Junk Food in the Age of AI

In November 2025, I was stunned to find AI-generated video channels on YouTube showing me saying things that I do not approve of. An even bigger source of distress was the realization that some of these videos garnered a million views within a few days.

This finding was worse than the misinformation and antisemitic hate spread a few months earlier in two YouTube videos created by an influencer, who suggested that my scientific practice as a Harvard professor is a fraud while at the same time labeling himself on his YouTube channel as “Professor Dave” — even though he never held a professorship, nor did he ever receive a PhD. Since his statements were “not even wrong” in the words of Wolfgang Pauli, they were of little concern to me. The irritation caused by this fake “Professor” felt as insignificant as a mosquito bite. However, the fake AI-generated videos, spreading misinformation in my voice and appearing in my image on the background of my office, felt more like a punch in the face. Both misinformation campaigns were motivated by the desire of their creators to make money out of my public fame.

My pain from the AI punch was exacerbated when an 83 old ecologist, named Abby Rockefeller (daughter of David Rockefeller), came to my Harvard office to tell me that she was extremely impressed by the content of one of the AI-generated videos and so she shared it with her friend group, only to be told by one of them that it was fake. Following this emotional rollercoaster, Abby was deeply frustrated not only by the fact that she was fooled by AI, but also because she wanted to believe the misinformation conveyed in the video.

Could AI trigger a misinformation instability in society by generating content that appeals to its consumers more than the truth? This would be akin to feeding the public with intellectual junk food that tastes good but is unhealthy to the mind.

Today, I attended an insightful WORLD.MINDS forum coordinated by Rolf Dobelli who hosted the brilliant German journalist Mathias Döpfner, chairman and CEO of Axel Springer SE.

My question to Mathias started with an analogy. The old world order involving the U.S. and the former Soviet Union was stabilized by the fear from mutual annihilation in a nuclear war. The existence of nuclear weapons created a stable geopolitical equilibrium because both sides wanted to survive. However, such an equilibrium is unstable if one side values martyrdom — in which case mutual annihilation does not pose a deterrence. This is the primary rationale for denying Iran access to nuclear weapons.

In the same vein, traditional truth-seeking reflected a stable equilibrium in the old world order of journalism, where evidence-based vetting was valued much more than engagement with the largest number of subscribers. In the current social media climate, where content creators gain more subscribers by spreading hate and misinformation, conspiracy theories thrive. But to all of us who value the truth, the misinformation spread by human influencers feels like a mosquito bite because their content is “not even wrong”. I explained the clear difference between evidence-based scientific research and social media in a recent essay, posted here.

However, the AI campaign of misinformation poses a much bigger danger. The reason is simple. AI can intelligently optimize misinformation by creating the ideal recipe of intellectual junk food. Once AI becomes more intelligent than its customers, members of the friend group of Abby Rockefeller will not be able to warn her about misinformation because they will all be fooled like she was.

In particular, I wonder how many scientific papers are already being published in scientific journals today with false AI-generated content, in which either the data or its interpretation is fabricated. The way to find out is to attempt to reproduce their data or interpretation. After all, the Sun did not start to move around the Earth just because the Vatican published the notion that the Earth is at the center of the Universe. The physical reality is independent of what influencers say about it and it is our duty as scientists to use experimental data in vetting what it is.

I therefore asked Mathias whether AI might break the stable equilibrium of truth-seeking journalism because readers might prefer to consume fake content which is more appealing to them. Just as with nuclear weapons, the old world order could be destabilized by a different value system of the humans involved.

Matthias assured me that most people are truth seekers and will reject the temptation of AI misinformation. I hope he is right. Until we find evidence for extraterrestrial technological civilizations, there is always the possibility that they died from self-inflicted misinformation wounds.

Finding technological signatures of alien civilizations is now more urgent in my mind as a practicing astrophysicist, because it will bring relief to my existential concerns about the future of humanity in the age of AI misinformation.

Speaking for myself, my peace of mind rests on scientific evidence both on Earth and in the Universe at large.

Before my morning jog at sunrise, I had received the following uplifting message:

“Hi Dr. Loeb,

I’m an 8th grade science teacher, and one of my students wrote you a thank you note for your research. Thank you for everything you do to continue to inspire young scientists. Have a great day!

Beth

Letter Pasted Below:

Dear Dr. Loeb,

I am an 8th grade student from Western Sierra Collegiate Academy. This email that I have written is to appreciate and thank you for your research and work regarding the Galileo Project. I recently came across your work on the “Chemical Classification of Spherules Recovered from The Pacific Ocean Site of The CNEOS 2014–01–08 (IM1) Bolide” (published here).

When I read it, I was captivated by what I learned and what discoveries you made at that site. It is really astounding to learn how many spherules exhibit elemental ratios that are distinct from known planetary bodies and how their origins are undetermined.

Thank you for your dedication and hard work in finding out how the “BeLaU”-type spherules chemical composition is unlike any known solar system material.

Looking forward to hearing from you,

Tanishkaa Raju”