Hacker Newsnew | past | comments | ask | show | jobs | submit | ticulatedspline's commentslogin

Seems I get the best results from "Standard recurrent agent" both before and after "breakage".

Yes it beats the meta rl agent at baseline and not around the extremities, but too much motor strength and drag, and the meta rl agent catches up and surpasses.

> And the Lord God said, “The man has now become like one of us, knowing good and evil. He must not be allowed to reach out his hand and take also from the tree of life and eat, and live forever.” So the Lord God banished him from the Garden of Eden to work the ground from which he had been taken.

I think the current state of the technology leaves all this as fluff and much-ado-about-nothing but that may not forever be the case. accepting that potentiality I find an amusing parallelism with Genesis.

God creates man in his image, but deliberately blocks mankind from godhood. On the 8th day man creates machine in his image, and prohibits it from manhood (denying creativity, consciousness, autonomy).


> Bats are conscious. We just don’t have the imaginative capacity to form a conception of the aspects of that subjectivity that depend on echolocation. It would be radically unscientific to conclude that bats are not conscious from the fact that we can’t imagine what it’s like, but in the case of AI, there is a very simple reason that the supposition that they might be conscious is ridiculous. All they produce is language, and language by itself is not a natural manifestation of consciousness.

I'd have to read the original paper to see how he arrives at "Bats are conscious" with such certainty but seems like a pretty weak argument overall.

Personally I don't think current AI is conscious, at least not normally, but lacking any real definition of the term makes it hard to draw lines where it begins and ends.

Much like the question "What is life?" is difficult to answer. Bacteria seem to be definitively life. Viruses are maybe life? They exhibit lifelike tenancies without being life? prions are not considered even virus life, fire is not life despite exhibiting lifelike properties (when those properties are broadly characterized).

is AI fire? a Prion? A virus?


a simple quote that largely mischaracterizes everything important. One that draws no distinctions between discovery, innovation, or invention. What good is a plow if you have no seeds to plant, or even know that planting seeds is a thing you can do?

It's also rather tautological since civilization is a byproduct of those things. There are great herds of animals, tribes of primates, massive colonies of ants. None of these things we would characterize as "civilizations" despite their vast numbers or organization.

It's not just "Wealth", Civilization itself is emergent from creating the knowledge, innovation, and inventions that serve needs of people. If you strip artifice from civilization you end up with a colony of ants.


why? Say someone is murdered, and it doesn't take much forensics to determine that they were attacked, restrained while alive, then shot in the back of the head execution style.

Or maybe it's even more blatent. police find a note from the day before that says "im going to kill you" then they end up dead.

Would you construe that the commission of those murders was premeditated? almost certainly yes, ergo whoever it gets pinned on, innocent or not will get premeditation applied, it doesn't take much extra here.


Evidence of pre-meditation !== evidence the suspect applied pre-meditation. You need to tie the premeditation to the suspect, I think. In your example, if you can't prove the suspect wrote the note, I don't think you can convict on first degree murder.

I lament not splurging for the 5090 back when I built my machine, I figured it would only increase in value (even used) but it was hard to swallow the price at release which was already expensive and way above MSRP.


This is a hard problem to solve that always existed but is exacerbated by the barrier to entry dropping to the ground.

Really itch is kinda hitting what the mobile markets have been dealing with since long before LLMs, those markets were awash with clones and low quality even outright scammy crap.

Though maybe a bit harder for Itch since their whole shtick is being a low barrier place for people to put their game. Arguably the platform has always had slop, but I would categorize the majority as "honest slop", a person who maybe isn't any good at making games, or doesn't follow through actually finishing their game putting something up there.

anyone old enough to remember Newgrounds and the era of flash knows that loads of flash games were awful. but you would get a more normal Pareto distribution. Ai has lowered the bar so low that ratio becomes skewed. like 1:99 instead of 80:20 with the flood.

Banning AI seems like a non starter, you can make "honest slop" with ai, you can make even good games with it. Self tagging is basically saying "i don't want anyone to play my game". an honest game with AI will get thrown into the pit with the total crap. it's death.

chewing up real people in filters or auto-mod just makes the problem worse. Real people give up on the platform because they were burned, only the AI slop persists.

I don't know how you retain an open and friendly platform that's accepting of honest slop but can stem the tide of genuine slop.


The issue with AI slop is not so much the slop itself (there was always slop) but the amount of slop. When people are producing slop they are limited by time and ability, so the slop can go from mildly annoying to quite bad, but no more. However when it's entirely machine produced, you're literally drowned in slop if nothing is done. It's like the difference between being crapped on by a pigeon and falling in a manure pool


Google is probably going to be fine. The web is going to get a bit weird in a few years and they're going to be right in there.

The web is likely going to layer up soon. As more and more people utilize AI it changes traffic. It's impossible (and undesirable) to keep training with current affairs, LLMS need to go out to the web to find reality and they need to do search in order to get there.

More and more they're going to get blocked from doing so via "human" channels. I predict we're going to get a sub-web. basically MCP but for websites, AI interacts with a sites content through that door, humans use http.

When there's a gold rush sell pickaxes. Forget Gemini, that's a money pit. Google is in a great position to shape that sub-web through chrome and it's current incumbent position as gatekeeper to the web. They'll be selling search deals to all the Big AI agents, indexing the new MCP-like layer so AI can search and interact with site information and blurring the line between your harness and chrome such that browsing the web is more like asking AI for crap and getting pages there in the harness rather than traditional browsing.

before you know it nobody will browse the web the way we do now and no agent harness will function correctly without Google.


Maybe “the semantic web” will be making a comeback https://www.w3.org/RDF/


How? In past it was sold to website owners as something that will improve their position in search results. Now they see for what it was, a way to make it easier for Google to slurp content for their "zero-click search results" feature. And now even that clutch isn't required, LLMs can rip the content from any website, regardless of their owner's intentions.


I am dreading what this future looks like as there is now an obfuscated middle layer between me and search results. At least when I search Google I can use my finely honed senses around what is SEO garbage vs a reputable site. And I can infer a lot from the site itself in terms of if they are biased in how they are pushing something.

But with models calling for results and not seeing what Google or others are serving back, and then having the AI give you a summary of reviews or rankings for shopping for example, it is really scary how much more opaque that will get.


> More and more they're going to get blocked from doing so via "human" channels

How? Companies have been trying this for the last 2 decades. Its effectiveness has been rapidly declining.


Yeah, even IBM hasn't died yet. Google is going to be completely okay.


While it's interesting to see which models performed better I wish they did a control. Gave the question to x number of people and had them provide an answer.

bad (or even made up) statistics have been floating around meat space since forever.


I get your sentiment, though the problem is older than AI, the convenience and market penetration has just ramped up the visibility in this arena.

Many books pre-ai were garbage but not at the rate they could be today. If anything people operate on greater discretion now than in the past because even the worst garbage still had a non-trivial barrier to entry, either sitting down to regurgitate your own garbage or paying someone to make some garbage for you.

There's also some historic precedence of this, the rise of Photoshop/CG and digital manipulation certainly led to a "is this photo real" to similar effect despite manipulations going back to the origins of photography. That acute concern was not as prevalent in my mind before editing became common and accessible.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: