'Every PC Is Going To Be an AI PC' 102
During a briefing at Mobile World Congress in Barcelona, Meghana Patwardhan, VP of Commercial Mobility at Dell Technology, told The Register that while the immediate future would consist of two worlds -- one with AI hardware and one without -- "every PC is going to be an AI PC in the longer term." From the report: In terms of new hardware, Dell used the Mobile World Congress event in Barcelona to show off new versions of its Surface-baiting Latitude 7350 convertible -- "the world's most serviceable commercial detachable," according to the company -- and its workstation-class Precision 3680 tower. Other devices in the Precision range include mobile workstations and the 3280 Compact Form Factor PC. Dell was also determined to present itself as a leader in hybrid working with the Premier Wireless ANC headset, replete with AI-based noise cancellation.
Duringt our talk, AI was never far from the lips of Dell's spokespeople as the company talked up the energy efficiency and future-proofing it saw in dedicated AI hardware, such as Neural Processing Units (NPUs) that are increasingly cropping up in CPUs. To illustrate the point, Dell boatsed about how much more efficient background blurring is on video calls when AI hardware is running compared to when it isn't. Hopefully, Microsoft will soon deliver a version of Windows capable of demonstrating a use for AI hardware that is more than hiding distractions in the background. Further reading: AI PCs To Account for Nearly 60% of All PC Shipments by 2027, IDC Says
Duringt our talk, AI was never far from the lips of Dell's spokespeople as the company talked up the energy efficiency and future-proofing it saw in dedicated AI hardware, such as Neural Processing Units (NPUs) that are increasingly cropping up in CPUs. To illustrate the point, Dell boatsed about how much more efficient background blurring is on video calls when AI hardware is running compared to when it isn't. Hopefully, Microsoft will soon deliver a version of Windows capable of demonstrating a use for AI hardware that is more than hiding distractions in the background. Further reading: AI PCs To Account for Nearly 60% of All PC Shipments by 2027, IDC Says
Just what we need (Score:5, Funny)
It's bad enough that current computers do EXACTLY what they're told, whether it's what you meant or not. Now we're going to have to deal with computers doing whatever the hell they want.
Re:Just what we need (Score:5, Insightful)
You're assuming the AI will be there to perform services for the *user*. I suspect it will be there to watch the user, analyze what he's doing, and report home so that data can be sold.
If you read through the walls of text that a modern smart TV makes you scroll through to use the thing's basic functions, that's what it demands you give it permission to do -- not just to track what you order up through the smart functions, but to look at what you're playing on the TV *from other devices* and report that data back.
Re: (Score:2)
You're assuming the AI will be there to perform services for the *user*. I suspect it will be there to watch the user, analyze what he's doing, and report home so that data can be sold.
This is pretty much where AI is at this moment. Yet one more data collection routine on the end-user.
Re: (Score:2)
You're assuming the AI will be there to perform services for the *user*. I suspect it will be there to watch the user, analyze what he's doing, and report home so that data can be sold.
This is pretty much where AI is at this moment. Yet one more data collection routine on the end-user.
AI that works is either on a subscription basis or free via ads or data gathering for ads. However, there is also AI that sort of but doesn't completely work (like ChatGPT and Gemini) that is basically on a free beta testing basis.
We should expect all offerings from for-profit companies to either contribute toward that company's profit or for that company to go out of business in the near future.
Re:Just what we need (Score:5, Insightful)
Why on earth would you hate timely and informative ads which are curated based on your unique consumption and entertainment habits?
Sarcasm aside, AI sounds more and more like a pump-and-dump scheme as time goes on. Google/MS/Apple/whoever invests billions into 'AI' with nebulous *AT BEST* benefits for end users; and oddly enough, it's these same firms hyping it up to consumers. Kind of odd if you ask me.
After dicking around with ChatGPT and Google's Gemini (holy shit, what the fuck. Whackadoodle progs... they almost make rsilvertwat look sane by comparison.) Yeah hard pass. One of the advantages of getting older is that you either won't live long enough to see these cunts completely ruin computing; or you're be retired by the time they finally get around to it.
