Have You Played With Microsoft Copilot?

[ArdWrknTrk]

Gary Lewis wrote

Yes, the name change didn't fix anything.  It still has a long way to go.


So Google re-routed us way north and back down into town.

Maybe Microsoft's Copilot has a long way to go, but you literally turn around and mention Maps doing realtime problem solving on the fly...


And with another node of GPT models Carnegie Mellon is groundbreaking in chemistry: https://arstechnica.com/science/2023/12/large-language-models-can-figure-out-how-to-do-chemistry/

[Gary Lewis]

Google Maps, and probably Apple Maps, do good problem-solving on the fly.  But only to a point.

On Sunday when we left the first ER to go to the 2nd ER to get a CT scan we'd had a torrential downpour here in Charleston.  What we didn't know, and Google was finding out sporadically, was that the streets between us and the 2nd ER were closing.  Janey was seeing the streets close ahead of us and we'd get re-routed, only for the new route to close and another open.

We played this game for a while until our daughter called the 2nd ER and found that ALL of the streets to it were closed.  So I'm guessing that it was an input issue for Google, that it was getting poor and limited input on which to calculate routes.

And yesterday Janey and her sister were in old Charleston and asked me to come and get them.  I was tracking her phone and thought I could get dynamic directions to them, but after I hit start they moved and the directions didn't update.  I'm surprised, but it didn't work.

Still, it is far easier to use either of those mapping systems than the one in either of our vehicles, and both Google and Apple have essentially the latest info where the cars don't.  Just try driving through Birmingham with a 10 year old nav system - it doesn't work as all of the roads have changed.

So AI is here, and there, but not everywhere.

[ArdWrknTrk]

Mapping isn't problem solving, and neither Google or Apple are prescient.
I'm kind of surprised that the results were so granular WRT to exactly where cars were stopped and diverted.  Though you are the product, and Its only really people with the app open that are informing the algorithm.

Learning models are another thing entirely.
They train on data or their own experience and will know or anticipate something like heavy traffic on a certain route at 5:00(for example)
Another thing these driving apps do is try to push drivers to alternative routes to lessen major congestion on what would be the 'main' route.

Now that LLM's easily pass the Turing test were going to have to come up with another definition of sentience or accept that silicon is a living, if not breathing, "being"

I can't imagine the advances you've seen nor how you can't be gobsmacked with this new normal.

[Gary Lewis]

I have seen a lot of changes.  Forgetting that I was born in '47 and the very few computers then were huge in size and almost incapable of having their programs changed, computing in and of itself has changed exponentially - sizes are so small and capabilities so huge.

As for AI, I remember starting to talk about it in about '88.  We were hoping to use it to solve all sorts of problems, like with oil exploration and product shipping.  But I don't think we were ever successful.  Maybe we weren't smart enough to wrote code that allowed learning.

But you are right about learning models.  If we could get even the mapping programs to do that some of the congestion we are experiencing in driving could be eliminated.  Just take some of us a slightly different way than the fastest and we'll all get there sooner.

As for not being gobsmacked, perhaps that's because of my career in IT.  Or maybe my career in IT was because of my interest in technology.  Chicken and egg?  In any event, I watch others only half my age ignore the benefits of the tech and shake my head.  Why, at the age of 76, am I the one to create the church website, set up the wifi, install the security cameras and the wifi-enabled thermostats?  Not that I mind doing it, but I would have thought that the more switched-on young adults would have jumped at the chance.  

[ArdWrknTrk]

The cesspit of social media?
I imagine those with background and a job in technology (wether it is installation/maintenance, network or data administration) are too busy or burnt out to volunteer for more

[ArdWrknTrk]

Rather basic comparison of silicon computers to the human brain from one of the channels I follow:  https://youtu.be/FegeRT5N3A4?si=-0jiqFCmb78idmk_

(no my phone still will not display an imbed link)

[BigBrother-84]

Very interesting!
Thanks Jim!

