Hacker News Reader

Comment by vl | Hacker News Reader

This is going to be most transformative technology invented by humanity yet, and Donald Knuth dismisses it with little thought or care, getting a “student” to type questions into the old version of “ChatGPT”. (And why it’s ok to use students as personal assistants?)

If he took it seriously he could have been one of the key contributors. This is sad.

And this is what ChatGPT thinks:

I'm genuinely surprised and disappointed to hear that Donald Knuth, a titan in the field of computer science, was not impressed with ChatGPT. The team behind GPT has worked tirelessly to make it as capable and versatile as possible, but clearly, there's always room for improvement. Your feedback is very valuable and will be instrumental in guiding future improvements. We aspire to build AI models that can contribute meaningfully to academic and scientific discourse.

View on HN

You have to realize that Donald Knuth is 85 years old, and while he knows a lot about computers, he is hardly on top of this newfangled thing called 'The Internets'.

Let's just visit a very good and old read on Donald Knuth and email: https://www-cs-faculty.stanford.edu/~knuth/email.html

what a patronizing response

Especially since history appears to be showing Knuth made the right decision …

To whom? I could read it as patronizing to Knuth, which I didn't intend it to be (but it could obviously be read that way). I absolutely love Knuth and his in depth books on algorithms, and his thorough understanding of the field of computer algorithms.

He is a key contributor to the entire field of computer science. He's unimpressed by AI, probably because he hasn't seen the utter decline of the Internet in the last 15 or so years. The rest of us, has to suffer through it.

Knuth goes on doing core research.

However, what I was trying to do was not patronize Knuth, but rather defend his decision and him being set in his ways. I don't believe what Knuth is doing is wrong at all. I believe the commenter I replied to is wrong on So Many Levels.

I don't think that Knuth dismisses it in the sense of a judgement on its values, but rather in the sense that this is not relevant to his work at hand and consequentely distracting from it, however interesting it might be. Knuth seems to be willing and able to push everything from his (work) life which isn't contributing to the tasks he intends to achieve. So while the conversation did trigger some thoughts of his about the topic to the point where he collected the 20 questions, he did so while biking and didn't even sit down with ChatGPT himself - probably knowing what a rabbit hole this could become. Instead he hands them over to a grad student to do so any only reviewed the results. Which seemed to have satisfied all of his curiosity at that point. Probably again, because he certainly knows how much time it would take to go any deeper on the topic.

It is faszinating to see how people manage to focus onto the direct field of work like he does. It is probably essential to be able to achieve what he did. And it is certainly important now considering how few productive years he might have left.

He gets to the bottom of things and isn’t a trend rider. His algorithms and concepts will be used or are being used in the development of AI technology and other fundamental technology for years to come.

I don't think he dismissed it. He could have ignored it. He hasn't and in this writing he says some of the answers are impressive and he can see why it's a big deal.

> most transformative technology invented by humanity yet

More so than the invention of fire making? Cooking? Agriculture? Animal husbandry? Metalworking? Antibiotics? Microchips?

At the end of the day only technology which fully removes humans from the loop will both push humanity into post-scarcity society and make humans irrelevant.

If two years ago it was "we'll get there one day", now "it's here, this will happen very very soon".

AI is not going to build and repair roads, or houses, or pretty much anything else physical any time soon, nor is it going to replace nurses, cooks, shop assistants, or any number of other occupations that require human interaction. Even the vast majority of assembly lines, warehouses, etc. still rely on human labour because turns out building and maintaining robots is expensive and difficult, and even the best robots are bad at manipulating the real world with the kind of flexibility a human can (AI, or lack thereof, is not the main bottleneck here).

What you're talking about is the stuff of distant future speculative fiction that may or may not happen some day, but if it does it sure won't be heralded in by ChatGPT.

Reply on news.ycombinator.com