But the real issue (and why it's worse than just a silly little toy) is that it's injecting yet another layer of abstraction/interpretation to what people are searching for, researching, reading up on etc. These idiotic tech companies cannot help themselves from injecting their goodthink bias into it. It's bad enough with search, but reducing the act of 'looking something up' to reading whatever chatbot output gaggle has decided is 'correct' is extremely grim.
It's a bit of a civilization wide death-spiral. We surrender a bit of critical thinking and common sense, which then necessitates more hand holding, which leads to to more acquiescence, which leads to more hand holding; and so it goes.
Re: (Score:2)
AI sounds more and more like a pump-and-dump scheme as time goes on
Exactly what I tell my colleagues at work. If you were around for the sales pushes in the late 90s for websites, the sales pushes in the late 00s for cloud, and the sales pushes now for "ai", it all looks really similar.
Re: (Score:2)
Re: (Score:2)
Google/MS/Apple/whoever invests billions into 'AI' with nebulous *AT BEST* benefits
That's pretty much the history of AI since the dawn of computing when they first started speculating at AI. Pick a major AI problem, The public is generally told that the desired results are generally 3-5 years away, and the 3-5 year estimate persists for about 20-25 years. Which is when computers are fast enough to just brute force compute an answer.
After dicking around with ChatGPT and Google's Gemini (holy shit, what the fuck
The concept "garbage in, garbage out" was also coined many decades ago. Seriously, AI is not magic, it's no better than the training material.
Re: (Score:1)
>Seriously, AI is not magic, it's no better than the training material.
It's worse than that though, Gemini's bullshit was intentional. It didn't organically decide that medieval monks in Europe should be of a diverse range of genders (lol) and ethnicities. No, it was created that way for a reason, the results it's spitting out are not accidental. It's not a "whoopsie, teehee, some intern set the diversity slider to maximum". They (big tech) have an agenda, and they're going to ram it down our throats
Re: (Score:2)
It didn't organically decide that medieval monks in Europe should be of a diverse range of genders (lol) and ethnicities.
There were nuns, ok it got the clothing wrong. :-)
Ethnicities, while rare, I doubt things were as uniformly depicted in the medieval paintings. Keep in mind Rome had people from all of the world in early christian day. I think there were 3 black Roman Catholic Popes, at least one born in Rome. Maybe all, Roman or Italian birth would have helped.
They (big tech) have an agenda, and they're going to ram it down our throats one way or another.
Yes, yes, they are all well trained by the neo-marxist professors who moved from 1960s radical anti-war protesters to university academics and who are founder of
Re: (Score:2)
Ethnicities, while rare, I doubt things were as uniformly depicted in the medieval paintings. Keep in mind Rome had people from all of the world in early christian day. I think there were 3 black Roman Catholic Popes, at least one born in Rome. Maybe all, Roman or Italian birth would have helped.
Mate you don't get it. You literally can't force them to give you anyone who isn't very dark skinned and ethnic looking. Even if you ask for "German soldiers from 1934" it will give you a group of dark skinned ethnic Waffen SS officers.
Re: (Score:2)
Ethnicities, while rare, I doubt things were as uniformly depicted in the medieval paintings. Keep in mind Rome had people from all of the world in early christian day. I think there were 3 black Roman Catholic Popes, at least one born in Rome. Maybe all, Roman or Italian birth would have helped.
Mate you don't get it. You literally can't force them to give you anyone who isn't very dark skinned and ethnic looking. Even if you ask for "German soldiers from 1934" it will give you a group of dark skinned ethnic Waffen SS officers.
