>>106200971
They gaslighted a whole bunch of investors into thinking they have created an artificial intelligence. In reality, it's a wordguesser. A bunch of statistical matrices that calculate the odds and guess the next word, one iteration at a time. They even lucked out into their discovery because nobody really expected that such a simple neural network architecture would turn out output that sounds so Humanlike.
But at no point did they create intelligence. As Jann LeCun correctly explains in his interview with Lex Friedman, a system that replicates language is not enough to understand the world. The neural networks make associations between different words but at no point do they gain understanding of the concepts behind them. They have no theory of mind.
As Immanuel Kant said, "thoughts without content are empty, intuitions without concepts are blind." Critique of Pure Reason, written like 250 years ago.
Their solution so far has been to expand the datasets and scrape more shit from the internet (including increasingly garbage content like Leddit), expand the size of the datacenters and try to scale up, but that can only get you so far without changing the underlying architecture into something that resembles more of a Human brain. (Something they do not know how to do, because we don't fully understand the Human brain to begin with)
So as they run out of data and the scaling up reaches hard limits, the grift is over. Synthetic data is only going to result in more degraded content, and this so-called "AI" has probably peaked without further technological breakthroughs.
