Posts by pip25 381 publicly visible posts • joined 17 Jun 2010
Training model still requires an impressive infrastructure, where even maintenance costs can be prohibitive. Chinese get away with it for the same reason they get away with relatively cheap electric cars for their internal market: government support/funding. That's not trivially replicated outside of China at the moment.
This claim that "new hardware will make AI so much cheaper to use" reminds me of how, despite hardware becoming several orders of magnitude faster, user interfaces remain floating around "mostly usable" in terms of speed and responsiveness. My first Motorola Android phone was slow to switch apps, and so is my current Xiaomi. Windows 3.1 had some hiccups on occasion, and so does your average webpage using the latest and greatest framework today.
Why am I bringing this up? Because just how UI ate up all those performance advantages on the hardware level for the sake of faster development (among other reasons), I am fairly convinced that AI models won't get an iota cheaper for the big players, in the short term most certainly. The models are exploding in terms of complexity, to the point that someone releasing an 1T+ parameter model today no longer raises any eyebrows. They will get even bigger still, and the new hardware will have its work cut out for it.
I guess that's the only thing missing from this "you might get sacked within one hour" love letter.
I'm not sure which is worse: a statement that makes it obvious how the bosses are fully aware that they're being asses and desperately try to put a positive spin on the decision (as we see here), or one that's so divorced from reality that they genuinely believe their nonsense.
Where I live we have the saying "among the blind, the one-eyed is king". There is such a rush towards AI solutions in a myriad of use cases, quite often pushed by people who have absolutely no idea what they're pushing for, that someone who does little more than manage expectations and serve as a rational voice the bosses actually listen to can go a long way.
They just didn't tell you. Private repos (and everything else not easily accessible from the open web) may have been the main exceptions. Now those are fair game as well, I suppose.
The poster above is right. They must be really desperate if they believe this will help improve their products in any way.
This new Steam Machine is being marketed as being capable of 4K output and run even the newest AAA games... but with the given hardware, the two are quite unlikely to be true at the same time. To say nothing about how future proof this is looking. Somewhat unlike Valve to prioritize short-term sales against long-term customer satisfaction.
I might not necessarily agree with the recommendation of trying to solve things yourself first. I can do that for most tasks I come across; the bigger problem is whether my solution is good (fast, maintainable, extendable, etc.) enough to be used. A bit of online research beforehand can increase the chances of that greatly.
Unfortunately, the same cannot be said for AI tools, especially those that suggest one single solution, providing little context for its advantages and disadvantages.
Look, if you do perceive that there are issues with copyright as it is currently, then we are not that far apart in opinion anyway. I understand that you believe those cons do not outweigh the positives, and you are entitled to that opinion. I don't want to convince you otherwise, because that wasn't my original point: what I was trying to say was that creators continue to create creative works even in today's climate, where the system that's supposed to empower them is not at all working in an ideal way (again, opinions on the extent of that vary). It would thus be an overreaction to claim that the abolishment of copyright (which, as I mentioned, is incredibly unlikely) would somehow lead to the global end of creative output. Creativity would limp along in some way regardless, just like it does today.
[...] and is reliable (Wikipedia [...]
I like Wikipedia, but it's been a while since I've seen that word and "reliable" in the same sentence. And I think that's not actually surprising, because if you take a sufficiently complex (moral) problem, you will find that that there are in fact no reliable sources to use, in the strictest sense. Every author brings their own prejudices and opinions into the mix. And even if our AI would be intelligent in a more substantial way than LLMs today, I do not see why we could hope it'd do more with those viewpoints than we could.
That's a bunch of overdramatic hogwash. A case can be made that copyright, in its present form, does more harm than good, and actually fails at what it was originally created to do, which is to support creators. Yet creators still create. Should it disappear overnight, other incentives would take its place, perhaps better ones.
But let's not kid ourselves, copyright won't just disappear, even if our AI overlords demand it; it's way too entrenched in current society. But if the whole AI brouhaha results in some well-needed reforms of copyright - in my book in might just have been worth it.
The limits of silicon were well-known and talked about decades prior. Whether Moore's Law is truly dead depends on whether someone with sufficient amounts of money manages to reinvent the basic building blocks from scratch. (Nvidia would be in prime position considering it's swimming in cash, but that's assuming those profits won't be spent on luxury yachts or whatever.) Also decades ago, there were talks of basing chips on light instead of electric current. I wonder what became of that research.
