The AI train only keeps accelerating
2023-05-20
The public release of ChatGPT (although officially still in research preview) really was and still is a quantum leap for AI and especially its public perception. With the release of Llama and a lot of leaks, the development in this area has reached an unprecedented pace, so let's take a look at what has happened, what will happen in the future and what an impact that will have on us humans...
Undoubtedly, ChatGPT was THE big drop last year enabling millions of users to access a cutting-edge AI chat bot without any specialized knowledge through an easy-to-use web interface. This created an immense pressure on other big tech giants to release similarly capable systems in time, before losing the AI battle once and for all. Facebook, oh sorry - Meta of course, rushed out to release LLaMA on February 24, 2023, while Google announced it's take called Bard already on February 6, 2023, but only recently opened it up for early access. In all this hurry, it is really no wonder that mistakes happen, which can be seen by the mysterious leak of the weights for LLaMa enabling hobbyists all around the globe to experiment with this technology in their basements on relatively cheap hardware.
The current trend of integrating these system with browsers, search engines, office suites, code
editors/IDEs and so much more really shows us the direction we're leading towards. In the next years an
increasingly high rate of content we create and consume will be AI generated and for the most part
without a meaningful loss in quality. But even though these systems certainly are very impressive and
capable of generating persuasive texts, they aren't the holy grail (yet). Fundamentally, they still are
text prediction systems without any checks to internal logic, which can lead to nice quotes like
3.5 > 4
. With the current, only increasing development speed however, I am very confident that these issues
will be resolved shortly.
Generally, the well-planned use of AI technology has huge potential for boosting productivity by offloading simple repetitive task, getting immediate answers to specific questions without having to read a lot of (sometimes sketchy) websites and ultimately starting a new era of work, where human beings can focus on what they love doing and specialize in what they are interested in instead of having to work in a stupid 8-hour-a-day job just to pay their rent and keep their head above the water. On the other hand, there also are a lot of (potential) risks like invasion of the users privacy to an unprecedented extent, unbeatably good targeted advertising or computer generated content on social networks (lowering the world's productivity even further than it's already abysmal level) and ultimately leading to mass unemployment, when large amounts of jobs can be automated and society isn't able to catch up with the development (by introducing an Universal Basic Income or similar measures).
Obviously, I don't know how long these developments will take (likely longer than shorter) or if my predictions are even remotely correct (they certainly are realistic though), but the current developments make me really optimistic about the future of human kind, if we as a species are able to adapt to the rapidly changing environment we are creating for ourselves. It's the closest thing to/the go-to tool for reaching the utopia of machines doing all the boring work and fixing the problems we created (like climate change) allowing us humans to really focus on enjoying live and being creative. Nevertheless, creating a real world Skynet also isn't out scope of what's possible in the long run - so AI overlords, please be merciful to me.
With that said, I am very curious about your opinions and predictions on this very interesting, relevant and rapidly evolving topic. Feel free to post them in the comments down below...
PS: If you are interesting in experimenting with these technologies yourself, I can highly recommend taking a look at llama.cpp and GPT4All .