This Post Has 8 Comments

1. Generative AI is not the best option for some fields, such as medicine – predictive AI is superior!
   Gary Marcus provides a valuable service by pointing out the issues with generative AI. He also makes it clear that, for some fields, such as healthcare, there are superior AI methods, such as predictive AI models. Marcus warns against risky shortcuts and writes “Only very rarely does the media call out such nonsense.” He concludes: “In the long run, we need research into new forms of AI that could keep us safe in the first place.” In healthcare, this will mean testing whether AI improves patient outcomes with pragmatic, randomized controlled trials. Marcus also provides these words of wisdom: “AI is almost always harder than people think.”

   Reply

2. A key reading for the context of AI governance
   Gary Marcus delivers a raw and critical examination of the current state of AI and the companies that develop it and promote it. He underscores that most of these companies are motivated by incentives aligned with profitability and growth at the expense of societal well-being and productivity improvements. Instead of benefiting society, they amplify among many problems, surveillance capitalism and the attention economy, leading to increased misinformation, polarization, and negative impacts on users’ mental health and cybersecurity. He highlights how the rapid pace of AI implementation surpasses safety measures, driven by growth and profits.
   Backed by extensive documentation—including academic articles, tweets, news reports, and other books—Marcus reinforces each point he raises with solid evidence. He presents the main risks that this globally scaled technology brings, affecting society globally.
   Offering his perspective on the moral and ethical decline of Silicon Valley, Marcus reflects on how it was originally founded with the purpose of positively impacting the world through technology by increasing well-being and productivity. He is especially critical of big tech for not having solutions to the flaws of generative AI.
   Marcus argues that there are more promises than realities in the AI industry, with hype often exceeding concrete results. His criticism is harsh and scathing toward big tech and venture capitalists who push the frontier without a responsible vision and without true containment measures.
   Finally, he provides a guide—a set of demands—that we must make to pertinent actors such as the government and legislators to regulate this industry and make it responsible, avoiding impunity over the negative externalities it causes.
   We are at a critical moment where each and every one of us must take action—from activism and communication to voting in elections for politicians who understand the worrying inflection point at which we find ourselves as human society.

   Reply

3. Amazing Insights
   Must read for everyone who is naively using gen AI without understanding its limitations

   Reply

4. Very timely and critically important book about AI’s benefits/challenges – and how we should prepare
   An accessible overview of current AI and its possible future paths. Well reasoned and clearly written – does a great job in discussing how governments and citizens need to respond to AI so that we can enjoy its benefits and avoid its clear potential dangers. Informative, enjoyable, and important.

   Reply

5. A Useful Overview of Current Circumstances
   The book draws together in a useful way many of the current criticisms of the ways machine learning technologies are being developed and sold as “AI” together with an analysis of the risks and dangers the technology poses. Marcus also presents some ideas about how “AI” should regulated in the public interest (as he thinks it should be) and an agenda for research aimed at developing a “trustworthy AI” beyond large language models and their focus on “mimicry”.
   Marcus notes that he wrote the book quickly out of a sense of “urgency and disillusionment” and it does present what amounts a warning about what is at stake in the development of “AI”. Already, he writes “the big technology companies are making some of the most important decisions humanity ever faced–on their own, without consultation from the rest of us.” (13) He argues that they should not be left alone to do so and presents number of suggestions about what citizens and governments should demand from the companies currently developing “AI” technologies.
   It’s important to note that while Marcus is critical of the ways in which “AI” is being developed at the moment–current research is, he says, amounts to “intellectual monoculture”–he advocates for continued responsible research and development in the direction of actual artificial intelligence.

   Reply

6. Succinct overview of what’s wrong about today’s AI development
   I got to read an advance copy of the book, and I must say it is excellent (note: Gary did not ask me to leave a review).
   I love the stories, which Gary knitted together clearly.
   The assessment of generative AI’s risks is practical, and I felt hopeful after reading this! The bubble is popping, the tide is turning, and we can do something about this.
   My view on AI is different btw. I think that machines reprogramming on surveilled data, power-hungry managers directing redesigns to automate more jobs, corporations competing to scale off extracted resources, and increasingly tighter evolutionary selection would converge on a monstrosity. Ie. my view here is closer to “let’s just actually stop AI”.
   Having said that, this book is absolutely excellent. I heartily recommend it.

   Reply

7. A must read book
   This book serves as a comprehensive collection of the critiques Gary Marcus has consistently voiced across his social media platforms, blogs, and public speaking engagements. As one of the foremost critics of AI globally, Marcus’ sharp insights and deep expertise are clearly demonstrated throughout the book, making it evident why he holds such a prominent position in the discourse on AI’s limitations and challenges.

   Reply

8. A great overview of how we’re failing to adequately legislate the tech industry, with potentially lethal consequences for people’s future, livelihoods and even the future of AI tech itself.
   Gary Marcus explains issues like the “echo chamber effect” in which AIs sometimes “learn and purvey nonsense generated by other AIs” which could lead to model collapse and cause the quality of the internet to degrade. Or that as the systems are large language models and not classical databases, it’s records can’t be selected, deleted, protected and so forth. “They are giant bags of broken bits of information and nobody really knows how these bits of distributed information can and cannot be reconstituted… [Therefore] We cannot say what hackers do and do not have access to.”
   He also examines its many misuses, including as a channel to spread misinformation more quickly; for political propaganda and individualised advertisements tied to personal information; for surveillance capitalism; to gain private information for scammers – did you know ChatGPT has “coughed up confidential passwords”? – and much much more…
   Read this book, then demand AI be made safer, better and more trustworthy through the implementation of clear legislation and statutory transparency and privacy requirements; layered, independent oversight and international governance; and government funding for research into a new approaches to building trustworthy AI…

   Reply