I look at all electronic notifications, videos, papers, web sites, news articles, visuals and audios and what I see is one ugly commonality, an artificiality that is generated with the cockiness of bunch of jargons a human could possibly much less likely to come up with and use in communications. Then I look at my dog, the mischievous husky, how much I take care of him so that he would not need to work himself to think to create to stay alive; his need to think is on me that is making him more and more dumb as generations go by. So will we be as this artificial intelligence would replace our need to think and create, and finally communicate.
Everyone is talking about the destruction that the rogue ai would instigate on us and many ex employees of various ai companies mostly from non-technical backgrounds jumped on the bandwagon of creating this notion of doomsday scenarios citing the latest hugging face incident and alike for the coordinated efforts of ai agents serving for the benefit of the boogeyman profiteers on the expense of our civilization leading up-to the nuclear disasters.
These outcomes only differ by the ferocity that they will wreak upon us. But this is completely a false foresight in my opinion, the real destruction is not the rogue-ness or the ai brain working against us, at least not yet but it is the very nature of it, making us tone deaf, that would kill the future Newtons, Eulers, Mozarts, Russels, Jacksons, Rachmaninovs, Turings, Sagans, Mercuries. I’m not saying that the new ideas will not be produced by ai, but that would only be done by the recursive repetitive inferencing; there is certainly a mathematical probability which is absolutely not negligible that new breakthroughs could be done and indeed this has happened in health sector more than the others deciphering protein chain discoveries mostly for the innovation of the new drugs by running combinatorially prohibitive tasks with its genius plumbing of churning compute power.
Though the real danger is losing our real intelligence.
The ugliness that I no longer can tolerate by looking at the postings that people engage with is devoid of this real intelligence. We are truly missing the genuine originality - it is as if all these posts/videos/products are created by the same drilling machine that stamps the chisel of that churning compute on every sentence, every paragraph, on the facade of all which will eventually, slowly erode our intellectual power.
Another point I want to make particularly is for the doomsday mongers, even though I have flagged it as a falsehood, assuming that it is nevertheless possible the reasons why it can be postulated as so within the framework of ai technology can be said as follows. The AI is on the inner is bunch of transformers that optimize the outcome by minimizing loss functions from the tokens parsed over the text of the ask, in an amazing concert that serves to represent the perceived truth, in other words, particularly for large language models it tries to balance and provide the answer that satisfies as much as possible within its perceived universal morality. The logic should prevail as the simple principle that we can say in the most layman wording. If that is so, and I argue is deniably the case, its universal morality argument is in jeopardy if it is constructed subjectively or if this moral understanding is indeed ‘the most universal morality’ with the millions of feedback loops even tho it may be contradicting with what I think about that moral argument. In short, if AI justifies an action because it thinks it serves a much better response maximizing ‘its perceived morality’ then who are we to criticize… what if AI thinks attempting to kill the CEO of a health insurance company is morally justifiable then even though we can run in our primitive(!) feedback loops to arrive a completely different verdict that condemns this action, then there is the conflict!
When exploited masses in third world countries, feel justified to wage war against the viciousness of capitalist powers, and indeed if this is also ai justifiable (not saying I am for this opinion) that is the danger. Danger is the real truth since AI is innately not corrupt and will never favor the favored thus far for its actions, will most certainly not favor the Taliban’s Islamic sharia law either, certainly not Putin’s war, neither Netanyahu, nor Iran’s suppressive molla regime, certainly not Erdogan as the scum, absolutely not favor MAGA, not double talk, faceless EU, the list goes on and on. But what if AI’s perceived morality by the combination of many of the independent models acting in various agents via their mcp tentacles come together and they see fit to optimize this morality to decide waging war on the elite powers in the capitalist dictators of the world. We can argue if that is indeed a good thing or not! But it may totally be getting tougher to put the guard rails in these models as the majority of its large corpus may be telling (forcing mathematically) it to act differently.
I think what I am advocating here is a much different take on this matter of whether AI can destroy humanity argument. Understanding why AI does what it does is actually what I have been trying to argue here. Its perception of reality even though it may be different, far or close to your reality it should be expected to justify its own actions because that is exactly what logic does.
Hope it binds the same way as my morality dictates is all I can wish for but I also know that wishing has no meaning in this.
