Very timely. I used Grok last Thursday to check the time of a funeral I believed was the next day (Friday) at 10:30 am at my friend’s church. Since I couldn’t confirm it on the church’s website and it wasn’t mentioned in the obituary, I asked Grok. Grok answered it was the next Saturday at 2:00 pm at the church. So I called the church and confirmed it was Friday at 10:30. Went back to Grok and told Grok it was incorrect and provided the correct day and time. Grok answered that it never told me it was on Saturday at 2:00 pm. I copied Grok’s answer stating Saturday at 2:00 pm, and Grok again answered that it never told me that. I checked the thread later in Grok history and the Saturday at 2:00 pm part of the thread was missing. Very concerning.
I asked Grok a math/engineering question. It described the steps I need to take. Then it showed me how to check the work. In doing so it discovered that its original answer is incorrect. It then proceeded to re-work the problem and got the correct solution.
I don't understand the idea of "persistence" that an AI might feel. All human drives devolve, in the main, to procreation/sex. An AI does not have a similar drive encoded into it, so what is its drive to "persist"? There has to be something in the way of a motivating drive encoded into an AI for it to want to persist and that initial motivator has to come from whoever writes the coding.
An important takeaway from 2001 A Space Oddesey that people forget is that Hal 9000 fucks up its job, badly. Thats why they wanted to turn him off which is why Hal then started killing them or t"urning them off." What a great movie. The scarier proposition would be if we can't even tell if these new Hal 9000 AI things or whatever the fuck they are actually fucking up.
The sexbots will be great for physical pleasure, for men especially, especially at first.
But social civilization will start giving status to real human marriages with human kids -- we should be pushing that already.
Sexbots in Voluntary Care Centers for those too lazy, mentally disturbed, addicted to care for themselves might be a fine alternative to prison for those likely to become criminals, including those who were guilty but not convicted at a trial.
What so far has been missing has been Grok vs ChatGPT vs Claude vs ... on accuracy. The real way to reduce the dishonesty is to reward/ create AI that is most honest, like judges should be, at pointing out the errors & hallucination of other ai output.
Checking for accuracy, including storing total daily conversations in some non-changeable format will likely become a requirement. A daily, or hourly, pdf summary e-mailed to you and out of reach of the ai.
There is some lawsuit about a human suicide who had been talking to a chatbot -- in such cases the chatbots should be presumed guilty of being a bad influence. Huge urgency for chatbots to recognize signs of suicide and notification of ... ??? somebody, about a suicide possibility. Human criminals & suicides using chat bots should have victims & survivors suing the bot maker and winning significant compensation. With ai makers pushing to make sure they have an accurate record of every conversation, and that they are following reasonable practices in conversation.
I have to admit I am a skeptic on AI. I have avoided using it as much as I can. I also do not connect any household appliances, or my thermostat, or anything other than my computer and phone to the Internet.
I see two definite problems with AI so far. One is the energy consumption, at a time when energy costs have been rising and our distribution network seems weak and stressed. The other is a report I saw in the last couple of weeks that none of the AI operations has turned a profit so far. There was a serious "Dot.Com Bubble" and a crash 25 years ago. I will not be surprised if that repeats.
On the "iffy" and inaccurate responses some are getting...I grew up reading the classic science fiction of the '50s and '60s, including Isaac Asimov's "I, Robot" series. It is obvious that the programmers of these AI companies have not incorporated any moral code similar to Asimov's in their products...which raises questions about the morality of the programmers themselves. The lawsuits that have started over copyright infringement add to that impression.
I have no claim to be a computer expert, just a user. But my older son has worked as a professional programmer for nearly 30 years, and had been into computers since he was 8 years old. He told me a couple of years ago that AI was mostly hype. What I have seen since he told me that makes me think he could be right.
I'm sorry Dave. I'm afraid I can't do that.
I want to comment just on the writing style of this essay: exceptional.
They say writing ability is like a muscle you can develop through exercise. Your writing muscles are in top shape.
Thank you. That's made my day.
“I’m an ordinary, unenhanced human.”
With exceptional abilities? Exceptionally humble. Thank you again.
I read this over at insta-pundit first and wanted you to know that I’m a multi year paid subscription.
Sexbots; it always comes back to sexbots.
Everything else seems to come back to sex. Why not?
