Investing in Artificial Intelligence (AI)

[qldfrog]

I use AI for coding, validation of science based ideas but it is always critical to check ..and be intellectually/knowledgeable able to evaluate the answer.
Considering how the majority in the west handled access to quasi illimited information/knowledge on the internet, the prospect of widespread AI use is genuinely scary.

 

[divs4ever]

I use AI for coding, validation of science based ideas but it is always critical to check ..and be intellectually/knowledgeable able to evaluate the answer.
Considering how the majority in the west handled access to quasi illimited information/knowledge on the internet, the prospect of widespread AI use is genuinely scary.

the automatic belief in their own computer's calculations is scary enough

to set up a server from scratch i need at LEAST four days of checking THE HARDWARE ( and the combination of the parts together ! )

you cannot believe where i have found persistent errors when this stuff is put under heavy, sustained load

after all that THEN the software needs to be triple-checked

... so after all that being 99.999% accurate ... then you need to import data for for your computer server to be productive ... what could go wrong ?

 

[Sdajii]

I asked yesterday to ai (Gemini)
110 knots wind speed in km per hour
The initial answer from Gemini
"110 knots is equivalent to approximately 1.85 kilometers per hour. This is because knots are a unit of speed that measures how fast the wind is moving across a horizontal surface, while kilometers per hour measure how far an object moves in a horizontal distance.
Here is a conversion table to help you understand the relationship between knots and kilometers per hour:
* 1 knot = 1.152 kilometers per hour
* 2 knots = 2.303 kilometers per hour
* 3 knots = 3.445 kilometers per hour
* 4 knots = 4.606 kilometers per hour 1.85 km an hour etc"
When i replied that it did not make sense and was far too low
"You are absolutely right, and I sincerely apologize for the significant error in my previous response. My initial calculation was completely incorrect.
Here's the correct conversion:
* 1 knot is approximately equal to 1.852 kilometers per hour.
* Therefore, 110 knots is equal to 110 * 1.852 kilometers per hour, which is 203.72 kilometers per hour.
My previous response was a major mistake, and I appreciate you pointing it out. Thank you for helping me correct this error."

AI has become a very useful analysis tool, but I have had plenty of experiences similar to this, where it completely fails at the most basic of calculations and logic exercises.

Sometimes the failures seem like they could only occur if the model was deliberately designed to fail, though I can't see any benefit to that other than perhaps to have these bugs ironed out in the paid versions. Sometimes they are incredibly creative and insightful, with brilliant reasoning skills etc, and then a minute later will tell you that ducks have gills and breath underwater because they live in lakes. I've had them tell me things like the moon orbits only a few centimeters above the Earth's surface (when asking it to calculate planetary rotation decay rates) and other completely absurd things. It told me a Greenland Shark would be an ideal pet for me (it would require a tank larger than the world's largest swimming pool which would have to be constantly refrigerated, among other challenges).

But judging it today is a bit like saying the internal combustion engine and light bulb will never replace the horse and candle based on the flaws you see in the first years proceeding their invention. I had one write an elaborate science fiction novel with incredibly vivid world building which was absolutely engrossing and had me reading until about 4AM, and I then asked it to wrap it up within 15 minutes in a satisfactory ending etc, which it did a brilliant job of (if I wasn't concerned about losing my session I'd have asked it to continue for several more weeks). A year ago it was absolutely useless at comedy, now it routinely has me in fits of laughter.

I doubt many school kids are writing essays any more, just delegating those tasks and maybe making a few tweaks to make it look less obvious. Heck, I bet plenty of CEOs are doing similar, and the AI is probably doing a better job than they would have in some cases.

...seems odd it can't convert knots to km/hr.

 

[qldfrog]

AI has become a very useful analysis tool, but I have had plenty of experiences similar to this, where it completely fails at the most basic of calculations and logic exercises.

Sometimes the failures seem like they could only occur if the model was deliberately designed to fail, though I can't see any benefit to that other than perhaps to have these bugs ironed out in the paid versions. Sometimes they are incredibly creative and insightful, with brilliant reasoning skills etc, and then a minute later will tell you that ducks have gills and breath underwater because they live in lakes. I've had them tell me things like the moon orbits only a few centimeters above the Earth's surface (when asking it to calculate planetary rotation decay rates) and other completely absurd things. It told me a Greenland Shark would be an ideal pet for me (it would require a tank larger than the world's largest swimming pool which would have to be constantly refrigerated, among other challenges).

