We all know that PDF's won't stop AI, and that everything created and put online will inevitably be used by AI. Hell, before we know it, almost everything in the physical world will be recorded/streamed and processed and used by AI.
1. I call that using a statical model to determine the most likely next character in a text string. The amount of time it takes AI to do that is irrelevant to the question of reasoning.
it doesn’t have to be… it just has to be better than humans at doing the select jobs we ask of it. The improvements in AI over just the past couple of years has been gigantic. There is no reason to believe that these improvements won’t keep on happening.
I didn't indicate that I believed that. In fact I wrote "...it may well be able to get past its current problems with PDFs..."
So you're merely arguing semantics... fair enough... I suspect this will be moot in a decade or 2 as whether an AI is "reasoning" will cease to be a concern. It will simply be.
I call it copying. That's all AI is capable of doing at the moment. It doesn't have the capacity to come up with novel solutions if someone else hasn't already done so. What it can do is the same thing computers have done already at a lower level: assemble facts and answers from disparate origins and put them together in a way the original thinkers didn't have the sources to do.
I don't think you understand the word "reason" at all. Just because we are astonished at what AI has been able to do is not remarkable nor is what AI doing "reasoning" the way a human would go about it.
Truly said. Much of what humans do for work is drudgery, it's the same argument that was made for things like vacuum cleaners and dishwashers. None of that makes AI suitable for doing things that require sophisticated thinking which includes things like emotional value, ethics, and long-term considerations. If I do "A" and "B" doesn't like it, what are the likely outcomes and will those be good for society or the world?
Again, a very specific claim for AI was made. The other poster first asserted the imminent capability of AI to extract information from PDFs, and rather than stopping there, they went on to make a second, bigger claim about its capacity for reason. They obviously believed it to be an important enough matter to be worth making a claim about.
I don't think semantics is irrelevant. I do think most people use language without understanding what they are doing in their heads. AI is never going to equal the quality of well-informed human thinking. It might be able to equal the average not well-informed person, which is why so many entry-level jobs and tasks are ripe for takeover. People are just going to have to get better education before even reaching college age. If your elementary and secondary schools are not up to the task, then woe betide you and your progeny.
I don’t think that poster meant what he typed… but, neither of us can read minds… so semantics it is…
It might be an interesting subject - but isn’t relevant to reading PDF documents…
That's a complete non sequitur. You start by saying you don't think they meant what they said, then admit that you cannot know whether or not they meant it, then assert that my argument is therefore a semantic one. How on earth does that work? If someone says something wrong without meaning it, it's still wrong, and whether or not they "meant" it has no bearing on whether or not a given counterargument is based on "semantics".
I’m merely stating that the use of the word “reason” wasn’t correct… he probably meant to say something like “decipher”… his point, as I understood it, was that AIs will be able to decipher, figure out, etc, PDFs far better than humans in the near future.
Them: Within 1-2 years AI will be able to extract information from PDF documents better than the average human and the sun will rise in the west.
So how is any of this relevant?
A good chunk of PDF utility is understanding the logic behind the decision making. "Why was this font chosen, in this size? What formatting is this? Why are there static rows instead of table separators?" A lot of complicated interconnected structures are maintained by the PDF format. Because that's the point: the first word in "PDF" is "portable" for a reason―they are designed to be platform-agnostic, requiring the bare minimum external libraries to interact with. Not everyone has a licensed copy of Microsoft Word on hand, but just about every device made in the last 30 years can open a PDF.
No, I'm living in reality.
That depends on the time scale. These data centers are being built on the back of the "promise" of AI, its supposed 100x ROI. They are built with VC funds and there is still no product. A product generates revenue and these models do not generate revenue―certainly not enough to justify their upkeep. The problem, that no developers have solved thus far is, "what do people do with our models, that generates profit?" Like, Grok and Claude have a lot of utilities, but do those utilities even begin to make enough money to cover their operating costs? Or are they basically subsidizing an unprofitable venture?
I don’t think you are… let’s address your inanity, point by point, shall we?
It isn’t too difficult - and they can read the vast majority of PDFs… and they’re improving day by day. What they couldn’t do last week they might be able to do by next week… whereas if the average human can’t do something, it will remain that way for millennia to come.
It’s not false advertising… results continue to occur - at a far greater success rate each day.
AI is still in its infancy. I know we are in a culture of “now, now, now!”, but reality is a harsh mistress. Rome wasn’t built in a day - neither is AI.
We have already crossed that line. And we WILL reap the benefits. We already are in many fields - medicine being a huge one in my books