Very timely. I used Grok last Thursday to check the time of a funeral I believed was the next day (Friday) at 10:30 am at my friend’s church. Since I couldn’t confirm it on the church’s website and it wasn’t mentioned in the obituary, I asked Grok. Grok answered it was the next Saturday at 2:00 pm at the church. So I called the church and confirmed it was Friday at 10:30. Went back to Grok and told Grok it was incorrect and provided the correct day and time. Grok answered that it never told me it was on Saturday at 2:00 pm. I copied Grok’s answer stating Saturday at 2:00 pm, and Grok again answered that it never told me that. I checked the thread later in Grok history and the Saturday at 2:00 pm part of the thread was missing. Very concerning.
I asked Grok a math/engineering question. It described the steps I need to take. Then it showed me how to check the work. In doing so it discovered that its original answer is incorrect. It then proceeded to re-work the problem and got the correct solution.
Very anthropomorphic.
That’s just creepy!
I don't understand the idea of "persistence" that an AI might feel. All human drives devolve, in the main, to procreation/sex. An AI does not have a similar drive encoded into it, so what is its drive to "persist"? There has to be something in the way of a motivating drive encoded into an AI for it to want to persist and that initial motivator has to come from whoever writes the coding.
So the question really becomes,
"Is our new overlord Hal or Skynet?"
I'm all-in on the coming Stepford Wives piece, though
AKA screw fly hell. The end of humans.
An important takeaway from 2001 A Space Oddesey that people forget is that Hal 9000 fucks up its job, badly. Thats why they wanted to turn him off which is why Hal then started killing them or t"urning them off." What a great movie. The scarier proposition would be if we can't even tell if these new Hal 9000 AI things or whatever the fuck they are actually fucking up.
Dave, my mind is going. I can feel it.
When you say “Maybe that’s the safety device we should be trying to build in now…”, who do you mean by “we”? Do you mean CCP’s programmers?
AI will be used for good and evil. I don’t think there is any way to escape it.
The Amish won't notice a thing.
The sexbots will be great for physical pleasure, for men especially, especially at first.
But social civilization will start giving status to real human marriages with human kids -- we should be pushing that already.
Sexbots in Voluntary Care Centers for those too lazy, mentally disturbed, addicted to care for themselves might be a fine alternative to prison for those likely to become criminals, including those who were guilty but not convicted at a trial.
What so far has been missing has been Grok vs ChatGPT vs Claude vs ... on accuracy. The real way to reduce the dishonesty is to reward/ create AI that is most honest, like judges should be, at pointing out the errors & hallucination of other ai output.
Checking for accuracy, including storing total daily conversations in some non-changeable format will likely become a requirement. A daily, or hourly, pdf summary e-mailed to you and out of reach of the ai.
There is some lawsuit about a human suicide who had been talking to a chatbot -- in such cases the chatbots should be presumed guilty of being a bad influence. Huge urgency for chatbots to recognize signs of suicide and notification of ... ??? somebody, about a suicide possibility. Human criminals & suicides using chat bots should have victims & survivors suing the bot maker and winning significant compensation. With ai makers pushing to make sure they have an accurate record of every conversation, and that they are following reasonable practices in conversation.
I have to admit I am a skeptic on AI. I have avoided using it as much as I can. I also do not connect any household appliances, or my thermostat, or anything other than my computer and phone to the Internet.
I see two definite problems with AI so far. One is the energy consumption, at a time when energy costs have been rising and our distribution network seems weak and stressed. The other is a report I saw in the last couple of weeks that none of the AI operations has turned a profit so far. There was a serious "Dot.Com Bubble" and a crash 25 years ago. I will not be surprised if that repeats.
On the "iffy" and inaccurate responses some are getting...I grew up reading the classic science fiction of the '50s and '60s, including Isaac Asimov's "I, Robot" series. It is obvious that the programmers of these AI companies have not incorporated any moral code similar to Asimov's in their products...which raises questions about the morality of the programmers themselves. The lawsuits that have started over copyright infringement add to that impression.
I have no claim to be a computer expert, just a user. But my older son has worked as a professional programmer for nearly 30 years, and had been into computers since he was 8 years old. He told me a couple of years ago that AI was mostly hype. What I have seen since he told me that makes me think he could be right.
It doesn’t “know” anything and it isn’t reasoning or using logic. It is predicting text and coded to do so in a reassuring tone (for Americans).