How could Artificial Intelligence become dangerous?

[Jstpssng]

News release today about hackers using AI (Claude) to probe and identify corporations with weak cyber security, then used Claude again to hack and steal data, and then used AI again to review the stolen data for high profit information...either by holding the data ransom, or using the data for gain.

We're standing on a cliff edge...and it's crumbling...

Some have been beating that drum for years, but security costs money.

 

[MinnesotaEric]

Meanwhile, why Files hits it out of the park with a short format video.

899K views · 35K reactions | If AI Didn't Scare You Before...It Will After This Video! | The Why Files

If AI Didn't Scare You Before...It Will After This Video!.

www.facebook.com

 

[paulsharvey]

I saw a video recently that Albania is having an AI program review contractors, consultants/RFPs, ect and make the selection from the list, because it is so Wildly corrupt over there, that the AI can't possibly be worse than a human, and likely Much better.

 

[California]

I saw a video recently that Albania is having an AI program review contractors, consultants/RFPs, ect and make the selection from the list, because it is so Wildly corrupt over there, that the AI can't possibly be worse than a human, and likely Much better.

And far faster. Contractor proposal evaluations was one of my roles, in a Large Public Agency.

Conflicting criteria, confusing declarations from firms who had done good work previously, late requests for evaluations from the project managers who would be using the contractor, various new surprise factors after completing an evaluation - Your Tax Dollars At Work but not very efficiently.

If Albania has the added factor of wild corruption, I can see how this would be a huge improvement. The human evaluator's role would be slimmed down to just evaluating if the AI did a good job this time. And of course reporting if bias was cooked into the AI's selection criteria.

Now a longwinded rant -

Nobody's going to believe this but maybe some will find it amusing: About the time I determined I had saved enough for comfortable retirement, the manager supervising pre-award contractor evaluations was on work-from-home maternity leave. Her student assistant, fresh out of high school, was tasked with handing out the assignment docs. Several times a contract manager said they had gone ahead with a contract award without our input because they never heard from us.

I investigated and found the project assignment doc was still in the hands of that student assistant, never forwarded to me. But in the eyes of the project manager who would employ that contractor I appeared to be responsible for the delay because the work had been assigned to me - just nobody told me.

That, plus then a new supervisor who cancelled all vacations to try to make himself look unusually efficient, was the last straw. I told him I was starting my vacation on July 1 and would be using accumulated leave through Thanksgiving. So he couldn't backfill my position before then. This killed his 'efficiency' program. The guy lost his trial promotion for that and some other equally stupid measures.

Your Tax Dollars At Work!

 

[kenmbz]

More ways AI can kill us:
AI data centers require a large and rapidly increasing amount of water for cooling, with estimates suggesting a 100-megawatt facility could use 2 million liters (530,000 gallons) of water daily, equivalent to the annual water usage of thousands of households. Global water consumption by data centers is already in the hundreds of billions of liters annually and is projected to rise substantially by 2030, creating significant water scarcity concerns in some regions

 

[California]

More ways AI can kill us:
AI data centers require a large and rapidly increasing amount of water for cooling, ...

I wonder how that quantity of water compares with nuclear generating plants, with their huge cooling towers. New plants will be needed for this additional load.

 

[kenmbz]

I wonder how that quantity of water compares with nuclear generating plants, with their huge cooling towers. New plants will be needed for this additional load.

Additive, so makes the end state go faster.

 

[riptides]

More ways AI can kill us:
AI data centers require a large and rapidly increasing amount of water for cooling, with estimates suggesting a 100-megawatt facility could use 2 million liters (530,000 gallons) of water daily, equivalent to the annual water usage of thousands of households. Global water consumption by data centers is already in the hundreds of billions of liters annually and is projected to rise substantially by 2030, creating significant water scarcity concerns in some regions

Neighboring county produced a detailed study on water use. Down to the aquifer levels. A quick peek showed half the county in the red.

They are considered the data center capital of the world.

Still building them out, and spreading to neighboring counties.

Transmission infrastructure build out underway, in the billions of dollars for one new segment to feed all the above.

 

[California]

Bah. Google's AI got me good.

I have a lot of Ryobi gear. Project Farm's latest review rated the Ryobi 6 inch pruning chainsaw P25013 third among a dozen tested. Buying anything but Ryobi would cost me an additional $100 to begin a new incompatible battery series. So I found a discounted new-in-box Ryobi saw on ebay, just listed, reputable seller. Researched its model# on Google. "Ryobi PCLCW01B vs P25013".

The PCLCW01B is the newer Ryobi ONE+ HP 18V Brushless 6 in. Compact Pruning Mini Chainsaw, offering brushless motor technology for better performance, while the P25013 is the older, non-brushless model. Choose the PCLCW01B for superior cutting efficiency and durability ...​

Project Farm had tested the P25013. So I bought the new-in-box Ebay PCLCW01B that Google said was better. For $80, $20 less than retail.

Then I went back to that same Google description and clicked 'Dive Deeper'. Bah. Now in this later paragraph, same report, Google AI reports just the opposite:

The key difference between the Ryobi PCLCW01B and the discontinued P25013 is the motor.The PCLCW01B has a standard brushed motor, while the P25013 has a brushles motor with HP (High Performance) technology. This makes the P25013 more powerful.​

Those two paragraphs in the same AI analysis are 180 degrees opposite.

 

[drssg]

Bah. Google's AI got me good.

I have a lot of Ryobi gear. Project Farm's latest review rated the Ryobi 6 inch pruning chainsaw P25013 third among a dozen tested. Buying anything but Ryobi would cost me an additional $100 to begin a new incompatible battery series. So I found a discounted new-in-box Ryobi saw on ebay, just listed, reputable seller. Researched its model# on Google. "Ryobi PCLCW01B vs P25013".

The PCLCW01B is the newer Ryobi ONE+ HP 18V Brushless 6 in. Compact Pruning Mini Chainsaw, offering brushless motor technology for better performance, while the P25013 is the older, non-brushless model. Choose the PCLCW01B for superior cutting efficiency and durability ...​

Project Farm had tested the P25013. So I bought the new-in-box Ebay PCLCW01B that Google said was better. For $80, $20 less than retail.

Then I went back to that same Google description and clicked 'Dive Deeper'. Bah. Now in this later paragraph, same report, Google AI reports just the opposite:

The key difference between the Ryobi PCLCW01B and the discontinued P25013 is the motor.The PCLCW01B has a standard brushed motor, while the P25013 has a brushles motor with HP (High Performance) technology. This makes the P25013 more powerful.​

Those two paragraphs in the same AI analysis are 180 degrees opposite.

And yet, I have to reconcile this type of experience with reports of companies replacing a large number of employees with AI. It will be interesting to see whether that turns out to be a horrible decision.

Computers are great at automation. They have speed and accuracy that no human can rival. When it comes to real intelligence, I'm still dubious.