0

I perform the same tasks an AI Bot does when answering questions here: I go online and make sure that my answer can be validated online as a self-check. Is there a practical difference between me and a current AI engine with regards to my activity as a Stack Exchange member? How philosophically is there a distinction in those two types of language contributions?

17
  • 1
    There are already bots on the stock exchange, so why not the stack exchange? They're identifying cancer, flying planes and driving trucks. Philosophy can't be that hard.
    – Scott Rowe
    Apr 5, 2023 at 1:57
  • 1
    I did ask some philosophical questions to ChatGPT. The answer looked coherent on the surface, but the logic was wrong in the end - just the logic steps.
    – Frank
    Apr 5, 2023 at 2:32
  • 1
    Let's experiment - take some questions from here, try them in ChatGPT, and see if you like the answers.
    – Frank
    Apr 5, 2023 at 2:33
  • 1
    This seems definitely a Meta question, and many related questions already exist in Meta sites.
    – tkruse
    Apr 5, 2023 at 5:08
  • 1
    I've revised the question to explore the basic philosophical distinction between AI and human language production.
    – J D
    Apr 7, 2023 at 15:08

4 Answers 4

1

I think that there is no difference. All living beings are made up of inanimate particles (electrons, quarks) and so are the non-living beings. You are asking because the universe has evolved to the state where you are asking. You don't have any free-will. We are sort of "programmed" by the universe. This implies that the AI bot certainly doesn't have free-will (which is programmed by us; we are sort of "programmed" by the universe). In fact, we are just "puppets" controlled by the laws of physics.

1

I perform the same tasks an AI Bot does when answering questions here: I go online and make sure that my answer can be validated online as a self-check.

Current AI technology may produce language, but it does not do it like the human brain and the very question of 'validation' is what distinguishes people from such technology. AI researchers for decades dreamed of having a system that thinks like people (see GPS), and one such formulation of the problem is the physical symbol system hypothesis and it's closely allied argument of the existence of the LoT or language of thought (SEP). But all attempts to create working software along these lines has led to the framing problem, and the second generation of cognitive scientists has led to counter arguments against the standard cognitive science model such as connectionism and embodied cognition (SEP).

Currently, OpenAI's success with it's current version of ChatGPT LLM (and others like Google's Bard based on LaMDA) has people again wrestling with knowledge about narrow AI and questions about AGI as a hypothesis with persons such as Altman and Goetzel (and other futurists) stoking public hopes and fears that machine systems are about to rise up as sentient beings or that the singularity is about to occur. Happily or sadly, depending on your view, AGI is nowhere near coming to fruition and people still struggle with fundamental critical thinking skills, such as Blake Lamoine, who makes claims that hold no philosophical water.

Is there a practical difference between me and a current AI engine with regards to my activity as a Stack Exchange member.

Yes, there is very much a practical difference. You are a person, not a LLM, and your contributions therefore are not marked by semantic hallucinations. When you contribute your language skills exhibit common sense, where as an LLM's do not. One way to think about this is that you can be distinguished from an LLM with the use of the Turing Test precisely because you are capable of common sense or exhibit other language usage indicative of psychological regularities such as cognitive biases. And a philosopher of mind has an additional tool box that a talented clinical psychologist may not, a vocabulary to describe and understand how minds work metaphysically. While some people are skeptical contemporary philosophy has any value, in fact, contemporary philosophy is more sophisticated than it has ever been as more and more philosophies-of come to maturity.

5
  • It's very analogous to language translation apps. Google translate has become effective enough to for me to send emails in German without the receiver really knowing if I can speak German or not. I believe the big "language" issues of AI will come from language translation algorithms
    – user64314
    Apr 7, 2023 at 15:38
  • 1
    It's the ability to mimic behavior that is concerning. It's ability to be anonymous. These are tools that can be used for sinister behavior without the Bot passing a Turing test
    – user64314
    Apr 7, 2023 at 16:03
  • It is concerning, because like nuclear energy, there are potential abuses that arise. People struggle to be rational, and giving one group of people a tool that creates the appearance of human beings where human beings exist not blurs the decision making process epistemologically for people who distinguish between reason and testimony when making choices. People will do for other people what they will not do which appears unreasonable otherwise. When computers can stand in for people, then what constitutes as testimony becomes dangerous when the motivations of tool users are questionable.
    – J D
    Apr 7, 2023 at 16:11
  • Indeed. In fact the differences are still quite large in my experience. I spent several hours with ChatGPT asking it questions about logic and it produced a load of self-contradictory nonsense. It was like talking to someone who is insane: it literally could not hold a coherent thought from one minute to the next.
    – Bumble
    Sep 3, 2023 at 5:34
  • @Bumble Given your perspicuity, I'd expect it. The logic of transforming symbols by semantic grounding you've internalized is fundamentally different that the statistical methods of token prediction which is, even with billions of data points, a shallow syntax processing. You, kind sir, are a living validation that meaning in natural language demonstrates semantic internalism. 'Meaning' is a universal which is more than mere utternace construction. It is a computation that draws from learning from the limbic to the neocortex; it is an exciting time that AI systems will continue to grow.
    – J D
    Sep 4, 2023 at 17:05
0

I lt looks like we are unable to distinguish ...

enter image description here

from

Mount Everest

0

I put a cryptic crossword clue to ChatGPT. Not only did the gormless bot suggest an incorrect answer, the answer suggested had seven letters instead of the six I had specified in my question, an oversight that ChatGPT was quick to acknowledge when I raised it. Perhaps if I upgrade to the paid version it will be better at figuring out how many letters are contained in a word. In the meantime, I will not be relying upon it for philosophical insights.

So the answer to your question is yes, there is a key difference. ChatGPT immediately recognises its mistakes when they are pointed out to it, whereas humans on stackexchange will stubbornly defend their delusions.

You must log in to answer this question.