# Why ChatGPT doesn't understand **URL:** \ **Category:** Great Debates\ **Tags:** ai\ **Created:** [February 26, 2023, 6:13pm UTC](https://boards.straightdope.com/t/why-chatgpt-doesnt-understand/980404 "2023-02-26T18:13:29Z")\ **Posts on this page:** 1\ **Showing post:** 34
**Author:** ![Half\_Man\_Half\_Wit](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/half_man_half_wit/32/21766_2.png) [@Half\_Man\_Half\_Wit](https://boards.straightdope.com/u/Half_Man_Half_Wit)\ **Post date:** [February 28, 2023, 10:33am UTC](https://boards.straightdope.com/t/why-chatgpt-doesnt-understand/980404/34 "2023-02-28T10:33:53Z")
> [@Dr.Strangelove](#): > > But I wonder. It’s entirely possible for someone to only ever have _read_ of Venus. Or, if they have seen it, not made the connection between the point of light and the object described by the books. How is such a person not in exactly the same position that ChatGPT is? One of the most common ways to address that worry is to appeal to something not reducible to structure, for instance, causality—the symbol is connected to the object because, by whatever means, it is caused by it. But I don’t think that’s a good option, because it gets into problems with misattribution—if, for instance, you mistake a cat for a dog, then in some ways, the ‘dog’ symbol was caused by a cat. But if meanings are due to causal influences, what’s the difference between the ‘right’ and ‘wrong’ instances? Shouldn’t the symbol then simply mean ‘dog or cat’? Another one is the ‘biosemantics’ approach due to, mainly, Ruth Millikan. This identifies the meanings of symbols by an evolutionary account: the symbol X means Y if it’s evolves function is to indicate the presence of Y. But you’re completely right to note that it’s an open problem, generally referred to as the question of ‘intentionality’ in the philosophy of mind. (It was the main motivation for coming up with my ‘[von Neumann mind](https://link.springer.com/article/10.1007/s10670-021-00467-w)’-model.) > [@Dr.Strangelove](#): > > A _cat_ is still connected to a set of properties: four legs, pointy ears, and so on. And each of _those things_, like the number four, is itself a network. Again, these will just be more relation—‘just more theory’. It will do nothing to address the problem (in fact, it will only make it harder). > [@Dr.Strangelove](#): > > You can say you have two, but that’s it. Exactly, and that’s all there is to the objects in the argument, and all that’s needed. The just form a set of n things, and that’s it—the rest, after all, is supposed to be supplied by means of relations. > [@Dr.Strangelove](#): > > I’ll repeat my earlier example: I’ve added Alice-Prime, Bob-Prime, and Charlie-Prime to our universe. They have the same properties as their counterparts. No one within the universe can discern more than three objects: if they try to enumerate them, they’ll never find more than three distinct things. But in fact there are six. If they show up in some relation, they’ll be ‘discoverable’ by those means. If they don’t, they’re also not relevant for anything, and we can just excise them from the domain. > [@Dr.Strangelove](#): > > So how exactly is it possible to determine that there are actually six objects here? And what experiment can be performed to determine whether two people are really talking about the same thing or not? If what you’re talking about includes a structure that is defined over those objects—i.e. if there’s some relation fulfilled by them—then they’ll be just as much part of the domain. If there isn’t, then, again, we can just do without, since they relate to nothing. It’s like saying there’s an electron that doesn’t influence anything else—if we talk about the universe, we can just forget about it. --- _[View the full topic](https://boards.straightdope.com/t/why-chatgpt-doesnt-understand/980404)._