Have You Played With Microsoft Copilot?

[ArdWrknTrk]

None of it surprises me but there seem to be an awful lot of people who should know better that believe it.

Cohen is a fine example.
As a lawyer who is looking at going to jail would you rely on some unsubstantiated babble from a chatbot and put it in front of a judge?  

[Gary Lewis]

The vast majority of people don't know that it is unsubstantiated babble.  And therein lies the problem.  I assumed that it wouldn't lie, but it absolutely does.  And I assumed it would follow explicit directions, and it absolutely refuses to do so.

It really needs to print a disclaimer that says "Everything I tell you is suspect as I make things up and cannot be bothered to follow your directions."  But people wouldn't read that, so why bother.

[ArdWrknTrk]

IOW, the VAST majority of people are complete idiots.

[Gary Lewis]

I would say "uninformed".  All of the hype in the news is about how it can produce fakes.  As you've found there are writings out there that prove that it lies and won't follow directions, but you have to go looking for them.  So unless you've played with it and caught it in the lies and/or not following directions you won't have been informed.

Having said that, there sure are a bunch of idiots out there.  

[ArdWrknTrk]

Maybe I'm just jaded and cynical?
Maybe it's the autism?

There's NO WAY I would ever put my fate in the hands of AI as it stands now.

[ArdWrknTrk]

https://arstechnica.com/information-technology/2024/02/amnesia-begone-soon-chatgpt-will-remember-what-you-tell-it-between-sessions/

Gary, you were saying there was no persistence of "thought" with Microsoft's chatbot.

[Gary Lewis]

That proves what I experienced was real, not just perceived.  And each chat has a finite limit, so just when you are "getting there" you have to quit and start over.

But while having it remember what it learns from chatting with me would help, that means it doesn't know what it has learned when you ask a question.  So somehow it needs to be able to connect things.  And that gets scary!  

[ArdWrknTrk]

All it takes is a neural network that constantly replenishes itself, and circuits get programmed like instincts in animals.

Musk's neuralink is scary indeed!
But when it's developed we'll find out if this is the matrix, or minority report.

Huxley was more prescient than satirical.  🤯
Just like Judd was with Idiocracy!

[ArdWrknTrk]

https://arstechnica.com/information-technology/2024/02/google-upstages-itself-with-gemini-1-5-ai-launch-one-week-after-ultra-1-0/

[Gary Lewis]

"It's impressive to process documents that large, but the model, like every large language model, is highly likely to confabulate interpretations across large contexts. We wouldn't trust it to soundly analyze 1 million tokens without mistakes, so that's putting a lot of faith into poorly understood LLM hands."

[ArdWrknTrk]

Interesting quote... 😕

https://arstechnica.com/tech-policy/2024/02/air-canada-must-honor-refund-policy-invented-by-airlines-chatbot/

😄😆😅😂🤣

[Gary Lewis]

That's funny!  And telling.

At least right now chatbots cannot be trusted to tell the truth, so disclaimers must be made.  But if you tell me that the information you are providing might not be accurate then I'm not going to believe anything you say.  So what's the point?

[ArdWrknTrk]

https://arstechnica.com/information-technology/2024/02/google-goes-open-ai-with-gemma-a-free-open-weights-chatbot-family/

[Gary Lewis]

Interesting.  I noted the Python capabilities.  I wonder if that's what could be used to port our Nabble posts to a new forum?

[ArdWrknTrk]

🙄. There may have been a reason I linked this....

[Gary Lewis]

Assumed so.

I know very little about Python as it was developed long after I passed my level of incompetence.  But it is intriguing to think that we could have AI write a program to actually migrate us over.

I can see two ways of doing that - write the program to play like a user and actually post exactly what's on Nabble.  But that would be tedious as we have ~2000 users and would have to create all of those accounts so it could log on and post.  Plus it would probably take a long time.

The other would be to study both file structures and recreate the files in the new forum's format.  But the user's passwords are encrypted so while we might be able to move the content I'm not sure we could set it up so they could automatically log on.

[Gary Lewis]

A guy I worked with at Computer Sciences sent me a link to this video from 60 Minutes regarding AI. It is slightly out of date, but is still interesting nonetheless.

I got a kick out of the reporter seeing what it had created with regard to her work experience, and it reminded me of the "Bob" that was a Ford engineer that Copilot created. Apparently they call it "hallucination" and they don't seem to be too worried about it. But to me it is a show stopper.

On the other hand, the ability for robots to learn to play soccer on their own is amazing. So I can see how this is going to be a game changer.

And the data center in Pryor is on the land that was the "powder plant" during WWII. I've hunted that land for quail, fished in the creeks, and picked up pieces of really hard aluminum from the scrap pile.

[BigBrother-84]

Lol!

Gary, I suppose this video contains some dark military secrets?  Not available outside US.

[Gary Lewis]

Well, that's a bummer!

One of the points is that there will be MAJOR changes in our lifetimes.  One of the cool things was that they built robots that didn't know how to walk, much less run, and they told them the rules of soccer - and the robots quickly learned how to walk then run and to play soccer - all on their own.

[ArdWrknTrk]

The carriage house I mostly grew up in was previously occupied by this guy.... 🙄
Maybe there's something in the water???

https://www.invent.org/inductees/george-devol#:~:text=In%20the%201930s%2C%20George%20Devol,a%20drum%20or%20other%20medium.

https://www.invent.org/blog/inventors/George-Devol-Industrial-Robot