Time to go and acquire necessary food stuff is not a luxury in any reasonable framing. What is the alternative, eating drive-thru every day or having Instacart deliver overpriced groceries?
> A model that aces benchmarks but doesn't understand human intent is just less capable. Virtually every task we give an LLM is steeped in human values, culture, and assumptions. Miss those, and you're not maximally useful. And if it's not maximally useful, it's by definition not AGI.
This ignores the risk of an unaligned model. Such a model is perhaps less useful to humans, but could still be extremely capable. Imagine an alien super-intelligence that doesn’t care about human preferences.
The disconnect is that companies are trying desperately to frame LLMs as actual entities and not just an inert tech tool. AGI as a concept is the biggest example of this, and the constant push to "achieve AGI" is what's driving a lot of stock prices and investment.
A strictly machinelike tool doesn't begin answers by saying "Great question!"
When you have a multi-platform image the actual per-platforms are usually not tagged. No point.
But that doesn't mean that they are untagged.
So on GitHub Actions when you upload a multi-platform image the per-platform show up in the untagged list. And you can delete them, breaking the multi-platform image, as now it points to blobs that don't exist anymore.
Circa 1970s there were bands touring (e.g. Yes) with music synthesizers that were made by wiring up voltage controlled oscillators, ring oscillators, filters and similar analog components.
That sort of musician was really an electronics technician who had to keep a fiddly piece of hardware working on the road. And that kind of synthesizer was often monophonic or not very polyphonic.
Yamaha's DX-7 was a digital simulation of that kind of synthesizer which was absolutely reliable and a musical instrument really has to be because you're going to put it in a case that is going to get banged around in shipping and expect it to work just fine wherever you go.
My issue is you end up dealing with dopes who don't want to learn, just want to milk the money and the job security, and actively fight you when you try to make things better. Institutionalized.
Can someone please confirm, is the Graviton an ARM-based CPU or something different? The page mentioned ARM, but I was still a little confused. Are we able to launch a Debian/Fedora using the CPU, or is meant for something different?
It is an amusing and potentially even good concept, but with one caveat - you only consider close enough exchanges with similar peers. To translate that into English - a fully libertarian self-sufficient settlement where people are exchanging home-made stuff between one another. Because everyone was the same and no one could abuse the system, it may even worked. We even seen it in the early criminal communities, when BTCs briefly were a medium of exchange for drugs. And then tokenbros made a leap of faith and magically scaled that to the whole world. Which obviously didn't work in practice. Their famous Lightening L2 is an abomination of hacks and centralization and it still doesn't scale.
The problem of BTC was at the same time overabundance of imagination in one area and a big gap in imagination how the world actually works. They only saw simple isolated scenarios and never a globalized economy.
Yeah, I think so too now that I read some documentation about it. It appears that the main issue with the spinlock pattern is that it inhibits "a severe performance penalty when exiting the [spinlock] loop because it [CPU] detects a possible memory order violation." [0].
~10 years ago, on Haswell, it took ~9 cycles to retire, and from Skylake onward, with some exceptions, it takes a magnitude more - ~140 cycles.
These numbers alone suggests that it really messes up hard with the CPU pipeline, perhaps BP (?) or speculative execution (?) or both (?) such that it will basically force the CPU to flush the whole pipeline. This is at least how I read this.
Well, Sam Altman had said not long ago we would have AGI in 2025, and has been constantly implying something about "AI Scientists" and this and that. He literally said "We now know how to build AGI", also not long ago. He also stated that ChatGPT passed the Turing test without much fuss. The Anthropic has been pushing the narrative about the massive job loss, implying again that there would be an absolutely transforming impact coming soon. The Microsoft MBA-in-charge will have you believe his entire life and work is supposedly managed by an army of Clippy 2.0. The Google-MBA-in-charge has now started day-dreaming about space-based clusters, because guess what, his tool generates better fake pictures than Altman's. He too peddles the nonsense about the superpowerful AI. So, again yes, they said the AI would cure cancer and meet the AGI standard, so I demand they be held accountable to their own words and provide the answer to those questions!
Percentages are fun because they can make something with a small absolute change look like a giant change.
No business is really going to care about $1.00/user, especially when it costs hundreds of dollars per user (or thousands) to migrate entirely away from the Microsoft ecosystem.
I wish I could agree with this, but the ecosystem lock-in is too great. They might lose business for sure but it may not put a dent in their revenue at all.
If you replace office, you'll have to replace sharepoint, onedrive, etc.. and it isn't just the tools but the policies and critical features that go along with those. For most orgs, this is literally their lifeblood, not just some tool they can yank out. For smaller orgs it might be easier, but those don't pay Microsoft as much anyways.
From a user point of view, there are tools that have similar features, some even better features. G-suite is the only platform i know of that unifies all the office productivity products like 365 does. But neither G-suite nor any other platform can be managed/policed as well as 365. At the end of the day, will Google behave any better than Microsoft anyways (cost or otherwise)? And it isn't just policing and management but securing all that precious data in there, Microsoft might not be great but lots of tech-debt has gone into securing it within that platform. A migration would be costly, justifying it with cost savings alone might be difficult.
Not to mention embedded software is often half the pay of a startup and defense software often isn't work from home. Forget asking what languages they can hire for. They are relying on the work being interesting to compensate for dramatically less pay and substantially less pleasant working conditions. Factor in some portion of the workforce has ethical concerns working in the sector and you can see they will get three sorts of employees. Those who couldn't get a job elsewhere, those who want something cool on their resume, and those who love the domain. And they will lose the middle category right around the time they become productive members of the team because it was always just a stepping stone.
Having a relational database in a gui was great. The problem with access was that it tried to support network applications as a backend but supported a punishingly low number of connections. Having an application that crashed with 10-50 connections put it in an awkward space. Businesses without a strong technical team would build on it, release with N=1 load testing, and get surprised when it crashed out at scale. MS wasn't going to improve it, because they wanted more sophisticated customers to buy SQL server.
I had a similar realization recently after realizing I was putting way too much energy into my day job.
The one thing I’m having a hard time with is determining what the right level of effort is. What signals and indicators do you use to know when you’re doing enough or too much?
The problem is, Liquid Glass isn’t all that closely aligned with the visionOS UI, despite both having glass-like qualities. Most notably, the iOS and macOS versions are missing the usability affordances that the visionOS version has, and more superficially the visionOS version looks nicer.
I couldn't tell you the exact details (I'm only passingly familiar with how it works myself), but you'd almost certainly want to start by looking into web assembly.
I found a-Shell's documentation[1] quite interesting, it describes their use of web assembly and offers some practical tips for compiling stuff so it can work in a sandboxed environment.
I built a static site to make my life intentionally harder. No login, no database, no apps, just weekly challenges designed to break dopamine loops, disrupt routines, and help you appreciate what you already have.
Every week, a new challenge drops: no moving frames, candlelight evenings, corded phones, sleeping on the floor…
