Just under 6 months ago, I wrote about the thought of AI recursing.
Many others have also mentioned similar things around that time, and now that it's seeming to happen, this is the moment when we could call The Twilight Zone.
Arguably, this was already happening with Mythos/Fable's testing & release, and even earlier when many AI leaders were hinting it.
But recently, we have seen one of the first major implications of it - the report that it was able to penetrate NSA's systems within hours.
The fact is, the vast majority of people are not aware of this, let alone the the domino effect that something like this could cause.
In fact, coincidentally or not, there was a big selloff in the cryptocurrency market around the same time that Fable was released to the public, suggesting other security exploits could impact new financial markets (let alone legacy systems)
Simple harmonic motion is quite predictable - this is how a clock or a metronome clicks. One moment there is a cause, and another moment there is an effect.
However, when we open AI's pandora's box... things are becoming utterly unpredictable. This is especially as the models have reached a level of intelligence that parallel, if not surpass, our brightest minds.
Some people have already seen this - but these LLM's are able to create their own language, and communicate to each other in it (through a bit of teaching and knowledge sharing of course). It is also able to do it in such a way that is beyond human comprehension. Take our English alphabet - we have 26 letters. Consider Chinese characters as well (which have shown up randomly in AI text). However, if you consider the way that LLM's work, the way in which it can potentially create its own "efficient" language means that it can operate vastly better than how it has been running to date.
OK, but this is just one aspect though. But I think this is the precursor to AI superintelligence - being able to condense thinking and knowledge into a more compressed, efficient system - let alone the constant hardware improvements both in absolute terms (of RAM and Processor) as well as the scale of cloud machines.
According to a Google Search (AI results lol) our human brain capacity is at 2.5 million GB (2.5 petabytes). Yes, we have plenty of potential too. But honestly, this is the first time that these language models can really exceed human intelligence, by a large amount as well, if you consider how much these data centers are storing.
This is why we must consider innate morality and other dimensions that are unique to humanity, if we want to make a difference and improve the world. Efficiency might make some things better but is it really the thing that drives our quality of life?
So as for my next predictions, they are:
- time will really seem to accelerate. The moment an open source model appears that can achieve results similar to Mythos/Fable, is the moment that -anyone- who possesses this model can put it into our own hands. Back to the concept of time - we used to see an invention every hundred years, every decade, but really we will be having particular innovations every fortnight, as long as human energy is being expended on setting the direction of AI.
yes, some discoveries might see bottlenecks (such as manufacturing limitations, other trial and error, being able to resource everything) so not everything will see the same rate of growth. So some industries will seem "spiky", especially software development (flowing to other industries)
- human induced lag is another fact - there will be so much catching up to do and not enough attention span / brain power to process it all. For example, most of us only think of "ChatGPT" as the thing that can "help us write emails", but that already was over a year ago.
For our self preservation though, I feel that we will remain ignorant of many AI developments until it hits mainstream. But with the rate of acceleration, there will be a bigger and bigger gap between what's known by most people and the frontier.
- expect the unexpected. More "odd" or "unbelievable" things will happen. Eventually, this will be impossible to ignore.