111–120 of 218 posts
Earlier quoted context omitted.
I spent 13 years chronically on reddit before stumbling into a exit hatch of the bubble chamber. Those people (well really it's teens and college kids) live on reddit, they are so far from an accurate representation of reality its insane.
Worse when you find out there’s a couple dozen of the same moderators running nearly all the top 500 subreddits.
The one thing I don't understand is this assumption that demand for GPUs for training is going to keep growing at the rate they grew so far. I get the demand for new applications, which require inference , but nowadays with so many good (if not close to SOTA) models available for free and the ability to run them on consumer hardware (apple M4 or AMD Max APUs), is there any demand for applications that justify a crazy…
Of course, cern is still going to use their FPGA hyper-optimized for their specific trigger model for the LHC, and apple is gojng to use a specialized low power ASIC running a quantized model for hello Siri, but I meant the majority usecase.
The one thing I don't understand is this assumption that demand for GPUs for training is going to keep growing at the rate they grew so far. I get the demand for new applications, which require inference , but nowadays with so many good (if not close to SOTA) models available for free and the ability to run them on consumer hardware (apple M4 or AMD Max APUs), is there any demand for applications that justify a crazy…
Earlier quoted context omitted.
AGI is not near. At best, the domains where we send people to years of grad school so that they can do unnatural reasoning tasks in unmanageable knowledge bases, like law and medicine, will become solid application areas for LLMs. Coding, most of all, will become significantly more productive. Thing is, once the backlog of shite code gets re-engineered, the computing demand for a new code creation will not support bu…
> AGI is not near There's also plenty of argument to be made that it's already here. AI can hold forth on pretty much any topic, and it's occasionally even correct. Of course to many (not saying you), the only acceptable bar is perfect factual accuracy, a deep understanding of meanings, and probably even a soul. Which keeps breathing life into the old joke "AI is whatever computers still can't do".
I think the main problem with AGI as a goal (other than I don't think it's possible with current hardware, maybe it's possible with hypothetical optical transistors) is that I'm not sure AGI would be more useful. AGI would argue with you more. People are not tools for you, they are tools for themselves. LLMs are tools for you. They're just very imperfect because they are extremely stupid. They're a method of forcing a body of training material to conform to your description.
But to add to the general topic: I see a lot of user interfaces to creative tools being replaced not too long from now by realtime stream of consciousness babbling by creatives. Give those creatives a clicker with a green button for happy and a red button for sad, and you might be able to train LLMs to be an excellent assistant and crew on any mushy project.
How many people are creative, though, as compared to people who passively consume? It all goes back to the online ratio of forum posters to forum readers. People who post probably think 3/5 people post, when it's probably more like 1/25 or 1/100, and the vast majority of posts are bad, lazy and hated. Poasting is free.
Are there enough posters to soak up all that compute? How many people can really make a movie, even given a no-limit credit card? Have you noticed that there are a lot of Z-grade movies that are horrible, make no money, and have budgets higher than really magnificent films, budgets that in this day and age give them access to technology that stretches those dollars farther than they ever could e.g. 50 years ago? Is there a glut of unsung screenwriters?
Earlier quoted context omitted.
AGI is not near. At best, the domains where we send people to years of grad school so that they can do unnatural reasoning tasks in unmanageable knowledge bases, like law and medicine, will become solid application areas for LLMs. Coding, most of all, will become significantly more productive. Thing is, once the backlog of shite code gets re-engineered, the computing demand for a new code creation will not support bu…
> AGI is not near There's also plenty of argument to be made that it's already here. AI can hold forth on pretty much any topic, and it's occasionally even correct. Of course to many (not saying you), the only acceptable bar is perfect factual accuracy, a deep understanding of meanings, and probably even a soul. Which keeps breathing life into the old joke "AI is whatever computers still can't do".
You consider occasionally being correct AGI?
Knowing history of past bubbles is only mildly informative. The dotcom bubble was different than the railroads bubble etc. The only thing to keep in mind is that all of this is about business and ROI. Given the colossal investments, even if the companies finances are healthy and not fraudulent, the economic returns have to be unprecedented or there will be a crash. They are all chasing a golden goose.
I’m concerned that the accounting differences mentioned between Lucent and Nvidia, Microsoft, OpenAI, Google just mean we have gotten much better at lying and misrepresenting things as true. Then the bubble pops and you get the real numbers and we are all like “yep it was the same thing all over again”.
Fiber networks were using less than 0.002% of available capacity, with potential for 60,000x speed increases. It was just too early. I doubt we will see unused GPU capacity. As soon as we can prompt "Think about the codebase over night. Try different ways to refactor it. Tomorrow, show me your best solution." we will want as much GPU time at the current rate as possible. If a minute of GPU usage is currently $0.10, a…
This is such a short sighted take glaringly ommitting a crucial ingredient in learning or improvement - both for humans or machines alike: feedback loops. And you can't really hack / outsmart feedback loops. Just because something is conceptually possible, interaction with the real rest of the world separates a possible from an optimal solution. The low hanging fruits/ obvious incremental improvements might be quickl…
LLMs are a great tool. But, the real world is far too nuanced to be captured in text and tokens. So, LLMs will be a great productivity boosting tool like a calculator or a spreadsheet. Expecting it to do more is science fiction.
Earlier quoted context omitted.
That Nvidia has to front the costs of the product at the beginning and arguably the risk that the allocation of assets end up not being paid off (bankruptcy, etc.) By carrying those costs early and the associated risk, Nvidia expects a return on that. If the risk is realized they'll lose but otherwise they'll gain. That has all the hallmarks of an investment.
Thanks for your reply, but: NVidia could protect itself against OpenAI bankrupcy by adding a clause to the lease saying that if OpenAI goes bankrupt, Nvidia gets its GPUs back. So the risk would only be that the lease would be aborted sooner than expected.
That is, in fact, the risk.
Earlier quoted context omitted.
> As soon as we can prompt… This is the fundamental error I see people making. LLMs can’t operate independently today, not on substantive problems. A lot of people are assuming that they will some day be able to, but the fact is that, today, they cannot. The AI bubble has been driven by people seeing the beginning of an S-curve and combining it with their science-fiction fantasies about what AI is capable of. Maybe t…
Getting gold in the math Olympiad is a pretty strong indicator of operating independently on substantive problems. A year ago they need an extensive harness to get silver, and two years ago they could hardly multiply 1000x10000. Terence Tao tweeted yesterday about using GPT5 to help quickly solve a problem he was working on.
Why did GPT5 help Terence Tao solve a math problem, because he gave it a prompt and the context etc.
None of these models are useful without a human prompting them and giving it tasks, goals, context etc, they don't operate independently, they don't get ideas of work to be done, they don't operate over long time horizons, they can't accept long term goals and sub-divide those goals into sub goals, and sub tasks etc.
They are useless without humans telling them what to do.
The gpu bubble is different. Nvidia is actually selling gpus in spades. So it’s not comparable to the telecom bubble. Now the question remains how many more gpus can they sell? That depends on the kind of services that are built and how their adoption takes off. So now is it a bubble or just frothy at the top? There is definitely going to be a pull back and some adjustment, but I cannot say how bad it is