I will not have a stake in this market.
This market resolves YES if someone proves the existence of at least 100 academic papers arguing that smarter than human AI is impossible. It resolves NO if nobody has presented evidence when the market closes.
Academic paper as used above is defined as any paper published in an academic journal.
Evidence could be for example by linking to the papers.
Different versions of the same paper counts once in total.
If an author makes multiple different papers they count 1 each.
Examples could include but is not limited to papers that argue in this vein:
- Godel's incompleteness theorems make it so computers can't know stuff we can know.
- You need feelings to for many types of cognition, and computers can't have feelings.
- Chinese room shows us machines are just doing mechanistic processing which obviously doesn't have consciousness, therefore they will never be able to surpass the human intellect.
- Humans think in stories while computers think in math. The equality operator is symmetric in its arguments, therefore computers can't model causality which is what you need to do planning and take over the world. Humans however have neurons that fire with directionality, thus they can model the universe's causal structure.
- *Says in the abstract that AGI is impossible which means we don't have to worry about it. Then later down in the paper defines AGI as "being literally identical to a human" in some verbose indirect way. Then proceeds to use this definition to argue why AGI is impossible because computers are made of silicon which is not what humans are made of. Then concludes that smarter than human AI is impossible*
- Computers will never be smarter than humans because they need to be programmed, which means they don't have free will. Without free will you have no intentional volition, which is a prerequisite for a moral agency. Clearly if you have no moral agency you don't have feelings etc.
- Computers will never be energy efficient enough to power a smarter than human AI.
- The halting problem shows that computers can't compute everything, this means they can't be smarter than humans.
- Brains are analog and computers are digital, so computers will never capture the nuance and continuity of human thought.
- Evolution took billions of years to make human intelligence. Since computers don’t evolve biologically, they can never reach that level of sophistication.
- Intelligence requires embodiment in the physical world. Since computers don’t have bodies that grow, feel pain, or die, they can never have humanlike cognition.
- True creativity requires "breaking the rules." Since computers always follow rules, they can’t be creative.
- Human thought is non-algorithmic because Gödel/Turing proved limits of formal systems. Since computers are formal systems, they cannot think the way humans do and won't compete with us many domains.
- Computers can only do what they are told (mechanical symbol shuffling), whereas humans can think about what they are thinking, so they are fundamentally different in what they're able to do.
- Consciousness is irreducible, and because computers lack it, they cannot achieve true intelligence no matter how complex they are.
1,000People are also trading
You should probably tighten the resolution criteria a bit by saying "published in an academic journal indexed by (WebOfScience or Scopus or Semantic Scholar or OpenAlex). Pick one depending on how strict you want the criteria for what counts as an academic journal to be. I've listed 4 main indexes, roughly in terms of how stringent their inclusion criteria are.
Note that I have not included Google Scholar, as the criteria for being included in Google Scholar are: a) Be a PDF b) Have a title, list of authors, and a references section c) Be publicly available on the web. As such, there are all manner of things that are not legit academic journals in Google Scholar (and thus in all the major LLMs too, caveat quaesitor).
@WilliamGunn I was just gonna vibe it out. I don't wanna do the work of figuring out which indexes to include, but I can tell you that I won't accept random google scholar slop.
@WilliamGunn I take it that means you think it shouldn't count. Well yeah that makes sense , although if it's an academic work published in a journal maybe it counts as a paper?
@marbinner I think it would make sense to modify the resolution criteria to include other kinds of publications, but a book hosted in a scholarly archive is still a book.
@marbinner A paper is shorter and generally focused on a single specific idea. A book has chapters and is a broader treatment of a subject. You can conceivably have a book assembled from a series of papers, and in fact I've contributed to such.
@WilliamGunn There are many very long papers, way longer than the philpapers paper. And there a very short books that are not delineated in chapers any more than the sections in a paper. There are papers that cover a lot of material and books that are dedicated to a very narrow topic.