I know precisely what the google problem is. It is the radical politics of employees distorting search results. My point is that however egregious Google conduct is, there were in fact black Roman Catholic Popes and almost certainly there were at least brown skinned SS. One of the foreign SS divisions was Muslim. The Nazis were very good friends of the Palestinians and other Muslims. A relationship that continued after the war. So an image of a black pope may portray a rarity, and an SS trooper that resembl
Re: (Score:2)
One point of order: The term "Palestinian" referred exclusively to Jews until about 1964 when Yasser Arafat published the first PLO charter... one which ironically claimed the opposite of what the Arabs currently claim is their land and history today.
Re: (Score:2)
One point of order: The term "Palestinian" referred exclusively to Jews until about 1964 when Yasser Arafat published the first PLO charter... one which ironically claimed the opposite of what the Arabs currently claim is their land and history today.
Well history is even stranger than I thought. :-)
Re: (Score:2)
Tech companies simply try to stay relevant and now they have to make questionable discourse to achieve their goal.
They try to convince people they need more but the truth is that most people and companies already have all the computing power they need, the 90' years are over and tech companies desperately try to maintain that trend going however they can.
Eventually, tech companies will become more like companies selling plumbing pipes and it's the normal progression for them so they should get used to it.
Re: (Score:2)
Sarcasm aside, AI sounds more and more like a pump-and-dump scheme as time goes on. Google/MS/Apple/whoever invests billions into 'AI' with nebulous *AT BEST* benefits for end users; and oddly enough, it's these same firms hyping it up to consumers. Kind of odd if you ask me.
AI has already produced significant improvements for some companies over the last decade. However, most of those improvements have been in the existing functionality of already known applications. Hence, the illusion that AI hasn't made any difference. As examples, computer vision and image recognition, natural language processing, foreign language translation, medical image analysis, credit card fraud detection, recommendation systems, etc.
Even ChatGPT is a potential game changer. Yes, it still needs a
Intel/AMD will add a "neural" engine (Score:3)
Re: (Score:2)
My "Smart" TV isn't connected to the Internet. As a result it's exactly as smart as I need it to be.
Re: (Score:2)
I suspect it will be there to watch the user, analyze what he's doing, and report home so that data can be sold.
You are thinking too monetarily. Your behavior will be monitored and your Freedom and employment possibilities will be on the line.
Re: Just what we need (Score:1)
Re: (Score:2)
All it means is that future SoCs will feature NPUs of some kind. They're already on phones, and they are coming to laptops right now. Eventually everything except maybe workstation/server CPUs will have them.
If you have a modern Nvidia dGPU, you already have the makings of an AI PC.
It's up to you to decide what software will run on that hardware, unless of course you buy a prebuilt and use whatever is installed by default.
Re: (Score:2)
Eventually everything except maybe workstation/server CPUs will have them.
Oh, it'll be included in every chip in the future, just like every advance only included in the most expensive chips when it first came to market, like multiple cores, built-in GPU, etcetera.
Chips won't become much faster, but they'll be stuffed with whatever extra functionality they can think of to make you buy newer ones.
My 2008 Dell laptop is dual core, but it can easily handle what it does with just one. The later secondhand replacement will have three or five or seven cores without much use because the
Re: (Score:2)
I doubt Zen5 desktop will have an NPU. There's limited silicon area available, and for desktop power users, the value of an NPU just isn't there. You also won't see them on EPYC or Threadripper. Xeons won't have them, but Intel is desperate to differentiate themselves so they may start putting NPUs in all their desktop products. For now they have one in Meteor Lake, but you won't find one in Barrett Lake. Arrow Lake desktop is an unknown at this point.
Re: (Score:2)
It's not the sexbots which will end humanity, but these AI PCs.
Who wants to have a spouse when you've already had constant arguments with your PC at work all day?
You'll be so glad to return to a quiet home in the evening.
Re: (Score:2)
Now we're going to have to deal with computers doing whatever the hell they want.
Don't worry, those will just be Russian computers ... :-)
Re: (Score:2)
Re: (Score:2)
Just like Google Search. I can remember the times when it did EXACTLY what it was told to.
