Why do people hate a title like that? It’s not hype, execration, sensationalism or hyperbole. Ignore what the companies SAY will happen, and just focus on what has happened. Within a few years we have seen an exponential rate of progress in intelligence. That is absolutely insane and utterly incomparable to the rate of anything with similar societal impact historically. Now LLMs using new techniques get 5/6 questions on the international mathematics olympiad. Something only a handful of the participants manage. These are the most mathematically talented teenagers participating countries can send.
I’m seeing a lot of either denial of capability or denial of risk. “We don’t have AI” “LLMs are not AI” “It’s just a new technology, why should we be afraid of that”
No, we’re approaching systems that match and will soon exceed the intelligence of the smartest humans. LLMs will probably plateau somewhere, but with the collective effort being poured in worldwide (AI infrastructure spending exceeded consumer spending in the US) that plateau will be temporary before a new technique allows for progress to continue.
Once AI exceeds humans by a nice clean margin, it won’t improve our productivity, allow us to understand more complex concepts. No it will outright replace us. When given agency, without knowing whether it is actually aligned or not it will be an existential threat to humankind. In the hands of a few powerful people aligned AI will allow them unprecedented power. The people in charge of these AI companies are petulant and irresponsible, just look at their twitter fights.
In the hands of everyone (the fabled opensource ASI scenario) aligned or not will in all likelihood create total chaos and destruction, however we have absolutely no way of predicting the exact outcome with our current understanding.
We aren’t ready for this. And sticking your head in the sand, saying it’s inevitable is not going to help either. Legendary academics everywhere are ringing the alarm. Are we going to let capitalism and libertarianism sort this out for us, like it has for all other problems?
[link] [comments]