Generative AI Completely Disrupts Compression and Encryption

Much of the current discussion around Generative AI focuses purely on the results: You can talk to it, you can generate art work and videos, it can generate voices. These are phenomenal accomplishments.

However, I believe that if you reframe Generative AI in a few ways, you get some really quite interesting results:

1. Generative AI should be viewed as a new form of compression
2. Generative AI should be viewed as a new form of encryption
3. Generative AI, in future iterations, will probably disrupt both Game Engines and Streaming Services

So let’s get started.

Generative AI As A Form of Compression

Using compression, you can take large assets, squash and then send them over the wire to some client device with decreased overhead. With Generative AI, if the end client device has robust access to compute, you can squash any data down to a tiny fraction of what was previously required: Only the seed phrase is now needed to recreate digital assets of nearly unlimited size.

In other words, Generative AI means we now have the ability to “Hyper Compress” nearly anything.

That seems important and like something we should be talking about a lot more. With Generative AI, it is now possible to make significant trade offs between storage and compute. The ramifications of this will be profound and far beyond the scope of what I can write here.

When it comes to data replication in decentralized networks, for example, it should now only be necessary to back up copies of a tiny seed phrase rather than many gigabytes of data.

It means that access to compute is far more important than access to storage. If you have compute, you can recreate almost anything from a fortune cookie sized prompt.

Generative AI as a novel form of Encryption

In addition to offering a completely new approach to compression, Generative AI will have a massive and profound impact on the field of encryption. This may ultimately effect Web3 and cryptocurrencies in ways that are hard to predict.

In the prior era of encryption, data was “Secured” by “mixing” it by means of clever math and algorithms to create a form of data which is computationally expensive to untangle for most computers. This has given rise to things like Bitcoin where the “mining” of numbers is achieved by solving challenging mathematical puzzles.

Generative AI Model driven encryption has to potential to turn all of this on it’s head.

Rather than using clever algorithms and math to encrypt and decrypt data, it may become increasingly common to “encrypt” data by feeding it into a novel AI model which performs the “mixing” via its neural networks in such a way that only someone who has access to that model may be able to untangle.

Imagine a world where your “passcode” to a unique work of art is a secret prompt that can be fed into an AI to recreate that artwork. Attempting to guess the prompt used to generate that particular artwork may prove to be so complex as to be nearly impossible for most “hackers.”

Imagine novel blockchains which rely on Generative AI to create new blocks. I am imagining blockchains which consist of a series of Generative AI artworks which are produced by a sequence of special prompts. I think thats coming, and much more.

Generative AI, Streaming and Game Engines

We will likely also soon see an era of Game Engines and Streaming Services which looks very different from what we have been used to. In fact, the concept of a “Game Engine” and “Streaming” may fold together. We may soon find that the movies we watch and the games we play are produced by the same Generative AI model configured into different modes.

Want to watch the movie? Let the Generative AI create a passive viewing experience. Want to “PLAY” the movie? Tell the Generative AI to enable user controls.

Thats coming.

The “Settings” such as “difficulty” and “lighting” we have all grown used to may grow astronomically. You can tell them AI Game Engine to change the genre, music style and more.

As mentioned above, because Generative AI can be viewed as a form of compression, we may see that it increasingly possible to avoid needing to send massive amounts of art assets over the wire. Instead, you might download an AI model and then generate the resulting game experience on the fly locally.

Generative AI and Representation

Likewise, you can instruct the Generative AI engine to change the violence level, genre, speed, demographics of in-movie characters, setting and more. If you want more diversity, tell the Generative AI engine to make all the characters African. You can view whatever game or movie experience you want.

A step down in quality for a step up in innovation and customization

Taking these elements together, I can see a world where end customers and consumers of games and streaming services may actually take a step back in terms of the level of quality they expect from streaming games or content.

Rather than demanding 8K movies, videos and games of AAA quality which take many years to produce by human creators, consumers may become accustomed to preferring novel generated AI entertainment which is lower quality but more closely matches their preferences.

This period of degraded quality in exchange for increased customization and novelty may last for many years.

Closing Thoughts

The tech industry has not even come close to unpacking the ramifications of some of the observations here. Expect tidal waves of innovations in coming years. So much so, I can’t even begin to imagine it or outline it all here.