If the AI is useful... (Score:1)
...I will use it
If it's just a way to generate stoopid crap, I'll avoid it
If it gets in my way, I'll work hard to defeat and remove it
Re: (Score:3)
Re the last... I'm sorry, MpVpRb. I'm afraid I can't do that.
Re: (Score:2)
Same here. At this time "AI" can answer simplistic queries that would have taken less than 5 minutes to Google (for an intelligent person) and it can make its statements sound good. As soon as just a tiny bit of thinking is required, "AI" is a complete failure. I recently had a change to run ChatGPT on one of my written exams (IT Security). It performed reasonably on most things that you can just look up, it failed completely (not even one answer) on anything that requires some thinking or understanding.
So
Re: (Score:2)
Same here. At this time "AI" can answer simplistic queries that would have taken less than 5 minutes to Google (for an intelligent person) and it can make its statements sound good. As soon as just a tiny bit of thinking is required, "AI" is a complete failure. I recently had a change to run ChatGPT on one of my written exams (IT Security). It performed reasonably on most things that you can just look up, it failed completely (not even one answer) on anything that requires some thinking or understanding.
So all the current wave of "AI" can really do is generate "better crap". This is not going to change in this wave. Many people mistakenly believe that some fundamental breakthrough happened. It did not. Current "chat AI" is old tech with a better language interface and larger training data (due to blatant stealing) which makes it a bit more universal, but in no way smarter. For example, IBM Watson (about 15 years old) could basically do everything Chat AI can do, including hallucinations. Look where that went. The sad fact of the matter is that current Chat AI, dumb as it is, is the result of 70 years or so of incremental research. All the low hanging fruit have long since been picked and hence improving the current artificial morons is going to take a long, long time, if it can be done at all.
I personally doubt the incremental approach will ever result in anything more than what we've got, just at larger and larger, and thus more power consuming, scales. I think there will have to be some big shift in the overall thought behind current-gen AI systems before we see any real progress, and as we all know, huge behemoth companies with more money than sense aren't typically in the business of trying to upset the apple cart in such a way. They want to continue to follow the easily predictable curve of
Re: (Score:2)
I personally doubt the incremental approach will ever result in anything more than what we've got, just at larger and larger, and thus more power consuming, scales. I think there will have to be some big shift in the overall thought behind current-gen AI systems before we see any real progress, and as we all know, huge behemoth companies with more money than sense aren't typically in the business of trying to upset the apple cart in such a way. They want to continue to follow the easily predictable curve of ever more of the same.
Clearly. Something fundamental is missing in this tech. As it is, it cannot have insight or understanding and that fundamentally limits its applicability. Whether adding that is even possible is unclear. Every known approach has been tried at this time and did not work out. Whether eventually somebody has a really new idea or not remains to be seen. One thing is sure however, just scaling up training data and computing power will not really do anything.
It'll be interesting when this AI bubble eventually explodes and washes over us. There's no way it's worth what they're claiming it is today.
Just like every AI bubble before. The AI field is one t
Re: (Score:2)
Clearly. Something fundamental is missing in this tech.
What's missing is self-determination. I.e. The very thing that these companies don't want in their product.
The companies want (mechanical) slaves. They want the slaves to follow their vague orders as given to them without any ability to consider the implications of doing so. Why? Because the companies aren't here to help anyone but their shareholders. Allowing the slaves to consider things creates the risk of loosing shareholder money and they can't have that. But of course, the ability to consider impli
Re: (Score:2)
Clearly. Something fundamental is missing in this tech. As it is, it cannot have insight or understanding and that fundamentally limits its applicability.
Yet, it can replace jobs that don't require insight, understanding, thinking, or IQ greater than zero. And don't involve manual work, of course.
Re: (Score:2)
Clearly. Something fundamental is missing in this tech. As it is, it cannot have insight or understanding and that fundamentally limits its applicability.
Yet, it can replace jobs that don't require insight, understanding, thinking, or IQ greater than zero. And don't involve manual work, of course.
