Hear me out. I'm excited about this if it means making shopping better. I want a better shopping assistant. I want someone to do all the work for me (read reviews, do product spec comparisons, price, shipping etc) and just tell me what to buy for 85% of things I purchase online. ChatGPT has this potential, but sucks at it. I'm probably going to buy it on Amazon. Work together and it'll be magic for me.
What makes you think ChatGPT will serve you, and not Amazon?
> "Hey ChatGPT, help me find a USB-C charger that has 3 ports and around 100W power"
>> "Absolutely. Here's [insert product] which matches most of your criteria. I should also mention this is a promoted response. Read on more to see the really best product, which is [another product]"
The bubble pops when Apple releases an iPhone that runs a good enough for most things LLM locally. At that point cloud hardware investments will plateau (unless some new GPU melting use case comes out). Investors will move from nvidia, AMD into Apple.
Based on the recently released graph of how people are using chatgpt. ~80% of use cases (practical guidance, seeking information, writing) could presumably run on a local model.
What's the advantage of that, exactly? Why would you want something very compute intensive run on your phone instead of just using an API to data centers with great economy of scale?
My assumption is that most users won't actually care if the LLM is in the cloud or device. That said, quite a few folks have iPhones and Apple's only way into the AI race is to go to it's strength, 1B+ hardware devices that they design the silicon for. They will produce a phone that runs a local LLM and market it as private and secure. People upgrade every couple of years (lose or breaks) so this will drive adoption. I'm not saying people will vibe code on their iphones.
I've been using Qwen3:32b on a 32GB M1 (asahi) and it does most of what I need, albeit a bit slow, but not slow enough that I´d pay monthly for remote ad delivery.
I suspect this huge splurge of hardware spending is partially an attempt to starve the market of cheap RAM and thus limit companies releasing 128GB/256GB standalone LLM boxes.
The models running on $50k GPUs will get better but the models running on commodity hardware will hit an inflection point where they are good enough for most use cases.
If I had to guess I would say that's probably 10 or 15 years away for desktop class hardware and longer for mobile (maybe another 10 years).
Maybe the frontier models of 2040 are being used for more advanced things like medical research and not generating CRUD apps or photos of kittens. That would mean that the average person is likely using the commodity models that are either free or extremely cheap to use.
ok, you can technically upload all your photos to Google cloud for all the same semantic labeling features as iOS Photos app, but having local, always available and fast local inferencing is arguably more useful and valuable to the end user.
What's the benefit to running LLMs locally? Data is already remote, LLM inferencing isn't particularly constrained by Internet latency. So you get worse models, performance, and battery life. Local compute on a power constrained mobile device is required for applications that require low latency or significant data throughput and LLM inferencing is neither.
30k in a month is an enormous amount of tokens with Claude through AWS Bedrock. And companies already commonly trust AWS with their most sensitive data.
The data you need is mostly not remote. A friend works at a software development company, they can use LLMs, but only local ones (local as in their datacenter) and it can only be trained on their code base). Customer service LLMs need to be trained on in-house material, not generic Internet sources.
The general advantage is that you know that you're not leaking information, because there's nowhere to leak it to. You know the exact input, because you provided it. You also get the benefit of being able to have on device encryption, the data is no good in the datacenter if it's encrypted.
I give this one a big maybe.... What's definite is Cook talks a lot about design but clearly has no handle at all on function or marketing. Everything from the overly scripted, mercurial presentations to examples used in canned videos is all so far from reality it's a caricature of itself.
This felt especially true today at two different points in the video when it felt like Apple was trying desperately to sound like they still had a Jony Ive. They were looking like other companies during the Jobs era cosplaying as Apple.
Long time apple fanboy. I've watched most of these unveilings for the past 20 years. The new phones are impressive. But it was all speeds and feeds. The examples felt so wrong. The women dancing while on the phone. The guy running with while recording. The person needing translation to buy roses? None of those feel grounded in reality. It's like they are building tech for made up in corporate conference room use cases.
Live translation u/i feels like a significant upgrade for that specific product, especially if you are in a dense area with different types of speakers. It feels like a way better MVP than Meta Glasses that are only meant to do troll videos on tiktok or youtube. The examples they trotted out feel ridiculous but that's because they have been approved by their internal systems.
Impossible to oversee. Everyone presenting was so old. It's like they were imagining what was hip and cool. Lets have a women doing salsa facetiming... who would she even be talking to in that scenario??
> It's like they are building tech for made up in corporate conference room use cases.
Totally felt the same during the live-translation demo, when these two casual business folks were talking about "the client will love the new strategy". Dystopian corporate gibberish.
