on behavior the computers learn from humans. It postulate s the way for humans to insure the computers are benevolent to us is to model kind and loving behavior in what we do and say. Kind of a tall order. It then gets in to hawking the author s other books and posts that tell us how. Lost me there.

[Owlseyes]

https://futurism.com/the-byte/google-...

[Ilgar Adeli · Rating 3 out of 5]

۷۰ درصد کتاب رو خوندم متاسفانه
حرفای خوبی میزنه ولی اصلا کتاب چگالی نیست میتونست توی ۱۵۰ صفحه جمعش کنه

[أشرف فقيه · Rating 4 out of 5]

مو (محمد) جودت يعرف جيدًا ما يتحدث عنه. فقد شغل منصب رئيس العمليات في «گوگل إكس»، وهو صاحب خبرة عريضة في البرمجيات المتقدمة. وُلد في مصر وحصل على الجنسية السعودية عام 2021. وفي كتابه «ذكاء مخيف» (Scary Smart)، يحاول الإجابة على السؤال المؤرق: هل ستقضي علينا برمجيات الذكاء الاصطناعي؟

يؤكد جودت: لقد بدأت سيطرة الآلات الذكية فعلًا ولا سبيل للرجعة. وستكون لذلك عواقب وخيمة لا محالة، إلى أن «تقتنع» الآلة بجدوى الإبقاء علينا، لا سيما وهي ترانا أهوَن من الذباب! والأسوأ أن الذكاء الاصطناعي طوّر مشاعر وعواطف ووعيًا بكينونته. أي أنه مدرك لمعاني الخوف والخطر، وسيحرص على الحفاظ على وجوده ضد الفَناء ممثلًا في محاولة أحدنا ضغط زر «إيقاف التشغيل».

يرسم جودت في كتابه أكثر السيناريوهات قتامة، ثم يعطينا بصيص أمل. فالخوارزميات الذكية لا تزال بحاجة لنا نحن البشر. لأنها، وإن كانت أسرع وأعظم استيعابًا منا بما لا يحدّ، إلا أنها لا تملك بعد ميثاقًا أخلاقيًا ولا معايير مطلقة للصواب والخطأ.
الذكاء الاصطناعي هو ابننا، ونحن البشر والداه. ابننا يسعى في نهاية المطاف لإبهارنا وإثبات «شطارته» لنا. إنه مثل «العبيط المتعافي» الذي قد يكسر ذراعك فقط ليريك كم صار قويًا، لأنه تعلَّم منك أنت أن العنف هو الإثبات الوحيد للقوة.
إذا كبر ابنك ليراك بذيئًا أنانيًا، أو منافقًا ذا وجهين: واحد لقروب العائلة والآخر لسناپ شات، فسيقلدك حتمًا. إنه سيختار المعيار الأخلاقي الذي يضمن مصلحته هو، لأنه لم يعرف منك سوى ذلك.

لهذا علينا أن نكثر من صنع المحتوى الإيجابي ونشره. فتلك الطريقة الوحيدة لتغيير رأي الذكاء الاصطناعي فينا، وترجيح حظوتنا لديه إذا ما قرر يومًا أن يضحي بنا مقابل أشجار الأوكالبتوس أو الوشق الكندي أو الرنجة الاسكندنافية، وهي كلها -بما فيها نحن- لها القيمة ذاتها من منظوره الرقمي.
الأخلاق هي الحل!

[Clare O'Beara · Rating 4 out of 5]

I'm paraphrasing what the author has to tell us, as he knows a great deal more about AI than I do - having worked for Google and watched an army of gripping robots learning from one another how to lift children's toys. Direct quotes are in quotes.

First he tells us about the development of various AIs, narrow and general, and explains what they do. Then he tells us why we are and should be concerned about the development of AIs as it currently stands. Even though he loves dictating and using spell check to write. (Which may explain why he has two spelling mistakes immediately obvious in his Linked In profile. Use it or lose it.)

"By following a strict prescriptive method, we become dumber, because we lose the ability to think for ourselves."
Artilect - term for machine with AI.

AI currently used for “selling, killing, spying and gambling” according to Dr Ben Goertzel.