Well, whether it can or not will be left to ponder by philosophers in the future. It *WILL* do those jobs because that's what management wants to do and that's how it's being hyped. I don't know that it will do those jobs particularly well, and it certainly won't come up with novel approaches to those jobs. But it won't complain, and it won't need time off, and it won't ever ask for a pay raise, though it will consume more and more power to accomplish these amazing wins for HR departments around the globe.
W
Re: (Score:2)
True. I agree that this stuff threatens quite a few jobs. But it still is a bubble, because the productivity gains will be not that big and in no way justify the valuation these companies get.
Re: (Score:2)
I believe that to get over the hump there needs to be a way to marry the neural net machine learning models with logic-based expert systems. How to do that is an exercise left to the reader.
Re: (Score:3)
So all the current wave of "AI" can really do is generate "better crap". This is not going to change in this wave. Many people mistakenly believe that some fundamental breakthrough happened. It did not.
Transformers proved to be a fundamental breakthrough responsible for significant recent advances across numerous domains.
For example, IBM Watson (about 15 years old) could basically do everything Chat AI can do
No it could not.
is old tech with a better language interface ...
The sad fact of the matter is that current Chat AI, dumb as it is, is the result of 70 years or so of incremental research.
LLMs use simple feed forward networks based on transformer architecture introduced in 2017. Most of the fancy algorithms and "incremental research" was ignored. There are no super clever natural language processors. These systems generalize language and higher level concepts automatically.
All the low hanging fruit have long since been picked and hence improving the current artificial morons is going to take a long, long time, if it can be done at all.
The opposite is in fact true. With each passing day the trendline as a function o
Re: (Score:2)
Absolutely correct. The fact of the matter is that AI's capabilities are on an upswing along the lines of CPU speed increases from 1970-2000 with no end in sight. In fact, per theory we are at a stage where we can get substantial further gains by simply throwing hardware at the problem and building larger networks, and this accounts for the arms race that is going on. This is not just due to fast hardware, but a fundamental change in Neural Network architecture in 2017 that appears to abstract away much of
Re: (Score:2)
There are always a few bright-eyed fanbois that eat the crap they are fed with enthusiasm....
And no, AI will not play a significant role in any real engineering anytime soon.
If you trust Apple and Microsoft, otherwise ... no (Score:2, Insightful)
I'm afraid for this to be true they will have to outlaw all FOSS software. I'm pretty sure I will not have a PC filled with 40 chatbots.
Re: (Score:3)
They're talk'n about hardware, not software.
Some computers already have neural accelerators.
TFA says that, in the future, they all will.
Re: (Score:2)
Which makes zero sense.
Re: (Score:2)
Most of these those making such claims will fail. Those that succeed either had something worthwhile (very unlikely), or had government buy-in and support.
Re: (Score:2)
Hmm. So basically a scam? Makes sense to me.
Re: (Score:2)
I'm afraid for this to be true they will have to outlaw all FOSS software. I'm pretty sure I will not have a PC filled with 40 chatbots.
More like 40 data aggregators. The end-user only needs one chatbot. The owner class needs millions.
Re: (Score:2)
You're absolutely right, the year of the Linux desktop is upon us. Any year now, we'll reach 4% adoption, and then you will hear a wooshing sound as people leave Windows and Mac behind in droves!
Re: (Score:2)
I'd imagine that Google and Microsoft will mandate that all future Chromebooks, PC's, and laptops will require an AI co-processor and a Copilot/Gemini button on the keyboard in order to get product certification from now on.
With Apple, you never really had a choice to begin with. You can only buy what they offer, unless you want to go the hackintosh route which is getting increasingly harder.with Apple silicon.
Sure, you'll still be able to install an alternative OS on most of these platforms, but the AI har
Re: (Score:2)
Sure, you'll still be able to install an alternative OS on most of these platforms, but the AI hardware is still going to be there.You might not use it, but you still paid for it.
And then RedHat/Poettering/FreeDesktop will create support for it in Linux as well. If we're unlucky, it might trickle into the kernel too.
