Completely OT: Let's chat

[stanl]

Believe previous thread hi-jacked enough with chatGPT posts. But I am curious. At work [which has blocked all access to chatGPT endpoints, so my PS script cannot execute]... there is a lot of talk about "training' data and they point out to what is intended as the best OSS


https://github.com/Decentralised-AI/falcon-40b/blob/main/README.md


Pretty interesting reading... but now back to the subject. I did put together a simple GUI with 2 radio buttons got (1) GPT  (2) CoPilot... something I was just playing around with ... but the question was brought up about using chat to help with Winbatch [or other code]... a few posts were made about this on the other thread which illustrated that chat understood a little about WB.


So I decided to push it [hoping to get back more specifics about 'training' WB for chat]


so I asked a question based on one of the latest WB features - the map


write winbatch code to create a map for months of the year with numeric value as key




my GPT came back with:

Code Select Expand



monthsMap = CreateObject("Scripting.Dictionary")
monthsMap.Add(1, "January")
monthsMap.Add(2, "February")
monthsMap.Add(3, "March")
monthsMap.Add(4, "April")
monthsMap.Add(5, "May")
monthsMap.Add(6, "June")
monthsMap.Add(7, "July")
monthsMap.Add(8, "August")
monthsMap.Add(9, "September")
monthsMap.Add(10, "October")
monthsMap.Add(11, "November")
monthsMap.Add(12, "December")



and the Copilot [same question]

Code Select Expand



@echo off
setlocal
rem Create a map for months of the year with numeric value as key
set "months[1]=January"
set "months[2]=February"
set "months[3]=March"
set "months[4]=April"
set "months[5]=May"
set "months[6]=June"
set "months[7]=July"
set "months[8]=August"
set "months[9]=September"
set "months[10]=October"
set "months[11]=November"
set "months[12]=December"
echo %months[1]%
echo %months[2]%
echo %months[3]%
echo %months[4]%
echo %months[5]%
echo %months[6]%
echo %months[7]%
echo %months[8]%
echo %months[9]%
echo %months[10]%
echo %months[11]%
echo %months[12]%
endlocal



If you have read this far, I trust you understand this not a negative post. It is just these chat bots harvest known information so maybe a year from now the query would return results that more approximate WB code style. Or, maybe 'training' comes in to short-circuit the time. Just fascinating.

[td]

I am still waiting to see if the training* I have done on an ongoing chatGPT "conversion" is pickup by the OpenAI chatGPT model. (I doubt it but one never knows for sure about these things.)  It would be nice because inside the "conversion," chatGPT can write a fair number of respectable WinBatch scripts.

FWIW, OpenAI has training* APIs that allow you to train your own copy of a large language model. Interesting stuff.

* AI buzzology with multiple meanings.

[stanl]

I tried chatting with Dall-E and asked to create an image of a German Shepherd eating a Taco.

[td]

I have read some about but have not worked with Dall-E yet. Assuming images do not have restrictive watermarks, I can see where it might come in handy for a Website project.

[td]

On the darker side, not everything is sweetness and light:

https://www.itprotoday.com/application-security/chatgpt-hallucinations-open-developers-supply-chain-malware-attacks

[stanl]

I have read some about but have not worked with Dall-E yet.

It seemed to boil down to a REST query to an OpenAI Dall-E endpoint, but script needed an extra function to pre-calculate the time to return an image adjusting for pixel size... went over my head

[stanl]

And how can you resist a black German Shepherd as a soccer goalie

[stanl]

or a programmer, confused over meat or cheese Tacos

[stanl]

Given the proposed class actions against Open AI, Microsoft and Meta, I found this new player - talk to Claude


https://claude.ai/login


....pretty interesting

[td]

"It is difficult to make predictions, especially about the future."*  I hope I am wrong but I suspect that the Techbro AI hype could produce a tech bubble unseen since 2000. The resulting bubble burst may have far-reaching economic and societal effects. Perhaps the legal challenges will cool the hot air enough to prevent untoward outcomes.

*<unknown>

[kdmoyers]

It gives me ironic pleasure to quote ChatGPT:

The quote, "It is difficult to make predictions, especially about the future," is often attributed to a number of notable figures, such as Yogi Berra, Mark Twain, Niels Bohr, and even Winston Churchill. However, its true origins remain unclear.

One plausible source is Danish poet and humorist Robert Storm Petersen, known as Storm P., who is said to have made a similar statement. The physicist Niels Bohr, who was also Danish, may have picked it up from him.

In addition, there's a version of this quote in a report written by the Danish Parliament in 1937, and it's used to describe a common saying in Denmark.

Still, the exact origin of this quote remains uncertain. It seems to be a witty observation that's been made by a number of people in different forms over the years.

[kdmoyers]

Claude says:

That quote is commonly attributed to Danish physicist Niels Bohr, though its exact origins are unclear. Some key information about the quote:

- The full version is typically phrased as "Prediction is very difficult, especially if it's about the future."

- Bohr was known for his contributions to quantum mechanics and atomic structure in the early 20th century. He received the Nobel Prize in Physics in 1922.

- There is no definitive evidence that Bohr actually said or wrote this exact quote. It does not appear in his published works.

- The sentiment expressed in the quote - the difficulty of predicting future events - does align with Bohr's general philosophy and worldview.

- The quote is often cited as Bohr's because it sounds like the kind of dry, witty remark he was known for among colleagues and students. But its true origins remain uncertain.

- Similar versions of the quote have been attributed to other sources over the years, including baseball manager Yogi Berra, physicist Dennis Gabor, and even a Soviet academic journal from the 1950s.

- The first printed source is thought to be a 1962 newspaper article, 11 years after Bohr's death. So if he did say it, it was likely an oral remark that was informally circulated.

In summary, while Niels Bohr is the most common attribution, there is no conclusive evidence he originated the quote himself. But it reflects his recognized views on the difficulty of predicting the future in areas like quantum physics. The statement remains an enduring though uncertain quotation about the limits of prognostication.

[td]

I vote for Niels Bohr for reasons I won't go into but might have something to do with the concept of the Universe resulting from a random quantum event...

[kdmoyers]

In a nearby parallel universe, Bohr actually said "Prediction is a bitch."  However, in that universe OJ was elected president and ranch dressing doesn't exist, so... take your pick.

[stanl]

In a nearby parallel universe, Bohr actually said "Prediction is a bitch."  However, in that universe OJ was elected president and ranch dressing doesn't exist, so... take your pick.

and Ancient Aliens had the first hit single

[td]

Another dire warning on the AI crystal ball front:

https://gizmodo.com/gary-gensler-ai-to-cause-unavoidable-economic-collapse-1850929797

[jmburton2001]

The only thing that concerns me is criminals or other "bad actors" (including nation states) that will use this technology for nefarious purposes without regard for safeguards and consequences. Criminals by their very nature commit crimes with zero concern for their victims or collateral damage.

I wouldn't doubt that Dr. Evil's lairs are brimming with programmers finding new ways to screw over the masses.

[td]

It is not all that clear, to me anyway, that the crooks are more dangerous than Big Tech.

[jmburton2001]

I didn't mean to insinuate that "Big Tech" doesn't have its share of criminals. I figured "bad actors" covered the entire gamut of criminal elements, no matter where they exist.  ;)