But judging it today is a bit like saying the internal combustion engine and light bulb will never replace the horse and candle based on the flaws you see in the first years proceeding their invention. I had one write an elaborate science fiction novel with incredibly vivid world building which was absolutely engrossing and had me reading until about 4AM, and I then asked it to wrap it up within 15 minutes in a satisfactory ending etc, which it did a brilliant job of (if I wasn't concerned about losing my session I'd have asked it to continue for several more weeks). A year ago it was absolutely useless at comedy, now it routinely has me in fits of laughter.

I doubt many school kids are writing essays any more, just delegating those tasks and maybe making a few tweaks to make it look less obvious. Heck, I bet plenty of CEOs are doing similar, and the AI is probably doing a better job than they would have in some cases.

...seems odd it can't convert knots to km/hr.

Agree with all your points, but i think it is us a VERY dangerous tool if taken as source of truth without critical analysis and basic knowledge in the subject.
Not that humans can not do even worse: look at Covid, Ukraine or even voting choices
Specifically IT programming: in the coming years already starting,no human will be trained anymore, the older more experienced programmers slowly weaned of their knowledge before disappearing and in 10y time,no one will be even capable to review AI generated code,the next level following the same path will be diagnostic..health, breakdown analysis and design for roads, bridges rockets missiles.
However rare the errors will be and statistically rightly less than human, i somehow prefer to be responsible if slamming in a wall at 100,km/h than an AI mistake :
even if statistically the error is less than the average human , especially considering that the average crashing human is most often a P plate on a thrill speed race, a drunk or drug high dimwit or a sleep deprived mum trying to quieten a kid in the backseat..you get the idea..
But statistically human loses..reverse Darwinism ahead
And this is nothing considering AI managing a nuclear plant, delivering remote GP diagnostics etc.
Errors will be made, and initially acknowledged then errors will not be detected anymore.pronlem solved.
It is a dangerous tool w/o care of use, but magnificent also, stunning.
Now, the last week has seen a cooling of the AI boom in the market.
Nuclear and Uranium stocks down ,AC providers, even warehouses and sheds glorified up to 6 weeks ago as data centers..
AI is not dead, it will carry on stronger and stronger and a few providers will become the Google or Siri of the past decades, potentially the same ,as they have the money to buy the emerging leaders ..
But i doubt it will lead to many profitable side businesses, or lead to an actual boom in energy generation.
We will see further concentration, how many people working for Google now in Australia , directly or not..next to nothing.
Just trying to convert ice to EV is a much bigger issue in term of generation when the West is destroying its own capacity via co2 suicidal paranoia
Market wise, stick to the leaders, do not bypass India/China and you should have AI related market covered IMHO

 

[divs4ever]

AI has become a very useful analysis tool, but I have had plenty of experiences similar to this, where it completely fails at the most basic of calculations and logic exercises.

Sometimes the failures seem like they could only occur if the model was deliberately designed to fail, though I can't see any benefit to that other than perhaps to have these bugs ironed out in the paid versions. Sometimes they are incredibly creative and insightful, with brilliant reasoning skills etc, and then a minute later will tell you that ducks have gills and breath underwater because they live in lakes. I've had them tell me things like the moon orbits only a few centimeters above the Earth's surface (when asking it to calculate planetary rotation decay rates) and other completely absurd things. It told me a Greenland Shark would be an ideal pet for me (it would require a tank larger than the world's largest swimming pool which would have to be constantly refrigerated, among other challenges).

But judging it today is a bit like saying the internal combustion engine and light bulb will never replace the horse and candle based on the flaws you see in the first years proceeding their invention. I had one write an elaborate science fiction novel with incredibly vivid world building which was absolutely engrossing and had me reading until about 4AM, and I then asked it to wrap it up within 15 minutes in a satisfactory ending etc, which it did a brilliant job of (if I wasn't concerned about losing my session I'd have asked it to continue for several more weeks). A year ago it was absolutely useless at comedy, now it routinely has me in fits of laughter.

I doubt many school kids are writing essays any more, just delegating those tasks and maybe making a few tweaks to make it look less obvious. Heck, I bet plenty of CEOs are doing similar, and the AI is probably doing a better job than they would have in some cases.

...seems odd it can't convert knots to km/hr.

not designed 'to fail ' , but done quickly within due testing and observation , much like recent vaccines

designed shoddily and possibly fed data that included errors and deceptions ( now one might suggest this was done to create self-learning ) , but some humans are indoctrinated to blindly accept automated answers and that would create a doom-loop