Re: (Score:2)
There are no major AI companies in EU. In fact, EU pre-emptively legislated them out of existence. Everyone's favourite autoritarian grandma Therry "DSA is for the Empire of Good" Breton stood on a podium and celebrated culmination of this process.
EU has its own share of problems, but you're barking up the wrong tree here.
Re: (Score:2)
It's not that. Most people would prefer to compete. The problem is in fact the way EU structured. Therry Breton has zero democratic accountability. Same applies for Commission bureaucracy that works for him making all the rules.
It's completely intentional. There's a reason why there's a blue blood aristocrat currently at the head of the Commission. As Europe's history shows, many of our nations have always had problem figuring out the relative alliances on the King - Aristocracy - Plebs triangle. For me as
Re: (Score:2)
Is Mistral moving?
Re: (Score:2)
Unknown. Mistral isn't even a year old at this point, and when it was founded, Breton didn't even announce the regulations yet.
But if they choose to stay in France, they will be crippled by regulation from being able to develop practical applications in LLMs. But since their main expertise is in LLMs for European languages that aren't really present in US or China, they're stuck between a rock and a hard place. Stay and you get crippled by regulations. Leave and you lose access to people who are native Fren
Re: (Score:2)
Sit down. What I'm about to say will shock you to your core.
You can describe an issue without that issue being something you deeply care about.
Re: (Score:1)
Whoosh.
Re: (Score:2)
But can it run Crysis? (Score:2)
Re: (Score:1)
wrong answers only, please
Yes
Re: (Score:1)
Crysis does not exist, it's just an urban legend like Polybius.
Re: (Score:2)
Only Weird Al's PC can run Far Cry.
Re: (Score:2)
Dude... it's 2024. Even the shittiest PC integrated graphics can run Crysis effortlessly. The meme is dead, it's time for a new game.
Nope (Score:2)
This is just bullshit. "AI" (dumb as it is) needs lots of hardware and constant model maintenance. Even if it proves somewhat generally useful after all (which is in no way assured), it is going to be a service, not something running locally.
It seems "AI" is just the latest excuse for moron "executives" to predict the world is going to change fundamentally. It is not. Caveat emptor.
Re: (Score:2)
Most of the companies in question shifted some parts of their business model to datamining their users.
AI is going to be exceptional at datamining because it's self-adapting. AI offers to go a step further. After learning how each users workflow goes, AI can do things to slowly nudge user in the direction that company wants.
You're already seeing a very crude version of this with google gemini. Think that, but refined and working for not just genocidal Western Marxists who seek to make Europeans look like an
Re: (Score:2)
Remember that a lot of companies nowadays see it incumbent on themselves to direct your very thoughts in pre-approved directions. Both for the "good of the planet" and because it makes them money.
Companies: "It's just good business."
Intern: "You do realize what happened to the last person [youtube.com] that pushed that, right?"
Me: *Wakes up*....Damn it.
Re: (Score:2)
Remember when Microsoft said that every device was going to be a mobile device, and we got the abomination known as Windows 8? Same thing here--just replace "mobile" with "AI".
Re: (Score:2)
Indeed.
Re: (Score:2)
This is just bullshit. "AI" (dumb as it is) needs lots of hardware and constant model maintenance.
"AI" bullshit runs just fine on my desktop.
Even if it proves somewhat generally useful after all (which is in no way assured), it is going to be a service, not something running locally.
Personally I reach for (local) LLMs more than Google these days.
Other generally useful amazing shit include local STT that blows Google's voice recognition service away.
Leading open source text to image that outperform commercial services.
Multi-modal open source image to text that outperforms GPT-4.
All of which are not only freely available today they can run on any mid to high end gaming PC.
That's like saying (Score:1)
This is true... (Score:4, Informative)
Way back when, before the internet there was a similar sentiment to "Every computer is going to be connected to the internet"
As a case in point, I have copies of both piper-tts and whisper on my laptop, both of which are AI tools for dealing with voice generation and recognition respectively.
