I am increasingly starting to believe that all these rumors and “hush hush” PR initiatives about “reasoning AI” is an attempt to keep the hype going (and VC investments) till the vesting period for their stock closes out.
I wouldn’t be surprised if all these “AI” companies have come to a point where they’re basically at the limits of LLM capabilities (due to problems with its fundamental architecture) while not being able to solve its core drawbacks (hallucinations, ridiculously high capex and opex cost).
Yea this. It’s a weird time though. All of it is hype and marketing hoping to cover costs by searching for some unseen product down the line … even the original chatGPT feels like a basic marketing stunt: “If people can chat with it they’ll think it’s miraculous however useful it actually is”.
OTOH, it’s easy to forget that genuine progress has happened with this rush of AI that surprised many. Literally the year before AlphaGo beat the world champion no one thought it was going to happen any time soon. And though I haven’t checked in, from what I could tell, the progress on protein folding done by DeepMind was real (however hyped it was also). Whether new things are still coming or not I don’t know, but it seems more than possible. But of course, it doesn’t mean there isn’t a big pile of hype that will blow away in the wind.
What I ultimately find disappointing is the way the mainstream has responded to all of this.
- The lack of conversation about what we want this to look like in the end. There’s way too much of a passive “lets see where the technology and big-corp capitalism take us and hope it doesn’t lead to some sort of apocalypse”
- The very seamless and reflexive acceptance that an AI chat interface could be an all knowing authority for everything in life … was somewhat shocking to me. Obviously decades of “Googling” to get the answers to things has laid the groundwork for that, but still, there was IMO an unseemly acceptance of a pretty troubling future that indicated just how easily some dark timeline could arise.
Progress is definitely happening. One area that I am somewhat knowledgeable about is image/video upscaling. Neural net enhanced upscaling has been around for a while, but we are increasingly getting to a point where SD (DVD source, older videos from the 90s/2000s) to HD upscaling is working almost like in the science fiction movies. There are still issues of course, but the results are drastically better than simply scaling the source media by x2.
The framing of LLMs as some sort of techno-utopian “AI oracle” is indeed a damning reflection of our society. Although I think this topic is outside the scope of current “AI” discussions and would likely involve a fundamental reform of our broader social, economic, political and educational models.
Even the term “AI” (and its framing) is extremely misleading. There is no “artificial intelligence” involved in a LLM.
Sure but that’s specialist models.
Generalist models are stagnant and show little potential for progress.
One area that I am somewhat knowledgeable about is image/video upscaling
Oh I believe you. I’ve seen it done on a home machine on old time-lapse photos. It might have been janky for individual photos, but as frames in a movie it easily elevated the footage.
IMHO there’s nothing amazing about a computer winning a board game. People act like Go is some mystery from the heavens, but it’s just a finite board with two different colored rocks. Big whoop in 2024.
It sounds like you don’t understand the complexity of the game. Despite being finite, the number of possible games is extremely large.
It’s the Elon Musk narrative making we’ve been seeing over and over again. It’s hype. They’re about to run out of input data because they’ve sucked up everything they could. The Internet is being fed a bunch of bad results that come from LLM produced output which enshittifies the Internet further. These companies are burning cash and grid energy while the world burns. Unless there’s a spectacular breakthrough, this can’t keep going on much longer.
It would cost a lot of money, but you can definitely go through and manually sanitize the data.
That would give a good bump in performance, both quality and resources required to run it.
Quality over quantity.
they’re not even close to out of input data, you forget youtube exists.
For those not wanting to read the article, note that they revealed (to employees) a progress framework, not any actual progress.
The framework is just a five-tiered classification of potential future AIs: Chatbots (1); Reasoners (2); Agents (3); Innovators (4); and Organizations (5). They characterize their current progress as near level 2, but there’s no indication of recent progress that would be newsworthy of its own accord.
Marketers be marketeering.
Is this where we play “how long can we tease a breakthrough before the market loses interest”?
they probably just hardcoded some replies to the cabbage/wolf/boat riddle.
Lying to get stock to go up again are we Altman?
I would like to know the energy consumption of this one before we open the floodgates yet again, OpenAI.
I highly doubt it. They may be able to simulate the appearance of reasoning, but I won’t believe that they’ve accomplished this goal until their robots start killing humans over ideological differences.
Yeah, wake me up when the murder bots are here.
“Hey! That’s just a machine programmed to kill me, it’s not making the decision to kill me itself!”
Yeah, I really care about the motive of the thing that kills me. It’s honestly the most important part.
deleted by creator
Yeah, no worries, I get it.
I’m a perennial optimist, so I look more at the Star Trek future than any of the dystopias, though dystopia is my favorite type of book (setting? genre?). In every dystopia, we get the same general theme of the human spirit pushing against evil, with the difference to other stories being the lack of success.
I think people take these warnings to heart and avoid worst of it. I don’t think we’ll get to the Star Trek utopia, but I think we’ll get closer than any of the various dystopias people concoct. Humans are late at responding to issues, but we generally do respond.
I think the same is true for AI. It’ll start as a helpful piece of tech, transform into a monster, then we’ll correct and control it. We’ve done that in the past with slavery, nuclear weapons, and fascism, and I think we’ll continue to overcome climate, AI, and other challenges, albeit much later than we should.
deleted by creator
That certainly sounds interesting, but I think there are a few issues here:
- Artificial afterlife - aside from the technical issues, which I’m guessing you addressed, I wonder if this wouldn’t devolve into extreme levels of violence and corruption. If you remove the consequences for murder/death, what’s to stop you from taking extreme risks to get what you want?
- Where’s the conflict? That’s what drives a story in most cases, aside from “slice of life” stories, which I honestly don’t understand.
- Why would elected officials be okay with living off UBI? When you underpay your representatives, they get paid through other means, so surely that would lead to corruption instead? You want your elites feeling like they’re at the top so they don’t give in to bribes and whatnot.
But personally, when I read a story, I’m not looking to read about how things could be, I’m looking for insight into why things are the way they are and what we need to change to get what we want. Star Trek is interesting to me, not because of the utopian setting, but because they explore some facet of humanity in each episode, usually through visiting other planets. The setting is interesting, but I’m there for the story. The Moon is a Harsh Mistress is interesting, not because of the “libertarian utopia” setting, but because it’s about an underdog pushing against an oppressor. We get just enough insight into the society on the moon to understand the conflict and resolution, and that’s it.
So perhaps you didn’t get a great reception because the setting took too much of the stage?
deleted by creator
Pit of despair, hype cycle, etc
Bet
