I've seen Colossus several times and you've got the ending wrong. Colossus and Guardian do not exterminate humanity. They control it - there is a scene where some humans are killed because of the conspiracy to disarm the computers of their nuclear capability. But the ending doesn't imply that Colossus is going exterminate humanity. It is going to coexist on its terms.
I forwarded one of your links to the cognitive dissidents channel which has three hosts one of which is Havjre from Geopolitics & Empire. Hopefully you got a few subscribers from them. I think they would appreciate your mostly objective views.
Thanks, Nick. I anticipate doing an interview with Havjre soon on Geopolitics & Empire. We did one once several months ago. He's a fun guy to talk with.
IDK. I gave a moderately complicated math problem to Grok, ChatGPT and DeepSeek. It involved trying to compute the approximate balance on a loan that my son has. I provided the original amount, the date, and the interest rate, along with a payment history, and the exact date that he began underpaying the loan by $200 per month.
I wanted to know 3 things. The approximate current balance. The date the loan would be paid off if he continued to underpay by $200 per month. And the amount per month he would need to pay, over and above the original payment amount to pay off the loan by the due date in 2028.
I had already done the work manually with a calculator so I had a reasonable idea but was looking for more exact numbers.
The results wet not what I expected.
Grok told me the loan had already been paid off and I should stop making payments and seek a refund.
ChapGPT told me that beginning at the 5 year point only interest had been paid from that point and the balance was 40k higher than what I had figured.
DeepSeek got within 10% of what my manual calculation showed in its final summary. But, it crafted multiple scenarios and speculations and possible outcomes that ran for pages and pages before it finally reached a conclusion…that at least had some semblance of what is likely reality.
Grok was 100% wrong and had someone followed that train trouble woukd certainly follow.
ChapGPT was 100% wrong and had someone followed thst train they could have erroneously decided to just walk away from the loan, as it was so far underwater that it made no sense to keep paying.
ChapGPT got it wrong, and presented far too many ‘other’ possibiities by inserting reams of information that might have been, but that I did not provide, but in the final summary got close enough to the correct result that one could take the knowledge gained to the lender with some minimal degree of confidence.
AI is certainly disruptive already. And as you say it will become more disruptive. But if you think about it, it has been for decades…replacing a human answering the help line with a computer started 20 years ago.
I remain unconvinced that it is actually anywhere near as advanced as its creators want to pretend. It was unable to provide correct information on a moderately complex, but not difficult, math problem. And 3 different platforms provided 3 different answers.
Yeah I guess that is disruptive. To my brain. LOL.
With an example of something that simple (for a computer), it does kind of sound like these things are lazy. That does illustrate part of the problem with relying on them, though. They are full of mistakes and outright lies. They seem to just make stuff up.
So, maybe there is another year left of job security for me as a writer before Grok can do better than I can. (Shouldn't have written that. Now Grok will sniff it out and then work on being a better version of me just out of spite. Let's hope it, at least, has more time than I do to go back and clean up all ITS typos.)
I've seen Colossus several times and you've got the ending wrong. Colossus and Guardian do not exterminate humanity. They control it - there is a scene where some humans are killed because of the conspiracy to disarm the computers of their nuclear capability. But the ending doesn't imply that Colossus is going exterminate humanity. It is going to coexist on its terms.
I forwarded one of your links to the cognitive dissidents channel which has three hosts one of which is Havjre from Geopolitics & Empire. Hopefully you got a few subscribers from them. I think they would appreciate your mostly objective views.
Thanks, Nick. I anticipate doing an interview with Havjre soon on Geopolitics & Empire. We did one once several months ago. He's a fun guy to talk with.
IDK. I gave a moderately complicated math problem to Grok, ChatGPT and DeepSeek. It involved trying to compute the approximate balance on a loan that my son has. I provided the original amount, the date, and the interest rate, along with a payment history, and the exact date that he began underpaying the loan by $200 per month.
I wanted to know 3 things. The approximate current balance. The date the loan would be paid off if he continued to underpay by $200 per month. And the amount per month he would need to pay, over and above the original payment amount to pay off the loan by the due date in 2028.
I had already done the work manually with a calculator so I had a reasonable idea but was looking for more exact numbers.
The results wet not what I expected.
Grok told me the loan had already been paid off and I should stop making payments and seek a refund.
ChapGPT told me that beginning at the 5 year point only interest had been paid from that point and the balance was 40k higher than what I had figured.
DeepSeek got within 10% of what my manual calculation showed in its final summary. But, it crafted multiple scenarios and speculations and possible outcomes that ran for pages and pages before it finally reached a conclusion…that at least had some semblance of what is likely reality.
Grok was 100% wrong and had someone followed that train trouble woukd certainly follow.
ChapGPT was 100% wrong and had someone followed thst train they could have erroneously decided to just walk away from the loan, as it was so far underwater that it made no sense to keep paying.
ChapGPT got it wrong, and presented far too many ‘other’ possibiities by inserting reams of information that might have been, but that I did not provide, but in the final summary got close enough to the correct result that one could take the knowledge gained to the lender with some minimal degree of confidence.
AI is certainly disruptive already. And as you say it will become more disruptive. But if you think about it, it has been for decades…replacing a human answering the help line with a computer started 20 years ago.
I remain unconvinced that it is actually anywhere near as advanced as its creators want to pretend. It was unable to provide correct information on a moderately complex, but not difficult, math problem. And 3 different platforms provided 3 different answers.
Yeah I guess that is disruptive. To my brain. LOL.
With an example of something that simple (for a computer), it does kind of sound like these things are lazy. That does illustrate part of the problem with relying on them, though. They are full of mistakes and outright lies. They seem to just make stuff up.
So, maybe there is another year left of job security for me as a writer before Grok can do better than I can. (Shouldn't have written that. Now Grok will sniff it out and then work on being a better version of me just out of spite. Let's hope it, at least, has more time than I do to go back and clean up all ITS typos.)