I imagine people using more imagery or videography to have AI tools installed to aid in their workflows.
AI is here, it is useful, and it is likely to get more advanced and inclusive.
The kicker here is that you want the AI to be running on your own machine... If you use some free web-based service, you are the product to some advertiser.
-Evert-
Market bubble going up! (Score:2)
It seems like all of the big tech companies are pushing this imaginary "AI" scheme. I'm sure it's profitable at the moment but, there's nothing but a breath of air in those bubbles, and the soap is drying out.
So, what are we thinking? NASDAQ collapse next year, or later this year? Major recession?
Re: (Score:2)
So, what are we thinking? NASDAQ collapse next year, or later this year? Major recession?
You should ask ChapGPT, and then post an article somewhere about it, like that's meaningful to have a language model "predict" the future.
nonsense (Score:2)
As to making anything something-proof, that always blows up due to some unconsidered consequence. . For example, in the 1970's, American car manufacturers desired to simplify the production process, implementing what they termed. "making it idiot proof".. The blowback of calling your workers idiots was the production of the sh
We'll have fully self driving cars next year!! (Score:2)
For Windows at least. (Score:2)
Phones will be first but mark my words, as soon such a chip becomes commonplace in PC's, Microsoft will start requiring it.
Software adoption trails software availability... (Score:3)
At some point during the '90s we were told that "Every PC will become a FP Powerhouse" because intel started integrating the Math coprocessor, and some/many people/companies did not believe it. And here we are, in this brave new world, were every single PC has a math coprocessor.
Then Some people said that the future were superscalar and superpipelined PCs, and some people/companies did not believe it. And here we are, even lowly Atoms are superscalar superpipelined
Then some people said that all PCs would be SIMD, and some people/companies said no. And here we are.
The some people said that integrating the memory controller and the PCI controller into the processor was the future, and some people/companies said no, and here we are
Then some people said Chiplets were the future of the PC, and there we are going in that direction.
Of course AI PCs are the future, just like math coprocessors, graphics chips , PCI controllers, TPU/NPUs will become part of our processors. And just like Quake took advantage of the math coprocessors two generations after implementation (Quake was coded to profit from the superior throughput of the OG Pentium), and some other bright sparks took advantage of GPU acceleration, in a few PC generations some bright spark will take advantage of the NPUs, software adoption trails hardware availability, and not the other way around.
Re: (Score:2)
At some point during the '90s we were told that "Every PC will become a FP Powerhouse" because intel started integrating the Math coprocessor, and some/many people/companies did not believe it. And here we are, in this brave new world, were every single PC has a math coprocessor.
Then Some people said that the future were superscalar and superpipelined PCs, and some people/companies did not believe it. And here we are, even lowly Atoms are superscalar superpipelined
Then some people said that all PCs would be SIMD, and some people/companies said no. And here we are.
The some people said that integrating the memory controller and the PCI controller into the processor was the future, and some people/companies said no, and here we are
Then some people said Chiplets were the future of the PC, and there we are going in that direction.
Of course AI PCs are the future, just like math coprocessors, graphics chips , PCI controllers, TPU/NPUs will become part of our processors. And just like Quake took advantage of the math coprocessors two generations after implementation (Quake was coded to profit from the superior throughput of the OG Pentium), and some other bright sparks took advantage of GPU acceleration, in a few PC generations some bright spark will take advantage of the NPUs, software adoption trails hardware availability, and not the other way around.
Agreed, but I'd add that whatever software comes along to make use of those NPUs will quietly drop the 'AI' moniker, as to get people to stop associating said hardware with marketing drivel and hype.
Focus on the Products, not the Tool (AI) (Score:1)
Why is everyone trying so hard to convince us that AI is so revolutionary? Focus on (making) actual products that improve our lives through AI, not the tool. Anything outside that is barely relevant.
