Latest in AI News(Test forthcoming)

[WildWestDesigns]

New AI experiences for Paint and Notepad begin rolling out to Windows Insiders

Hello Windows Insiders, today we are beginning to roll out updates to Paint and Notepad to Windows Insiders in the Canary and Dev Channels on Windows 11. Paint (version 11.2410.28.0) Generative fill

blogs.windows.com

Now, I know everyone here loves MSPaint apparently MSPaint and even Notepad get AI abilities. Going to have even more AI abilities interwoven into the Windows experience.

I do have to wonder what the minimum specs are going to be for Windows (with AI integrated into quite a few already bundled programs) and what the min. specs are going to be for the additional programs that have their own AI services bundled in to the programs that professionals will need to use adding on top of what Windows will need for it's AI needs.

 

[White Haus]

I do have to wonder what the minimum specs are going to be for Windows (with AI integrated into quite a few already bundled programs) and what the min. specs are going to be for the additional programs that have their own AI services bundled in to the programs that professionals will need to use adding on top of what Windows will need for it's AI needs.

Good point, this will probably give the computer manufacturers an excuse/opportunity to sell premium builds designed to handle AI's demands, where as the basic (cheaper) out of the box computers eventually won't have the processing power to handle the new technology.

 

[WildWestDesigns]

Good point, this will probably give the computer manufacturers an excuse/opportunity to sell premium builds designed to handle AI's demands, where as the basic (cheaper) out of the box computers eventually won't have the processing power to handle the new technology.

And the irony is, this tech isn't really necessary, not even from an improved efficiency standpoint (it's mainly a scraper at this point, kit bashing the results (which they may not have the license to do so, which means that the results are of dubious license) with various levels of success, akin to auto conversion of years past). Those that do find it efficient probably (but not always the case) have a harder time to get to that point on their own to begin with(this will especially be the case with future generations that have only been taught this tech).

Those that know what it's abstracting and be able to fix the results are going to be fewer and further between when starting to get to the 3rd and 4th generation of users (just have to look at the younger users of desktop computers now to see how that will play out, but I digress).

 

[JBurton]

Honestly

I do have to wonder what the minimum specs are going to be for Windows

Am I wrong in reading that these will not necessarily be onboard ai, as it will require a logged in account? So minimum specs won't be impacted, aside from always on internet...

 

[WildWestDesigns]

Honestly

Am I wrong in reading that these will not necessarily be onboard ai, as it will require a logged in account? So minimum specs won't be impacted, aside from always on internet...
View attachment 174741

No, there is local processing. Akin to telemetry as well. How much extra that it has to gather in order to have "quality" results is the thing. The online account would be to connect to the service and actually have it to use (which we all know that is yet another SaaS to go up in price). I'm sure that there is going to be some base usage that may be free, anything extra would be paid for, of course, I'm willing to bet that the telemetry will be there regardless). Also, even if some of this work is offloaded to a server, will have latency as well, internet connection, power that one has to have the program gather/send/receive/parse/show the results. Add more and more details, more and more services, this gets impacted quite a bit.

I'm willing to bet that it's going to "need" a whole lot of telemetry to help improve the results as well. And that telemetry gathering, collate, sending, handle the results given back is where the hit is going to come from.

I remember they said that this was only going to be with specific hardware (NPUs etc) for best results, and it appears that that isn't the case.

 

[Gino]

Glad I'm on my way out...... this sh!t is getting too weird for me. If I ever retire, I gonna draw things with my hands and eyes and paint pictures. Then make something in my woodshop.

 

[WildWestDesigns]

Glad I'm on my way out...... this sh!t is getting too weird for me. If I ever retire, I gonna draw things with my hands and eyes and paint pictures. Then make something in my woodshop.

Older that I get, the more I go that way as well. Although my painting skills are moderate. I'm more towards drawing/animating. Taught my kids (even though computer animation (both 2D and 3D) were coming into strides when they were younger (especially the youngest) the old animation techniques (flip books, zoetrope, phenakistiscope, thaumatrope, cut out and stop motion (my youngest always liked the Harryhausen movies)).

Why things are going just takes all the heart and soul of it out. But I digress.

 

[Solventinkjet]

AI isn't super resource intensive once the model is trained. That won't stop the computer manufacturers from selling you an "AI ready" computer with overblown stats for what people will actually need.

 

[WildWestDesigns]

AI isn't super resource intensive once the model is trained. That won't stop the computer manufacturers from selling you an "AI ready" computer with overblown stats for what people will actually need.

It's always being trained. Some of the problem now is more incestuous, it's training on it's own renders or other chatbot renders. Of course, also have things like nightshade trying to poison the well.

The best change when it comes to more or less "fully trained" AIs are going to be the more local ones (which big companies will more likely have to have if wanting to trademark the results here stateside (unless they have the pull to get the laws changed, which may be possible), that don't have access to the WAN and once they have a certain threshold of internal learning, it is only a matter of the iterative results after that. However, when talking about these major ones that people will subscribe to, that have access to things like Adobe's stock site (or any other stock site), those are going to be more or less constantly learning (unless it reaches to the point to no one is uploading more of the human created renders).

Tinfoil hat (although the G rated version), I can see big corporations liking this, because they can have a skeleton crew of the more junior artists (thus pay less) and have them churn out more "content".

 

[Graphic Extremes]

New processors are having a separate section just for AI to use.

 

[WildWestDesigns]

New processors are having a separate section just for AI to use.

That is only some of them and they were just recently released and at the time, MS was saying that those were the only ones that were going to be able to run AI on them (which turned out to not be true (same thing with Recall (another AI feature as well)), but these updates are not just going out to those NPUs, they are coming out to the masses as well.