Is Silicon Valley in a state of moral decline thanks to generative AI? NYU psychology and neural science professor emeritus Gary Marcus thinks so, but he’s not convinced it’s a new phenomenon.The US 2008 financial crisis spurred a Silicon Valley shift toward value extraction and the prioritization of startup valuations over actual profits or sustainable business models, Marcus argues in his new book—Taming Silicon Valley: How We Can Ensure That AI Works for Us—citing correspondence with early Facebook investor Roger McNamee. This short-term outlook in Silicon Valley has trickled into the generative AI we see today, where AI’s risks are deemed worth it because of the hypothetical concept and dream of artificial general intelligence, or AGI, which has not yet been publicly developed by any AI firm to date.But Marcus has more than just an axe to grind with generative AI. It’s more of a broad, sweeping warning, cautioning us that if major changes aren’t made soon, AI could dramatically unravel our digital and real-world lives. Providing numerous examples from around the world, Marcus explains that there are at least a dozen immediate threats posed by the generative AI tech we have today, mainly because there’s little to no regulation in the US and these tools are flawed by design. They’re also free, or cheap, and widely available. Political disinformation, market manipulation, accidental misinformation, defamation, nonconsensual deepfakes, increased crime, cybersecurity threats, bioweapons, and discrimination are just some of the many threats detailed with great specificity in Marcus’ book. He zeroes in on our ongoing societal anxieties around generative AI, exposes the dark side of big tech and anti-regulation sentiment, and lays out a plan of action for regulators and everyday citizens. While Marcus criticizes OpenAI, Google, and Meta, he notably takes a somewhat different stance toward Apple, which “hasn’t quite fallen as far from the tree as the others,” he writes in Taming Silicon Valley. Apple, he argues, is more in the business of selling “sexy productivity tools” than your personal information, agreeing with Facebook whistleblower Frances Haugen that Apple doesn’t have an incentive to deceive the public about its business strategy. In a written interview with PCMag, Marcus shares his unfiltered stance on big tech and the rise of generative AI. This interview has been sparingly condensed for clarity.PCMag: You have written a book about AI previously, Rebooting AI, which came out in 2020 before ChatGPT’s rise. What inspired you to write Taming Silicon Valley? Why now?Gary Marcus: The previous book was a look at why AI is hard and why then-current approaches were inadequate technically. Five years later, I still [think] most of what we said was right. The new book is about what is happening to society, right now, as a very immature form of AI (generative AI) is spreading rapidly. It’s ultimately about something more though: it’s about why getting to the right kind of AI is urgent, and why citizens should get involved.PCMag: Some of the biggest names in the GenAI game right now are telling us that AGI is on the way. Are they right, or wrong?Marcus: Wrong. Whenever people say that, I know they don’t have a good feel for how complex intelligence really is, and for how far we have left to go. And most of the people who say that don’t want to actually defend their beliefs; I offered Elon Musk a million dollars against claims like this but he didn’t dare take it. Even the latest model (GPT o1) can’t reliably win at tic-tac-toe or consistently make legal moves in chess; it has trouble with floating point arithmetic too. Each new model suffers from hallucinations and stupid errors, just like the last, and all we have is an endless litany of promises, not a principled solution.PCMag: In your opinion, can AI ever really ‘know’ anything in the way that humans can?Marcus: I see no reason why not. A GPS-nav system effectively knows where your car is and where you are heading and what the roads in between are and computes an efficient path. That seems like knowledge to me, and reliable knowledge at that.LLMs are much less tractable, much less reliable black boxes. I am not sure I would really want to credit them with any knowledge beyond statistical tendencies. But that doesn’t mean that other approaches can’t be conceptually deeper. Just means we should stop wasting so much money on GenAI, and that we should start exploring different approaches.
Recommended by Our Editors
PCMag: Some Silicon Valley execs and VCs claim that regulation inherently stands in the way of innovation. What do you make of arguments like this?Marcus: They are total nonsense, from greedy people who are willing to say anything to make a buck. Do you think commercial airlines would be safe if there was no regulation?PCMag: Your book discusses the many ways AI is currently being used for harm. AI is also incredibly energy-hungry, in a time where human-caused climate change, the fossil fuel industry, and global pollution remain a concern. Given all this, do you think the AI of today is a net-positive for humanity? Or is it a net-negative today? If negative, what needs to happen to flip the impact?Marcus: I would say that AlphaFold, Google Search, and GPS-navigation are all tremendously net positive, but chatbots and GenAI have been pretty mixed, and could easily wind up being net negative. I think (a) we should be looking for better technology than GenAI, which is the technique du jour but very flawed and (b) holding the companies more responsible for the downsides of their tech (misinformation, harm to the environment, nonconsensual deepfake porn, covert racism in job hiring, etc.) and that if we do, the companies will figure out better approaches. If we don’t hold them responsible, we are looking at a mess.PCMag: Some big tech firms, like Meta, Google, and Microsoft, have already seen backlash to their AI products. Surely, Meta is aware that viral AI-generated Shrimp Jesus imagery isn’t moving the needle to make society a better place, and Microsoft’s Recall feature has been criticized as a security concern. Why do you think these companies keep trying to push generative AI features on users, often without giving them a choice to opt out?Marcus: Obviously they are all looking for an ROI [return on investment], but the technology isn’t working out nearly as well as they wanted the public to believe, and both the public (and big companies) are realizing that they had been sold a bill of goods. So you have companies like Microsoft that seem desperate to find a use for GenAI, but not convincing all that many people. PCMag: As you discuss in your book, AI-generated images are crowding out legitimate historical or real-world image results on Google. Google’s AI Overviews have also been a big problem that users can’t really turn off. How should everyday people navigate the web as Google and Bing become increasingly filled with AI-generated results?Marcus: They should “just say no” and tell the companies that they don’t want a bunch of climate-crushing, copyright-violating AI systems they didn’t ask for, and maybe even consider boycotting AI if Silicon Valley doesn’t get its house in order.Taming Silicon Valley is out on Sept. 17 from MIT Press.
Get Our Best Stories!
Sign up for What’s New Now to get our top stories delivered to your inbox every morning.
This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You may unsubscribe from the newsletters at any time.
About Kate Irwin
Reporter
I’m a reporter covering early morning news. Prior to joining PCMag in 2024, I was a reporter and producer at Decrypt and launched its gaming vertical, GG. I have previous bylines with Input, Game Rant, and Dot Esports. I’ve been a PC gamer since The Sims (yes, the original). In 2020, I finally built my first PC with a 3090 graphics card, but also regularly use Mac and iOS devices as well. As a reporter, I’m passionate about uncovering scoops and documenting the wide world of tech and how it affects our daily lives.
Read Kate’s full bio
Read the latest from Kate Irwin