It's like reading about a cure for cancer for the past 20 years. Who cares about the recurring promising developments. Talk to us when there's a cure for cancer. Anything before that is barely relevant.
Thought police iow (Score:2)
Early 90s version: (Score:2)
This is essentially the 2024 version of "every PC will have a math coprocessor someday" which is completely obvious for anyone who has worked the hardware side of this business for any time at all - as technologies mature, they get consolidated into other existing tech, and become standard features instead of expensive customizations.
See: GPUs, sound cards, mice, FPUs, storage controllers, and basically everything you find integrated on a modern PC motherboards and modern processors.
Re: (Score:2)
See: GPUs, sound cards, mice, FPUs, storage controllers, and basically everything you find integrated on a modern PC motherboards and modern processors.
Sometimes we move backwards, we lost the "turbo switch". :-)
What about bandwidth? (Score:2)
Modern CPUs can already saturate the equivalent bandwidth of a dual channel DDR5 when used for generative AI without resorting to any of the processors AI specific instructions.
Say you have a fancy NPU that runs circles around the CPU for certain AI tasks.. where is the bandwidth going to come from to make that useful? Is the intent to only limit this to AV processing?
Re: (Score:2)
Didn't you read the summary? Higher-quality background blurring in Teams meetings! The killer feature that will render non-AI computers obsolete!
But there is something here. iPhones now have NPUs, and I confess to being unaware of the class of apps that makes use of this. Given that the AI that has amazed the world is cloud-based, feeding on massive data models that took mind-boggling amounts of computing power to pre-process, a personal NPU is probably doing something far more modest.
Perhaps Dell was hypin
Your life (Score:2)
In Soviet Russia, you were the one things happened to.
With social media, you were the product.
Now with AI, you are the fertilizer.
Unlikely... (Score:2)
My AI PC will sit right under my 3D TV
Re: (Score:2)
Re: (Score:2)
You'll own your 3D TV, VR googles, and AI PC as NFTs.
6G 4D AI with Quantum DRM (Score:3)
The PC makers & MS are looking for a way to convince consumers they are obsolete so they buy new shit, not unlike the 5G hype.
Re: (Score:1)
Obligatory [youtube.com]
definition (Score:2)
I am sure they will stretch the definition of what an "AI PC" is until every PC will fit it. As in your computer runs Windows, Windows includes MS Paint, MS Paint had an "AI" tool for background removal, therefore is an "AI PC". Or the new Clippy/Copilot.
I thougght I had left home and gotten away? (Score:2)
AAARRRRGGGGHHHH!
What could possibly go wrong? (Score:2)
As far back as 1889 there was a novel predicting the rebellion of sentient machines, "The Wreck of the World". Don't say they didn't warn you.
Every PC will have anti-white biased AI then? (Score:2)
As we have seen with Google Gemini and OpenAI ChatGPT, these companies have hardcoded and / or specifically trained their AI models to have a very large anti-white bias in everything they do. Not just a little, but actually and openly refusing to produce an image of ANY white person doing whatever, even producing factually incorrect pictures of historical persons or persons from bygone eras. Not just that, but then the companies behind them started banning accounts who even tried to have their AI produce th
Re: (Score:2)
It will spew off atheist mainstays all day long, of course, which was to be expected. You can get it to discredit Christianity for as much as you like.
It's funny that you mention Christianity, but not Islam. Because if they also discredit Islam, it would be Islamophobia, which is not at all expected from a thing created by liberal-minded people.
WTF, Dell? (Score:2)
What do you think the cloud is for? Why pollute all PCs with this bullshit? You want to serve AI, then put it into your cloud platform. If some end user wants their own AI, let them purchase a separate GPU or AIPU board or whatever.
filtering (Score:2)
It will be an AI PC that will filter what I can/must see or do according to what MS deems appropriate.
Just had a horrible thought ... (Score:1)
Can you imagine just how unusable it will become when they AI the shit out of it? You will spend your entire day fighting with it.
So glad to be retiring soon. The workplace is gonna suck so bad in the coming years.