9 minutes ago, booga said:

AI tools are untrustworthy when it comes to factual information. They frequently state falsehoods as if with absolute certainty. Nothing they say should be passed on without extensive questioning and research - which begs the question: why use them for factual queries in the first place?

I accept they may have other uses - I have been told they're quite useful for generating code, for example. 

Hi @booga

How are you?

It’s not just for code, but a there are already practical implementations of technology from [Customer Request evaluation -> Requirements -> Design and Coding -> Testing]

…in many sectors of the industry.

49 minutes ago, P_7878 said:

Hi @booga

How are you?

It’s not just for code, but a there are already practical implementations of technology from [Customer Request evaluation -> Requirements -> Design and Coding -> Testing]

…in many sectors of the industry.

Indeed. By the time I left university, I could make out that professors were using it too!

1 hour ago, booga said:

I've never asked an AI tool an in-depth factual question regarding a subject about which I am reasonably knowledgeable and not received an incorrect answer.

Then you don’t know how to use it. 🤙

1 minute ago, UrgentSiesta said:

Then you don’t know how to use it. 🤙

Sigh.

3 hours ago, booga said:

AI tools are untrustworthy when it comes to factual information. They frequently state falsehoods as if with absolute certainty. Nothing they say should be passed on without extensive questioning and research - which begs the question: why use them for factual queries in the first place?

We have to be careful while making such statements. Toos are essentially input/output functions.  You cannot expect a tool to provide accurate output, if the input itself is inaccurate. With inbuilt feedback mechanisms, the AI tools are now becoming more and more powerful, in industrial analysis.

In (our) aerospace technology evolution, FBW and AI are getting connected. Modern FBW systems use complex, rule-based algorithms (a form of AI) to manage flight. We are always uncomfortable to let go of some of that human control, but that was exactly the vision of Bernard Ziegler, a former French Air Force Pilot, legendary Airbus Test Pilot, and the father of FBW. He knew how to fly but wished to make it easy for others...

Here is a snippet from my (own) writeup:

"One of the most memorable and famous of Bernard Ziegler's quotes is this, "I want to design an airplane that my concierge can fly..."...🙂...a provocative expression of the visionary's thought-process, that has evolved (and still evolving) to forms of FBW that are the norm in today's modern aviation, irrespective of whether with an Airbus sidestick, or with a traditional Boeing control yoke...because it's said that Bernard Ziegler knew that not everyone can fly like a US Air Force test pilot..."

People resisted his vision with opposition and skepticism (just as we do for AI), while we also now find ample presence of FBW in 787 and (to be) 777-X. Likewise, AI, when, utilized properly, will provide advancements in technology, unheard of before. We need to adapt to it within its limitations...or be left behind...just as what happened with FBW...noting that the A320 family that introduced FBW for the first time, has now overtaken the 737 family...as the most-delivered jetliner...

Edited March 2Mar 2 by P_7878

55 minutes ago, UrgentSiesta said:

Then you don’t know how to use it. 🤙

If I do not learn how to use AI fast enough or/and refuse to learn it, then, I might be summarily dismissed, but my children are relatively young, so, I have very little choice in the matter ...🙂...
 

1 hour ago, Joshua757 said:

Indeed. By the time I left university, I could make out that professors were using it too!

Joshua:

You are the new generation...🙂...Some of us here learned programming in the university with Fortran punched cards. 

Nobody even talked of any kind of AI then, though the ideas certainly existed.

1 hour ago, P_7878 said:

If I do not learn how to use AI fast enough or/and refuse to learn it, then, I might be summarily dismissed, but my children are relatively young, so, I have very little choice in the matter ...🙂...
 

I feel ya. 
 

but at least I’m still ahead of my kids (mainly because I’m an IT guy 😁)

2 hours ago, booga said:

Sigh.

You can “sigh” all you want:

if you don’t know how to get correct answers out of  AI, then you don’t know how to prompt it. 

Again, just like talking to kids. 

16 minutes ago, UrgentSiesta said:

You can “sigh” all you want:

if you don’t know how to get correct answers out of  AI, then you don’t know how to prompt it. 

Again, just like talking to kids. 

I know how to ask an unambiguous question. 

Anyway, back to the thread: do you think that the AI response trumps stearmandriver's?

25 minutes ago, UrgentSiesta said:

You can “sigh” all you want:

if you don’t know how to get correct answers out of  AI, then you don’t know how to prompt it. 

Again, just like talking to kids. 

By the way, you yourself said that "AI is the proverbial word not allowed savant, and it just can’t be trusted with many technical subjects." Either you don't know how to prompt it, or it can't be trusted - which is it?

35 minutes ago, booga said:

I know how to ask an unambiguous question. 

Anyway, back to the thread: do you think that the AI response trumps stearmandriver's?

You probably do. 
but asking unambiguous questions are not the key to success with AI. 
again, very similar to speaking with children, you have to not only know what you’re after, but how to elicit the answer you need. Which many times can not be obtained with but a single “unambiguous” question. 

Which leads us to:

24 minutes ago, booga said:

By the way, you yourself said that "AI is the proverbial word not allowed savant, and it just can’t be trusted with many technical subjects." Either you don't know how to prompt it, or it can't be trusted - which is it?

I know how to elicit the information I want, especially when it comes to subjects with which I’m conceptually familiar. 

I get what I want because im constantly using it and I know where it tends to go off the rails. 

So, many times it ends up being more like a conversation than a question with answer. 

Context, as they say, is everything. 

The example I gave wasn’t referencing AI’s well known challenges, it was giving an example of what happens when one uncritically accepts its initial answer. 

eg, I’m sure stearman could get very good flight info out of AI - because he already knows a lot about it. 

But if you just go at it like that kid I mentioned and you don’t know enough to spot some errors or inconsistencies, then yeah, you’re gonna run into trouble pretty fast. 

Like I said- it’s a tool. And it does actually take practice and learned skills to get the right information out of it. 

29 minutes ago, booga said:

By the way, you yourself said that "AI is the proverbial word not allowed savant, and it just can’t be trusted with many technical subjects." Either you don't know how to prompt it, or it can't be trusted - which is it?

Hi @booga

AI is Model driven. Depending on the problem and task on hand, you can choose the most appropriate model, such as "Thinking AI" that spends more time than normal to think the problem out in greater depth; or the "Auto AI" where you let AI do the whole lifecycle of the project; or the "Human input" based models, where you, as a human, is allowed to provide oversight (i.e., prompt for actions)...etc.

Each model has its pros and cons, and that's where the human judgment factors in. These techniques are all evolving (rapidly).

I have been mandated to attend a series of such courses over the last several months. I will next see how well it applies to my (own) specific (job) objectives.

@UrgentSiesta may add more. He seems to be more into it...🙂...

52 minutes ago, booga said:

I know how to ask an unambiguous question. 

Anyway, back to the thread: do you think that the AI response trumps stearmandriver's?

Pardon me, but I wish to stay out of it...🙂...

What happened to this thread?