There is nothing wrong about the specific objection but, to borrow an analogy from Scott Aaronson, it is rather like dismissing the Wrights’ early airplanes as being slower than a train and only carrying one passenger. What is missing is an appreciation for how intractable the problem seemed five years before.
It's not really like that? A small plane with one passenger is a concept that you can extrapolate to a bigger, faster plane. Afaik an often-wrong generative AI is _not_ a concept that extrapolates to a never-wrong generative AI: that's just not how it works at a fundamental level.
Although presumably very smart folks are working on it.
For me, it’s already a jetliner in the context of coding assist. It’s correct more often than top hits provided by a top search engine (and any coworker), and it is a very enjoyable user experience (no ads and SEO garbage to filter out). I’d say the wright brothers version was something like BERT or earlier GPTs.
I see that the analogy is having the unintended side-effect of apparently being predicated on a supposed utility of ChatGPT’s direct descendants. The point I want to get across is how difficult anything like it seemed before, regardless of whether it begets anything markedly better afterwards.
Can we stop making analogies like this. It’s so bizarre. The circumstances around each invention are completely different. To always reference the conditions of some past technology is to be stuck in this bizarre form of thinking.
A few days ago some poster tried to compare denying LLM’s to being a naysayer in Gallileos time. Is this all we can do? Make references to past events and fail to evaluate the present properly?
Of course every case of anything is different in detail. On the other hand, identifying patterns is often useful in understanding human reactions to change. Do you disagree with this, or do you think there is something special about inventions?
My analysis here is that the young person in question appeared to focus only on current utility. You have not yet explained why you say it is a failed analysis of why that response may seem surprising.
A lot depends on the definition of "understand" you want to use. According to some definitions there is already some kind of understanding going on in the current generation of LLMs, according to others true understanding requires a soul in the Christian sense and thus it is unattainable an LLM, and of course there is a spectrum of opinions in-between.
If you want to limit your range of opinions to credible domain experts speaking scientifically, the range goes from "definitely not" to "maybe a little if you have a very loose definition".
Which is not that far from what I was saying. And if you take the domain to be philosophy or psychology rather than deep learning or CS, I'm sure the range will be much wider.