Thanks for the explanation, but this does not seem to mesh with how I saw passkeys being used. My only interaction with them so far has been through the Proton Pass extension, which stores my passkeys encrypted in the cloud somewhere. I can retrieve them on any device I install that extension/app on.
My thoughts exactly - thankfully the text is easily extracted. I've made an effort to validate it by skimming parts of the article, and it seems valid enough. Done by Gemini Flash 1.5.
----
This article argues that the tech industry has entered a state of "Rot Economy," where the relentless pursuit of growth at all costs has led to the degradation of user experience across virtually all digital platforms. The author contends that this isn't a deliberate, Machiavellian plan, but the consequence of countless short-sighted decisions prioritizing short-term financial gains over user needs.
Specific examples are given, including Spotify's redesign prioritizing video content over its core music function, Sonos's app update removing accessibility features, and Meta's frequent, disruptive redesigns of Facebook and Instagram. The author highlights the pervasive nature of this issue, citing manipulative design choices, excessive advertising, microtransactions, and constant notifications as contributing factors to a degraded user experience that negatively impacts users' psychological and social well-being.
The author also draws a distinction between their theory of "Rot Economy" and Cory Doctorow's "Enshittification," arguing that the Rot Economy reflects a broader, more fundamental issue of growth-at-all-costs capitalism impacting the entire digital ecosystem, not just specific platform strategies. This relentless pursuit of growth, driven by figures like Jack Welch and Milton Friedman's neoliberal philosophy, leads to the prioritization of metrics over user satisfaction and ethical considerations.
The author uses a detailed anecdote involving purchasing a budget laptop to illustrate how this affects even low-cost devices, which come pre-loaded with a slow, ad-laden operating system and forced into specific ecosystem usage. This highlights how widespread and impactful the problem is across different socioeconomic groups. This negative user experience is compounded by the rise of manipulative algorithms, AI-generated content, and widespread scams and disinformation.
The author concludes by calling for greater awareness and accountability, urging readers to recognize the widespread nature and psychological harms inflicted by this system. They name and criticize several prominent tech CEOs (Sam Altman, Tim Cook, Sundar Pichai, Satya Nadella, and Mark Zuckerberg) for their roles in contributing to and profiting from this system. Finally, they call for a collective effort to resist and challenge the status quo, emphasizing the importance of public acknowledgment and discourse as a catalyst for change.
I can rent a H100 SXM on-demand for $2.99 per hour on RunPod, or $1.75 if I allow my workload to be interrupted. (This is probably not the cheapest provider either, just one that comes with easy-to-setup Docker templates.) That $2.5 to $6.74 offer from DigitalOcean is rather uninspiring in comparison.
Well... duh.
As much as I agree with AI PCs being a load of nonsense currently, anyone who bought an ARM-powered Windows machine for gaming simply did not do proper research. It's not like this is the first, or even the second time people got burned by Microsoft's ARM offerings.
It can be useful exactly because the LLM output is error-prone. It's just text. A "tool" on the other hand is an API. Its input commands can be validated, not just syntactically, but also whether they make sense in the current context.
Let's say you have a cocktail-making machine. Its hardware can follow a certain list of basic commands, but not all of them are always valid: there's no point in shaking the liquid container if it's empty, for example. You give the LLM the recipe of the cocktail in plain text format, like something you copied from a website, and it needs to translate it into a list of "tool" commands the machine can use. If it does anything else, starts lecturing you about the dangers of alcohol, or just outputs an invalid command, the system will immediately know that the LLM bugged out and may attempt to recover by restarting the process with a different random seed, for example.
With a heavy heart, because I think it's a great piece of software, and we'd probably fit the criteria for a free license - for now. But the yearly license renewal sounds like a disaster waiting to happen; if they change their terms for any reason, we're screwed. That's not something we can risk.
Aero, then touch operation and the interface formerly known as Metro, Cortana, augmented reality, and now so-called AI. I assume much of the latter will be gone too by the time Windows 12 rolls around.
Could we have a no frills version, perhaps? I'd even pay money for it.
