blihp

This was just the author's current issue on Android. In another month or two it would have been something else. I fought similar battles for the better part of a decade before finally giving up when Google's policies made it so that even keeping apps running (at least in my case) was an economic non-starter. The sheer amount of bureaucratic B.S.[1] they constantly fling at you while simultaneously bit-rotting existing applications is insane.

[1] Sometimes it's related to their store listing policies which are constantly changing, sometimes it's related to taxation in a specific country, sometimes it's related to laws in a specific country, sometimes it's actually related to software (on-device or web services!) they are forcing you to update/change etc. etc.

This was the company that made all sorts of noise about how they couldn't release GPT-2 to the public because it was too dangerous[1]. While there are many very useful applications being developed, OpenAI's main deliverable appears to be hype that I suspect when it's all said and done they will fail to deliver on. I think the main thing they are doing quite successfully is cashing in on the hype before people figure it out.

[1] https://slate.com/technology/2019/02/openai-gpt2-text-genera...

I use FreeCAD on a very regular basis and can understand why it's not more popular: it's very powerful but has some very sharp edges that will often have me using it in a state of near rage. Topological naming comes to mind but there are other various issues that I've hit like a brick wall (in that you can't work around the bugs/limitations so much as you must rework your design to avoid them which can be tedious and frustrating) when designing something non-trivial.

That said, each release continues to improve it just has further to go than most open source projects.

Yes, but as I recall it was a rather tiny subset of the database market unless one was very forward thinking and had Internet access and/or was part of a BBS community that shared information electronically which put you in the minority of the minority at the time. Financial market data was probably the largest one early on. Mostly individuals and small businesses were just rolling their own solutions and entering the data themselves until the late 80's / early 90's. Larger businesses had many of islands of systems and data but it was nearly all internal or used to generate paper documents... even internally.

Roughly speaking: prior to the 80's one (often) wrote their own programs, prior to the 90's one entered their own data (i.e. for structured content like databases and spreadsheets), prior to the 00's one created their own content (i.e. documents etc.)

There were lots of PC databases, of which dBase was one of the more popular ones.

Beyond the technical reasons for the limit, it provides for a relatively painless way to begin to build out/for RISC-V[1] without an uncomfortable transition. For those who just want a better next iteration of the controller, they have it. For those who build tools, want to A/B test the architectures, or just do whatever with RISC-V, they have that too. All without necessarily setting the expectation that both will continue to coexist long term.

[1] While it's possible they are envisioning dual architecture indefinitely, it's hard to imagine why this would be desirable long term esp. when one architecture can be royalty free and the other not, power efficiency, paying for dark silicon etc.

The tweet thread makes it clear: AMD has consistently based future strategy on the past which is why they were so ill-prepared for every major non-PC trend this century in CPUs until Ryzen. (which was basically catching up + much better value than Intel) This also translates to their GPUs where they seem to have absolutely no consumer GPU vision beyond 'we want some of what nVidia's getting.' Their current strategy seems to be weaker hardware + weak compute drivers + a little cheaper than nVidia = success.

In the consumer space at least, WiFi was nowhere to be seen on a typical PC when Apple adopted it. Same with USB. So while it technically originated and existed elsewhere, there was no serious traction on it prior to Apple adoption.

What you say is also true: many people weren't ready to ditch the old when Apple decide to deprecate it.

I suspect it was all about the money. Those following him viewed him as their best bet for a good financial outcome for themselves.

The author of the library has done numerous YouTube rants on how bad he thinks the AMD compute drivers are.

Yes, some of those early cards were monsters usually due to some combination of discrete 7400-series logic (i.e. economies of scale weren't there for most ASICs yet) and memory chips (individual DIP chips as SIMM/DIMM modules weren't a thing yet) in the 8086 and into the 80268 era. I remember issues with some of the early clones not having full length/height slots causing an issue with some of those cards. By the 386 era, those monster cards had mostly died out on the desktop side at least (there were a few specialty cards that persisted into the late 80's/early 90's probably before enough customers complained about not fitting in their new system.)

The point being made is that AT&T should have never been allowed to reform itself as it did, probably along with numerous other acquisitions/mergers that the government approved that they should not have. There was a period of time late in the last century when many of us had numerous (i.e. 5-6) options for a while.

Interesting article... I've been wondering why things seemed to go in a strange direction at Pine64. That article provided some clarity.

They don't have a choice. FPGA's aren't like vanilla logic chips: they are highly proprietary both in terms of their specific hardware functionality as well as software tool chains and you can't second source them. Some of the larger customers are using them for government contracts (think military and space applications) where they are signing up products they are contractually on the hook for from years to decades.