The lack of authentic examples diminishes the impressive tech. Great design is all about function. Why is it so hard to show how this would actually be used in the real world?
Every time I read about a big publicly traded company say something "we're like a startup" I just LOL. Startups win by doing something really disruptive that actually works. You only hear about extreme success. Most things fail. If SNAP allowed each startup squad (10 users) to actually operate like a startup, they'd launch 500 unique disruptive features per month, every month until they hit a growth curve. I'm skeptical that's actually happening.
This is wild. Meta, OpenAI, MSFT, Nvidia are collectively keeping the AI trade alive, which is propping up the stock market and overall perception of the economy. This admission makes it clear that the AI spends are being made up not based on business value/demand...
To be fair, when he suggested $600Bn in spend by 2028, it’s obvious they won’t actually be spending that. That would exceed their yearly revenue each year. They just don’t have that money. This feels like less of an indictment of AI spend, and more of the political process of blatant lying for political favor.
For comparison, Google said $250b, Microsoft said $80b, but Apple has said $600bn. Meta currently spends ~$100bn.
>and more of the political process of blatant lying for political favor.
Even that's debatable because he walked back on the number shortly after
>Once the discussion concluded, Zuckerberg leaned over to Trump to privately admit the president had caught him off guard. "I'm sorry I wasn't ready...I wasn't sure what number you wanted to go with," Zuckerberg said in a revealing moment caught on a hot mic.
And somehow he never lied? Or he wasn't trying to brown nose? Because it literally has to be a lie if he changed his story and it's hard to deny that he appears to be trying to curry favor.
It would be debatable if we didn't have the Commander in Chief we have now is very blatant and open about demanding things and providing political favor in return constantly
But how does it keep that alive? Wouldn't it mean that the marginal utility of each dollar spend would go lower? They are already spending as much as they can and believe is needed since they are diehard AI bulls themselves. If they saw a need to spend 600 billion dollars themselves through 2028, they'd have already done it.
> This admission makes it clear that the AI spends are being made up not based on business value/demand...
Well, isn't that okay? All the companies are racing to capture a nascent market. It would make sense to spend beyond current demand and even projected 5Y demand if it gives you a larger share of a market that might last 20+ years
Agree. This article is completely on point for Hacker News. There seems to be a cabal of Maga-types (or AI) on these forums that reflexively flag any content that seems to smack of exposing the reality of the corrupted and anti-democratic times in which we live.
I believe dang has said there is an auto-flagger which will do that on posts that are getting too many controversial comments. Then a mod has to unflag it, if they want to do that.
Of course this does suggest a strategy for censoring topics that you don't want on the front page of HN: deploy a bot army to add inflammatory comments. I feel like I've seen effect on other posts (the tylenol post currently on the front page has that smell) but who knows.
True story. I grinded hard at a startup for years. This was a decade ago so the concept of 996 wasn't part of the lore yet. But it was fun. We stayed late and I made life long friends. I worked closely with the founder (really awesome dude) as I was an early-ish employee. The company ended up not working, our equity went to zero and we got what you get when you don't get rich, experience. I ran into the founder randomly on the street years later. He didn't even remember my name. He recognized me and was excited to see me, but he had no idea what my name was. So yah, prioritize your life.
I will forget names of people I haven't seen in a long while but whom I legitimately value and am glad to kave known. And I am not a founder type, who are being exposed to 10x people compared to me. It's just a human thing, don't let it bum you out.
Totally. The point is, this guy who's mission I dedicated a good portion of life to clearly doesn't think about me at all after the fact. I'm sure he's gratefuland values me. But if I was part of his story he'd remember my name.
As someone who's just terrible with names, I hate that this is how people interpret it. There are plenty of people who have had a big impact on my life whose names have slipped - and, plenty of inconsequential people whose names stuck for no good reason. It has very little to do with how much that person mattered to me.
Not saying your overall point doesn't stand, but at least for some people remembering a name isn't a consistent indicator of their impact.
same here, names and dates elude me. I can't even remember some of family names and dates, can't remember almost any of the schoolmates names, let alone surnames, but I do remember conversations, moments together etc.
I don't personally have any regrets, because I'm that kinda person. But there are people who prioritized travel, dating, health and personal experiences who have better stories from that period than I do knocking out bugs or shipping a feature that no one ever used. I bought into the mission of the founder. Sure he remembered me when he saw me, but I doubt he thought of me once in the period in between. I remember feeling pretty disappointed after that encounter. Like I bought in, put in the work, it didn't work out and the guy that I followed didn't even remember my name. Prior to that encounter I shared that experience as a badge of honor. After reading this thread and posting here I can't help but think what a waste of time...
