Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

How should we talk to AIs?

blog.stephenwolfram.com

Post reply on HN
Print view View original

41–46 of 46 posts

Re: How should we talk to AIs?

#41

Post by veidr »

dkural wrote: post #27

Wolfram is the ultimate company failing in product thinking. For decades they yell "You can do anything with this!", but fail to identify, specifically, one thing it does better than other products that are designed to specifically do that thing. Thus for anything it could do, other products spend more attention and do that specific thing better/cheaper, with a tighter UX and more content. It turns out most people do…

But you can only really say they are failing, if you are looking at them through a short-term lens. Interviews with Stephen Wolfram give the impression that he is thinking long-term (as in, his entire lifetime and beyond).

Mathematica does actually seem to have dominated its niche for decades. People did and do all manner of things with it.

Wolfram Alpha, their "computational knowledge engine", has only been live for 6.5 years. It's sort of baffling who it is for (but I do, very occasionally, find it extremely useful), but it seems what they are trying to do is bigger than a normal startup that helps you find a taxi or rent out your extra bedroom or search the news.

They may well fail, but I don't think you can say that they have failed yet.

Re: How should we talk to AIs?

#42

Post by hjek »

gooseus wrote: post #15
TeMPOraL wrote: post #4

Earlier quoted context omitted.

It's Standard Wolfram Style. You can see the same pattern in his "New Kind of Science" book - an interesting piece of writing, but the amount of "I", "I've", "my", etc. is plain distracting.

I understand so little of the topics of that book or this review, yet still very much enjoyed reading the takedown: http://bactra.org/reviews/wolfram/ "So much for substance. Let me turn to the style, which is that of monster raving egomania, beginning with the acknowledgments..."

That "New Kind of Science" review was a curious read. Surprising how the absurdity of software patents is mirrored in physics: "Wolfram has since retreated from normal scientific life, in to, on the one hand, tending the Mathematica empire, and, on the other, his peculiar scientific vision and method. The vision is of the universe as, if not exactly a CA, then a simple discrete program of some sort. The method has involved an enormous number of man-hours on the part of subordinates who are, as it were, enserfed to him, scanning the behavior of likely-looking CAs and signing over the rights to their discoveries to Wolfram; their efforts are supplemented by frequent lawsuits and threats of lawsuits against those whom Wolfram feels have infringed on his turf"

Re: How should we talk to AIs?

#43

Post by caligastia »

Wolfram is going off into the weeds. The whole point of NKS is that the universe is digital. We live in a computational universe, therefore we are living and breathing and interacting with natural (not artificial) intelligence as a function of merely existing.

Assuming that by AI, Wolfram is referring to a computational device built by humans, the communication has already happened in the act of designing and constructing the machine. Real communication is over once you flip it on.

What Wolfram is proposing is to then run a Wolfram Language interpreter on top of this new machine and use that to communicate, presumably to ask it for the meaning of the universe. Instead of creating a lingua franca between man and machine, this will prove only that the machine can emulate a von Neumann computer and run the Wolfram interpreter on top. You are NOT communicating with the machine at that point, you are communicating with a parser and a nifty collection of mathematical routines.

Think of a parrot. You teach it to say "Polly want a cracker!" - and the parrot may even learn a few human words - but are you really communicating? Can the parrot express to you the thrill of flying, can it explain to you what a total mind-fuck it is for it to be stuck in a cage begging for crackers when it should be flying with its flock and living its life? No, the parrot can only ask you for a cracker, just the way you taught it.

Man will never have a relationship among equals with a machine, regardless of how much software he pours on top - the machine has a frame of reference that can never be communicated to a human, it's different at the electronic level, never mind the symbolic or linguistic. The machine will have a reality so different from that of the human that even if it could communicate something of substance, it would be as futile as asking your parrot to help you with a math problem - the parrot is not stupid, and in fact can do many things you would find highly mathematical, but it cannot help you with your homework because you are in two different mental universes, no human-invented language can bridge them.

Re: How should we talk to AIs?

#44

Post by dkural »

veidr wrote: post #41
dkural wrote: post #27

Wolfram is the ultimate company failing in product thinking. For decades they yell "You can do anything with this!", but fail to identify, specifically, one thing it does better than other products that are designed to specifically do that thing. Thus for anything it could do, other products spend more attention and do that specific thing better/cheaper, with a tighter UX and more content. It turns out most people do…

But you can only really say they are failing, if you are looking at them through a short-term lens. Interviews with Stephen Wolfram give the impression that he is thinking long-term (as in, his entire lifetime and beyond). Mathematica does actually seem to have dominated its niche for decades. People did and do all manner of things with it. Wolfram Alpha, their "computational knowledge engine", has only been live for…

This is what Wolfram says to their engineers who question it. Wolfram / Mathematica has been around for a while. What's the long-term idea? What exactly will it do better than some other tool in that category? You can't reach there if you don't know where you're going. It's worse than failing - they don't know what they are supposed to be failing at.

Re: How should we talk to AIs?

#46

Post by ddingus »

If it's an AI, then I should be able to talk to it like I would other intelligence.

That is how we talk to an AI. This article is framing a set of powerful computational entities as an "AI", and that is not yet defensible.

So much of what we communicate among humans depends on implied, common context. We know what would make sense to other people, or have a good idea of that, based on our understanding of ourselves and other people.

When I think, "AI", I think about something that knows me other than a set of attributes and rules, and I can know it in the same way. Imagine a little kid type intelligence connected to those powerful compute entities and with a great memory. Better than our memory. And it's fast.

It's that "little kid" type intelligence that is missing! We have the fast, we have the great memory, we have lots of powerful compute entities too.

What we actually don't have is the "intelligence" to complete the idea of "artificial intelligence"

Post reply on HN