Yes, there were 2-3 other (mostly web-based) services that stepped in with Feedly being the main one. However, by that time the popular client-side tools had been pretty much been wiped out due to years of Google's free offering. For some of us, that's what killed the user-facing RSS market. (I was so fed up I ended up cobbling together my own solution... far from anything polished but I'm happier with it than I was with any of the offerings available at the time)

Because history has shown that the money is in selling the picks and shovels, not operating the mine. (At least for now. There very well may come a point later on when operating the mine makes more sense, but not until it's clear where the most profitable spot will be)

No, the schemes being used are decidedly low tech for the most part and trivially detectable. Google's result quality isn't dying a death of a thousand cuts due to hordes of disreputable actors, it's dying a death of Google being dependent on a small basket of sleazy companies that they're economically in bed with.

Yep, but what do you expect from an industry near its end. They tend to do better with the front page stories, but a lot of the local/filler stuff now looks like this. You're not going to have meticulously crafted articles with editors to ensure its up to the papers standards when you don't have the ad (or subscription) dollars coming in to support that level of output. So now it often looks more like a blog than a newspaper... plenty of more prestigious news organizations are looking just as bad these days.

Sure, but the question is would OpenAI be around to see the investment bear fruit after such a spend? In the 90's infrastructure companies (you know, the 'rock solid' infra companies who were going to be around forever) spent billions to lay a lot of dark fiber that was to be the future of the Internet, went bankrupt during the dotcom bust and then Google and other Internet companies gobbled up said fiber at fire sale prices. Of course it was the future of the Internet... just not for them.

On second thought... go for it, Sam! I'm looking for some powerful AI hardware at great prices...

That's because he doesn't design for where the state of the art chips are, he designs for where they are going to be.

My take was that they currently have a working chiplet design and are looking to move to a wafer-scale (i.e. not cut up) design. Thermal/power/yield are all issues and the design has to take all of that into account. Cerebras has done it for their NN processors, so it has been done before.

When you really analyze at how many/most of the richest people in the world got so rich, they tended to extract (i.e. more parasitic) much of the value rather than create (i.e. more symbiotic) it. They have also often done so at the direct expense of others simply so they could have more... when they already had much more than they could ever use. Our economic system allows for this, but to praise these folks for much of what they did leaves a bad taste.

Right, that's the problem: getting mainline support while the processor is still relevant rather than years after the fact... if at all.[1] Also, the bar has moved. OpenGL was good enough 10 years ago. What about OpenCL? Vulkan? NPU support? HDMI in? etc. (I'm very interested in the functionality the 3588 offers but not if that functionality is just on paper with the hope that maybe someday someone gets it working or if I need to run a cobbled together kernel build that won't see meaningful updates)

[1] There have been plenty of devices where someone was working on drivers and then the developer moved on or a new revision was released etc and the original thing never gets the support people bought it anticipating.

I was actually looking into the Orange Pi this weekend... the driver situation remains a deal killer for me. Sadly the Raspberry Pi offers crap hardware with almost good drivers while most of the other SBC's offer good hardware with crap drivers and/or binaries only for specific kernel releases. It's impressive how actively bad ARM is for anything remotely user facing on Linux... nVidia wasn't even this bad for this long.

Assuming it's the same 'tile' tech they're using in their GPUs, it's a chiplet with a Xilinx FPGA with hard IP blocks where appropriate.[1] So it's probably a matter of reverse engineering the bitstream as well as the functionality of those hard IP blocks.

[1] My bet would be that it's mostly (90%+) hard IP blocks since most of the functionality for deep neural nets is well known and the FPGA logic just glues everything together and gives them some wiggle room as architectures evolve.

Don't be so sure that all, or even most, junior employees know any such thing. I've seen junior employees fired for doing silly things in prod before[1]

[1] Of course whatever more senior bozo granted the junior the rights to blow up the thing(s) they did should have been fired instead. That's not the way things work in the corporate world.

Anything that pulls in income and/or pushes out expenses results in a higher tax liability to the IRS. It makes a lot of sense in the world of real estate and equipment expenses (i.e. items which have very long useful lives to a business), less so for labor and other short term expenses.

While it's slowed down in recent years, it's still active. Re: Pharo, that's just because they market it hard and don't give much (any?) credit back to Squeak which is what it was forked from. (and still shares more DNA with Squeak than the Pharo folks like to admit)

I think so. SqueakJS fills the space Amber used to occupy.

The problem with a text search is that you have to get your keywords exactly right. With LLMs you ask inexact questions like 'has the topic of X ever been discussed?'[1] without needing to have an exact match on X. An LLM front end which could return references to the full text seems like the best use of both.

[1] For example, your query might have had X be 'crime' and the transcript would have references to multiple specific types of crime such as 'muggings', 'vandalism' etc. which a full text search isn't going to match. Further, with the LLM front-end you could refine the query to ask about violent crime etc.