Recently, I did some tests on my computer with Lama3 and found it bizarre how rapidly artificial intelligence technology is advancing.
Just like torrents, artificial intelligence models are expected to spread across the internet in a decentralized manner, where each individual could have a copy of the neural network on their computer. Similar to how blockchain technology functions, we might see people selling language models as a 'cure-all' for various issues. This could fundamentally change the way AI technology is distributed and accessed, potentially making it more democratized.
Initially, many websites might offer payments for users to join their platforms and allow their GPUs to run computations. This could create a new economic model where computational power is commoditized. However, a major challenge will be validating the information produced by these decentralized models. This would require a Proof of Stake (PoS) system rather than Proof of Work (PoW), as PoW is not suitable for this context. PoW doesn't work in this context because there's no way to verify if the client’s deep learning was genuine. That's why ChatGPT uses evaluations in its responses as positive and negative feedback for a PoS (Proof of Stake) learning system.
One significant development is that ChatGPT and similar AI technologies have begun using user processing power to extend their computational capabilities. For example, WebTorrent, which can be found at WebTorrent GitHub, is a technology enabling decentralized peer-to-peer sharing. This approach could potentially be adapted to distribute AI models, making advanced technology more accessible
The rapid advancements in AI are significant. AI systems are no longer just simple tools; they are creating books, artworks, films, and even driving vehicles autonomously. This swift progress requires adjustments across various sectors. As AI technology evolves, its impact on society and industry will be profound, necessitating continuous adaptation and regulation.
CAPTCHAs are increasingly ineffective because distinguishing between bots and humans is becoming more challenging. Scammers are using deepfake technology to commit fraud, such as manipulating KYC (Know Your Customer) processes. As a result, there is a growing need for more effective solutions beyond traditional CAPTCHAs.
In a few years everyone will have a fork of ChatGPT (or Lamma) on their computer. There are already thousands of forks of Lama (Facebook's AI) running on the Internet, some of which are even uncensored and programmed to have no rules.
I'm very happy to be witnessing this whole machine revolution. 🥇