Question Confused newbie questions about AI and Co-Pilot ?

[Minaz]

I thought I was getting a handle on AI until the new Co-Pilot PCs came out. Before that, I understood that AI was sort of like a server-side technology, where you input some question or task, it gets sent to a remote AI in the cloud and you get a response/art piece/homework done/etc...

Now, it seems that AI is living on your computer. What? What exactly is going on?

Firstly. the name Co-Pilot PC seems confusing considering all I have seen is Co-Pilot laptops. Are there Co-Pilot desktops?
Why is Co-Pilot only available on laptops if that is indeed the case?

Then I heard that Co-Pilot processors or NPUs, only work with Snapdragon, but Intel is coming out at the end of the year, yet at the same time, Lenovo and other mfgs are saying you can buy an Intel Co-Pilot laptop right now? Again what? Did they just move the schedule up a couple of months?

And what exactly is Co-Pilot supposed to do? I heard about the recall function, frankly that is the last interesting to me, I don't think I will need it. But there are supposedly a whole bunch of other integrated AI things. I fear this is going to be like Amazon Alexa, where there are so many niche functions that you forget what the AI can and can't do, and it becomes really frustrating to find out. What are the killer applications for Co-Pilot?

Also, it seems you need at least 40 TOPs or you can't use Co-Pilot. This is different from system-wide TOPs apparently, which you may have alot of, but apparently MS doesn't care, it only wants NPU TOPs for Co-Pilot. Why? What is the difference? And is there a way I can find out how many TOPs, if any, my ancient PC has systems wide? Also, apparently, RTX40XX cards have some system TOPs, what about RTX30XX, or is that too old for TOPs?

And how many TOPs is ideal for buying a new computer/laptop anyway? More the better, or is 40-45 enough? What will getting more TOPs do for you anyway? Considering from a typical consumer end-user POV: i.e for someone who might use their computer for a mixture of gaming, productivity, media-consumption, blogging/vlogging, photo and video editing, coding, and network management?

Hopefully someone can straighten this out for me, thanks! (And yes I Googled it, no good - just got more confused).

 

[Ralston18]

Not (full disclosure) a technical answer....

= = = =

Regarding:

"Now, it seems that AI is living on your computer. What? What exactly is going on?"

The cynic in me feels that "AI" wants to do and rule everything rightly or wrongly as long as it all yields clicks and thus profits.

What I really believe is that my cynical side may be correct.

Just my thoughts on the matter.

 

[onespeedbiker]

This is not intended to be definitively accurate, contains a lot of questionable personal opinions and glosses over a lot.

First AI, is nothing more than an advanced algorithm designed to learn data and mimic human responses. It also can only learn what it's taught, so it's incapable of creating new thoughts or ideas. It also is incredibly processing hungry which lends itself to traditional GPU type processing incorporating NPU (ARM) over old Intel/AMD CPU x86/x64 processing ( Windows 11 on Arm can run x86 and x64 applications through emulation).

The hardware requirements for Windows CoPilot+ is built on top the hardware requirements for Windows 11, which have been around since 2018, but the Windows 11 CoPilot+ hardware requirements have only been incorporated into computers in large part just this year 2024, meaning there is software and operating difference between Windows 11 and Windows 11 CoPilot+ and another difference with CoPilot+ whether you are running an ARM processor or new Intel processor and if the software is running applications emulating x86/x64 or when/if those applications with be ported for ARM. Windows big issue seems to be convincing most computer users that any of this is a substantial improvement over Windows 10, which so far has not been very successful.

 

[ubuysa]

First AI, is nothing more than an advanced algorithm designed to learn data and mimic human responses. It also can only learn what it's taught, so it's incapable of creating new thoughts or ideas.

I'm no expert here but generative AI (as opposed to discriminative AI) is capable of creating new data and new models by analysing the training data and models. It basically looks at the distribution and probabilities in the training data and generates new data and new models based on those probabilities and distributions. You're absolutely right that generative AI is not thinking, in any sense of the word, and it cannot make intuitive leaps like humans do, but generative AI can 'learn' in the sense that it can move from what is known (the training data) to create something new that didn't exist before but which conforms to what it's learned.

This is where I suspect (but I don't know) that generative AI will be used in products like Windows. You can think of it as though generative AI uses what most users have done in a particular set of circumstances and can thus offer an 'intelligent' set of (possibly new and different) options to this user who is doing something different but similar.

 

[onespeedbiker]

Since generative AI cannot create “original” images or data from zero, it needs a pre-existing background/data input to feed the algorithms through a machine learning model. It also doesn't have a sense of morality or truthfulness or even humour, something to keep in mind considering the fact that Generative AI primarily uses the Internet as it's learning data base. In many cases, where data is limited or non-existent, it simply creates information by parsing unrelated data together or plagiarizes long series of text. One thing that is quickly becoming an issue is the lack of data available for learning. Apparently, there is a huge dearth of data on the mechanics of how the human mind works on a day to day basis to accomplish even simple tasks. One issue that was learned when trying to mimic the human brain when trying to train robots to act like humans, is it turns out we use the majority of our brain's capability to filter out stimuli that diverts attention from task(s) at hand (distraction); in psychology, the inability to filter out the onslaught of sensory stimuli, is one of the key symptoms of schizophrenia. The point being if everything seems equally important, you are incapable of making rational decisions and will quickly go insane.

From Forbes magazine, "Despite their impressive output capabilities, GenAI applications are limited in their ability to tackle complex, multi-dimensional societal issues. They excel in defined, narrow tasks but lack the general understanding needed to address broader challenges such as strategic decision-making or ethical dilemmas." May 9, 2024