“Some reason” is you talk to the thing about anything and it answers and a lot of the time its answers make sense. I’d like to remind you that one year ago this was sci-fi!
Incidentally, "AI" (the original field called AI, not the latest definition of the term) definitely made it to 5. Except these days, we call some parts of it "SQL", other parts "the Web".
I once read an old book talking about AI including something called "expert systems" which are.... human-written if-then-else trees. The saying that we stop calling things "AI" when they become well-understood is real.
I'd definitely say LLMs are stuck on step 3: "Failure to live up to the hype"
They are certainly impressive, but their utility-to-hype/gimmick ratio is incredibly low right now, which could cause a crash. The greater the disappointment the greater the crash.
I'm reminded of 3D TVs. Remember those? Avatar came out in 2009. By 2016 the trend was dead. Despite the cries of "this time it's different." Of course, that time it was not different. The tech was impressive. Much more than the previous time the fad was around in the '80s. Remember the blue/red glasses? Absolutely not a single person talks about 3D TV today.
The 3D TV was a technical success but it was too much of a gimmick that it died out. My Facebook feed is a never-ending stream of AI generated garbage. I think people are going to tire of it, realize the images it makes are about as goofy as a 2004 MySpace page, and maybe it will stick around to fill out the useless corporate email and document bureaucracy and boilerplate framework code monkey BS.
But ChatGPT isn't writing Breaking Bad or The Sopranos anytime soon.
When every LinkedIn thoughtfluencer is writing 3-page screeds about how they're thoughtfluencing through AI you really know you're firmly in step 2.
Likewise when breathless reporters keep asking non-AI companies what their AI strategy is, you know you're firmly in step 2. Remember when Walmart was expected to have a "metaverse strategy"?
Also worth noting that many (most?) technologies do not have a step 4 or 5. They're just permanently/indefinitely dead after the hype train goes off the rails (see: personal jetpacks)
Something being overhyped implies grifters will appear but grifters don't imply that something is overhyped. I'm sure people were the same with cars and transistors. People definitely were breathless about smartphones the same way when they came out (everyone was selling apps that did exactly the same thing as websites) and they've changed the world for better or worse.
Crypto and AI both attract get rich quick bullshitters but I think AI right now is actually a crazy unexpected sci fi tech while crypto wasn't good for anything except fraud, gambling and the black market.
I partially agree. I think the last giant hype cycle around crypto had no pants on the entire time. It was all one giant delusion, and also why I don't think crypto will have step 4-5. It's just dead outside of incredibly niche uses (many of which are crimes).
I also agree that there's something there there with LLMs... but also that it's hopelessly overhyped right now.
Smartphones are a good example of this - nowadays we tend to think about iPhone or BlackBerry as the start of the smartphone "era" - but that wasn't the actual start of smartphones.
The first smartphones were called PDAs, and there was a hype cycle around that! Lots of companies wanted in! But adoption was abysmal and the whole thing fizzled out. BlackBerry and iPhone were the steps 4-5 of that cycle.
The state of LLMs right now is the Palm Pilot. Whiz-bang. Cool. Tons of press. Lots of imagined applications and attempts at mainstream adoption - but honestly nowhere near good enough to achieve mainstream breakout. Died a slow death without fulfilling its most lofty promises, and the space was relegated to a niche status until the actual entrants arrived to actually achieve mainstream success.
I think LLMs will have a step 4-5 with actual mainstream success. I just don't think the current players are it, and also that the vast majority of the current players have no pants on and are just pure grift.
Yeah, I think "AI" as a term has always been super compromised to the point of uselessness. ML in general is firmly in steps 4-5 - it's integrated into our lives in so many places and generally without the users having to think about it.
Car crash detection, automatic photo editing, heart rate sensing, etc. We use this stuff daily but there's generally little hype about the underlying tech (though some hype about specific applications).
What's in step 2 is "Generative AI", which IMO is also a misnomer for "large language models". The viability and uses of these models is far from proven out yet.
The LLM hype is maybe blinding us to all the other use cases that the new powerful GPUs will provide. Maybe the real progress was the ability to train increasingly large models. I don’t think LLMs will solve most problems but there will
be other models that can learn from their success
This is why I hate the term "AI". Ai pathfinding, ai upscaling, and generative art are completely different pieces of technology that falls under the same marketing term. Strictly speaking, the decades of machine learning that's been going on in acedemia is also all "AI" as well.
I don't think Jetpacks are even in step one. Tech's been there but I don't think any are commercially available. They've certainly been hyped for decades, but more in the same way most sci-fi concepts are glamorized in general. Worldwide instant communication was very much a sci-fi concept 50 years ago, and today it's definitely not all rainbows and butterflies
Jetpacks aren't here because they're fundamentally incompatible with real-world society. Imagine giving a typical driver a jetpack or a flying car, and then think about how a traffic accident would play out.
Or in other words, we already have flying cars - but the form of a flying car that's compatible with reality is called a helicopter, and piloting one comes with a fuck ton of expensive hoops to jump through.
That's the overall problem with all the cool sci-fi tech - it's cool in an action movie, when the protagonists are the only ones who get to use it. It stops being cool and becomes either useless or dangerous, once every rando gets to use it in their daily lives.
> Jetpacks aren't here because they're fundamentally incompatible with real-world society. Imagine giving a typical driver a jetpack or a flying car, and then think about how a traffic accident would play out.
Oh yeah, imagine a transportation technology that killed people every week. No way that would be legal. Except if it's cars, for some reason they magically get a pass.
> Or in other words, we already have flying cars - but the form of a flying car that's compatible with reality is called a helicopter, and piloting one comes with a fuck ton of expensive hoops to jump through.
We could get rid of those hoops and flying cars would still have a lower death rate than the regular kind. But they can't replicate the "our oopsies are someone else's problem" field that cars have. That's the hard part.
> Oh yeah, imagine a transportation technology that killed people every week. No way that would be legal. Except if it's cars, for some reason they magically get a pass.
Imagine a transportation technology that killed orders of magnitude more of people every week. That's the reality if you just magically s/car/jetpack/g for everyone.
> We could get rid of those hoops and flying cars would still have a lower death rate than the regular kind.
Not really. Driving a car is trivial compared to flying a helicopter; the hoops in question are mostly about ensuring pilots are properly trained (vs. half-ass bullshit trained, "you'll learn the real thing on the road" that is getting a driver's license) and actually meet some health standards. Number and difficulty of hoops differ in various areas of aviation, but they all recognize just how much easier it is to kill yourself with an aircraft, and how much more death and destruction an aircraft can cause.
> Imagine a transportation technology that killed orders of magnitude more of people every week. That's the reality if you just magically s/car/jetpack/g for everyone.
Where is the problem: those people who don't have this risk affinity don't need to buy/use a jetpack. Similarly, not everybody should go ice climbing or BASE jumping. Thus I see no reason to outlaw jetpacks just because of their danger.
They're not outlawed per se. They just don't make sense at the intersection of economics and safety regulation, which is why you don't see them outside some experimental work.