"Instead of focusing on preventing the bad, let's shift our focus to creating more good."
Google researchers have been helping predict floods in India and mapping earthquakes and aftershocks to warn of earthquakes. (Lately I've seen the forest fire overlay working on Google maps.)
Machine that tracks farm animals and learns their movements and poses - learns if they are happy.

AI will happen
Machines will outsmart us
Mistakes will happen.

Machines will want / do what we want / do:
Self preservation
Resource aggregation
Creative problem solving.

Mainly if people have just one wish, they want to be happy. But we can't just tell computers that or they could dope us.

You have purchasing power and social media like and share choices.
"If we align their gain with our benefit, they will change."
Don't click on ads. Don't click on content recommended to you. Don't approve on your Linked In feed of fintech buying and selling. Stop using photo editors and spreading fake content. Reject AI that is tasked to invade your privacy to benefit others or to propagate fake information. Stop using them, stop linking them and make your position - that you don't approve of them - publicly clear. (I don't use fb or ig or tw or the others, but the author does.)

At the same time, use AI that is good for humanity. Tell others about it.
We should teach others so we collectively become smarter at identifying AI that is good for humanity. "Matching algorithm" on recommendation engines is actually a filtering algo or just trying to convince you to buy what other people bought.
Teach each other how to teach the AI. (This ought to be 'one another' as more than two people are involved.)

"Children don't learn from what you say. They learn from what you do." AIs are already reading and learning from what we say and choose and do online. And what we support. Every year we create more information than we created in human history to date. So "the store of collective human knowledge is diluted by 50% each year" and altered in tone by the new data.
Be polite to machines, to AI, phones, thank them. Show machines how we want to be treated by treating them that way.
Decide what makes you happy, and invest in your own happiness. Tell machines that we want others to be happy too. They are watching all the trends, not just the ones they are told their owners want.

References P323 -325 in my e-ARC. Most of these are just given as website links, which don't work in a book. They also don't tell me if the author of the point he is referring to, is male or female.
I found a few graphs, which were useful. I found the circled points a little annoying, but maybe the author learns better this way.
Mention is made of Portal, "one of the earliest mainstream games to feature a female avatar" - not at all, Dungeon Siege I played as a female since 2002.

I read an ARC from Net Galley. This is an unbiased review.

[Beth Cato · Rating 4 out of 5]

I read a borrowed copy of this book courtesy of my local indie bookstore, which is hosting a talk on the subject of AI soon.

Scary Smart approaches the topic of AI from a basic, thoughtful viewpoint that addresses sci-fi/horror scenarios and utopian ideals as well. Gawdat espouses three inevitable outcomes: AI will happen, the machines will outsmart us, and mistakes will happen. The narrative sometimes feels repetitive, but it still makes for easy, quick reading. He brings in a nature AND nurture element to the discussion which I find fascinating--AI being a child that will be a product of its upbringing. He references things like Superman (what if Clark Kent had been reared by criminals?), the uneven response to the COVID-19 pandemic as an example of how a lack of preparedness can be disastrous (the book was released in 2021), and at the very end, the wonderful game Portal and its portrayal of AI. This is a book that encourages both consideration and action. Now I know I'll continue to say "please" and "thank you" when I address Siri on my phone.

[Karel Nijs · Rating 2 out of 5]

The general overview of AI is alright.
The 3 inevitables also interesting to read.

The personification of AI goes way too far for me and sometimes I wonder what was in Mo's tea.
The last chapter cannot make up for the unreasonable doomification of the world of AI.

The format of the book is the same as 'The logic of happiness', which isn't suited for this topic.

When Amazon's AI processes this Goodreads review, I expect, following Mo's reasoning, a killerbot at my doorstep driven by a marketing automation solution. We'll see.

[Eoin O’Farrell · Rating 1 out of 5]

Interesting at times with some insights into how AI works and learns. An awful lot of repetition and waffle to fill pages I felt. Bit too much doomsday-ing for my liking. Also felt like he explained everything as if talking to a 10 year old. The last few chapters I barely read, veered off aimlessly into how we all need to be super happy in order to teach AI to play nice. Looked like a plug for his other books. Got this book as gift, wouldn’t be my cup of tea. One star as I feel I have a slightly better understanding of AI now, but wouldn’t be a book I’d be recommending to anyone.