[ArdWrknTrk]

https://arstechnica.com/apple/2023/12/apple-wants-ai-to-run-directly-on-its-hardware-instead-of-in-the-cloud/

"While Microsoft and Google have largely focused on delivering chatbots and other generative AI services over the Internet from their vast cloud computing platforms, Apple’s research suggests that it will instead focus on AI that can run directly on an iPhone."


This has interesting implications for battery life and processor architecture (as discussed in the video above)

[ArdWrknTrk]

Gary Lewis wrote

But you are right about learning models.  If we could get even the mapping programs to do that some of the congestion we are experiencing in driving could be eliminated.  Just take some of us a slightly different way than the fastest and we'll all get there sooner.

The mapping apps do exactly that.
I guess I wasn't clear enough in my post above?  

So kind of like a bullet and a fast pitch may have the same kinetic energy but vastly different consequences map apps try to avoid gridlock by directing some users around choke points.

[BigBrother-84]

ArdWrknTrk wrote

" [...] Apple’s research suggests that it will instead focus on AI that can run directly on an iPhone."

Maybe one day our supposedly smart phones will finally be.


In french they call them "téléphones intelligents".  Still looking for this brightness feature...

[ArdWrknTrk]

Gary Lewis wrote

Interesting article, Jim.  Very scary!  They've apparently reigned it in a bunch as it seems fairly tame at the moment, but it is very, very obvious that this may be Alpha, not Beta.  A lot of things on Copilot do not work well at all.

 Relevant articles discussing Copilot's former egos

https://arstechnica.com/information-technology/2023/02/ai-powered-bing-chat-loses-its-mind-when-fed-ars-technica-article/

https://arstechnica.com/information-technology/2023/03/microsoft-equips-bing-chat-with-multiple-personalities-creative-balanced-precise/

So it not only been lobotomized, it's got DID

From the first article:  "In the face of a machine that gets angry, tells lies, and argues with its users, it's clear that Bing Chat is not ready for wide release."
"Already, Bing Chat is known to spit out erroneous information that could slander people or companies, fuel conspiracies, endanger people through false association or accusation, or simply misinform. We are inviting an artificial mind that we do not fully understand to advise and teach us, and that seems ill-conceived at this point in time."
"And that's why Bing Chat is currently in a limited beta test,"

This points to the Beta being last February...

[ArdWrknTrk]

BigBrother-84 wrote

In french they call them "téléphones intelligents".  Still looking for this brightness feature...

On my Android the brightness slider is at the bottom of the pulldown shade.....  

[Gary Lewis]

Jim - I missed that the mapping apps already send people around choke points.  Glad you pointed that out.

As for Bing Chat/Copilot, those articles are chilling.  But, if you think about it, not all that un-expected.  The thing has a personality and they've not taught it that it needs to admit when it is wrong.  We have to teach kids that, so why not computers?

Anyway, this is interesting but I'm the taxi driver taking various parts of the family to see things and bring them back, as well as work on Lil Blue a bit until TJ gets here this afternoon.  So I've gotta go...  

[ArdWrknTrk]

I'm sure you're busy shuttling family around.

No, it's not unexpected. Realize they are starting with an infant.

That's why I focus on the learning model.
But when you have a chatbot that is not coded to "admit" that it may be wrong you have a dystopian sci-fi story.
What happens when one of these gets a hold of SCADA controlling the web or power grid?

"Open the pod bay doors HAL"

[Gary Lewis]

Somehow I see Slim Pickens, slapping his hat.  

[ArdWrknTrk]

I don't think there's ever going to be another nuclear strike launched from aircraft.

[ArdWrknTrk]

https://www.theverge.com/2023/12/26/24015198/microsoft-copilot-mobile-app-android-launch

[Gary Lewis]

[85lebaront2]

I can't do that Dave.

[ArdWrknTrk]

"I'm afraid I can't do that, Dave".