Nuclear powered Air-Planes, Hashcash, and the AI Revolution

René Pfeiffer/ April 28, 2023/ Scuttlebutt

[This article is part of the monthly publication on our scuttlebutt mailing list. Not all the scuttlebutt messages are published on our blog. You are encouraged to subscribe to our mailing list.]

Dear readers,
the world of information technology and information security is driven by trends. This is very similar to the fashion industry or other aspects of our society. However, the impact on all of us is much bigger when a trend shifts the attention of the whole IT industry. Let me give you an example from the world of physics.

During my time at the university, I read two books with anecdotes from the life of Richard Feynman. In the context of his work at the Manhattan Project, he told the story that someone from the US government asked him about the use of nuclear power outside the field of the military. Basically, it was a brainstorming session, and the key question was which uses of nuclear power for propulsion systems Feynman could imagine. In conversation, he talked about using this type of power source for aircraft, rockets, and submarines. He did not pursue the technical details of the applications. The notes of this conversation were later turned into proposals and patent applications in order to ensure that the US government would have the patent rights. The New York Times even wrote an article titled “U.S. at Work on Project to Apply Atomic Power to Planes, Missiles” on February 23 1947 (source is at the end of my letter). So what do you think of strapping a nuclear reactor to the end of a missile? Better not tell Space Karen.

To set the perspective right: Feynman did not propose to do this. He just let his imagination run free. Maybe you can do this, maybe not. The first phase of any design is to collect ideas regardless of their engineering or scientific limits. The second phase is all about risks, effort, costs, efficiency, and other aspects. ChatGPT and the current hype about algorithms of artificial intelligence are currently in this first phase of design. Now everyone is thinking about what the alleged new technology can do for us. People and companies come up with all kinds of ideas. I am waiting impatiently for cars, missiles, ships, and elevators powered by AI (“powered” in the sense of “energy”). Of course, this sounds absurd, but strapping a nuclear reactor to anything that moves is on the same level.

There is a big difference. First of all, ChatGPT and its relatives are not algorithms of artificial intelligence. Their code does not think about anything. It is a language model that can produce convincing texts. The algorithm combines parts of its training data and creates a combination of “knowledge” that is existing. It creates nothing new. Its uses are centered in areas where language plays an important role. This includes programming languages, which are a collection of vocabulary and grammar rules, just like their counterparts used for communication and culture. Again, there is no thinking involved. ChatGPT cannot find new ways of implementing the containers of C++’s Standard Template Library (STL), but it can tell you that maps use red-black-trees and that unordered maps use hash tables. It cannot tell you why. It also cannot find better or new ways to implement these data structures. Likewise, the rise of computing hardware and Moore’s Law of computing performance have not rendered mathematics redundant, quite on the contrary.

This year’s DeepSec has AI algorithms in its focus. The IT security industry has seen the labels “machine learning” and “artificial intelligence” on products by vendors before the rise of ChatGPT. The advertisement for these security products reminds me of Feynman’s nuclear brainstorming after the Manhattan Project. Yes, machine learning and AI algorithm can solve some problems. For example, they are often used as a filtering system. This is nothing new. Filtering information and looking for clues or patterns is the primary task in IT security. Spam filters do this for over 20 years now (and they even could benefit from language model algorithms, if these would not use so much resources). Strangely, the anti-spam proposal called Hashcash (dating back to 1997) was reborn with a slightly different design as Bitcoin (fans will object, of course, but the underlying principle was explored by Hashcash). So maybe email systems are the real innovation source of IT. Regardless of these historical facts, we want to know what AI can do for IT security and which applications are a threat to both the technological world and society. Manipulating language has a long history, and it has not only been used for the benefit of people. We like to go beyond the obvious applications. Yes, ChatGPT can write code and create phishing messages. This is no surprise, because that is what you do with language models. We are looking for a few steps further, just like the US government official who interviewed Richard Feynman.

And there is the defence, too. How can you defend against manipulating language? Sounds familiar, because this is also the key question for politics facing threats to democracy. Despite having AI models with vast collections of training data, we still live in the Age of Disinformation. The Semantic Web and free access to the Internet have not changed this, sadly. Judging from my experience, having libraries does not help. You also have to convey information and deliver it to the audience. Social media platforms have shown that this can be used for either side. This has also an impact on IT security defence, because arguing for sensible protection measures can be influenced by the hype of the day. Technology alone cannot solve a problem.

So if you work on problems or solutions in connection with artificial intelligence algorithms and maybe even information security, please let us know. Let’s turn on the first design phase and think about how all the shiny tools we have can be used for defence. And please, do not use nuclear reactors in your IT infrastructure. Thanks!



Feynman, Richard P. (1985). Ralph Leighton (ed.). Surely You’re Joking, Mr. Feynman!: Adventures of a Curious Character. W. W. Norton & Co. ISBN 0-393-01921-7. OCLC 10925248.

Feynman, Richard P. (1988a). Ralph Leighton (ed.). What Do You Care What Other People Think?: Further Adventures of a Curious Character. W. W. Norton & Co. ISBN 0-393-02659-0.


Share this Post

About René Pfeiffer

System administrator, lecturer, hacker, security consultant, technical writer and DeepSec organisation team member. Has done some particle physics, too. Prefers encrypted messages for the sake of admiring the mathematical algorithms at work.