Sobering and compelling essay--unless, of course, Chatbot had been directed to compose an essay casting doubt on the accuracy of empirical assertions by Chatbot, so that our revered blogger could settle back, entrust all future essay composition to the sociopathic mind and incorporeal hand of Chatbot, while evading authorship suspicion on the grounds that 'Hey, he warned us about Chatbot veracity--he would never use it himself.' BTW, this note was not composed by Chatbot. It was composed by Felicity van Crevalt-Pflessenheim, DPhil, Oxon. Honest.
I'd like to suggest that the future of AI will be multiple Agents - a Doctor will use their expertise to select the Agent that fits best with the current diagnosis. The Agent will be like the tests Doctors use to confirm a diagnosis - which many times are overkill.
Asking the Congress to regulate AI is like asking them to set tolerances for a blacksmith.
Trust is the key element of any relationship. ChatGPT is the belle of the ball right now. But what happens as the fickle public learns it can't be trusted?
I've been using ChatGPT to write some first drafts, I'm beginning to notice that it is regurgitated pablum. It sounds good, but real detail is lacking.
I listened to some of the Gatsby. Something is lacking. I had a MIDI program years ago - it had a "Humanize" function - which introduced random slight errors to make the music sound like a human was playing it.
Thanks for this piece. Fabricating multiple source documents to verify a person's background is what intelligence agencies do to give deep cover to spies. The newfound ease of generating many mutually supporting, prima facie plausible frauds is a very interesting—and disturbing—possibility that will only accelerate loss of trust. The fact that Sam Altman invested his fortune into a startup to create reliable a global digital ID for humans to prove they are humans, a step toward ending internet anonymity, is very telling.
Toiling in Oceania's Ministry of Truth, 1984's protagonist Winston Smith devoted himself to re-writing history. Anyone who had been physically erased at the Ministry of Love also had to be erased from memory, so Winston revised books and newspapers and class lists and employment records and whatever else it took to disappear their existence. Sometimes he had to do the opposite: invent somebody to play a part in an imaginary story, such as the tale of a heroic sailor who died fighting East Asia, and whose story is necessary to inspire the populace. Smith could create a well-documented life story for such people.
Our future dystopia will have no need of Winston Smith; ChatGPT will sit in his cubicle, well supplied with reliable power and good internet connections. And of course, ChatGPT won't have to head to the employee auditorium after lunch for the Two-Minute Hate; ol' Chat will just ZOOM in.
I am so glad I read this. I used a chat AI recently and was provided with prevalence data that was incredibly valuable. I asked for the source and was given one that seemed legit. So far, I have been unable to find it! When I quizzed the Chatbot, it insisted it was there. A PMID was provided that PubMed cannot find. Very strange.
Sobering and compelling essay--unless, of course, Chatbot had been directed to compose an essay casting doubt on the accuracy of empirical assertions by Chatbot, so that our revered blogger could settle back, entrust all future essay composition to the sociopathic mind and incorporeal hand of Chatbot, while evading authorship suspicion on the grounds that 'Hey, he warned us about Chatbot veracity--he would never use it himself.' BTW, this note was not composed by Chatbot. It was composed by Felicity van Crevalt-Pflessenheim, DPhil, Oxon. Honest.
I'd like to suggest that the future of AI will be multiple Agents - a Doctor will use their expertise to select the Agent that fits best with the current diagnosis. The Agent will be like the tests Doctors use to confirm a diagnosis - which many times are overkill.
Asking the Congress to regulate AI is like asking them to set tolerances for a blacksmith.
Trust is the key element of any relationship. ChatGPT is the belle of the ball right now. But what happens as the fickle public learns it can't be trusted?
I've been using ChatGPT to write some first drafts, I'm beginning to notice that it is regurgitated pablum. It sounds good, but real detail is lacking.
I listened to some of the Gatsby. Something is lacking. I had a MIDI program years ago - it had a "Humanize" function - which introduced random slight errors to make the music sound like a human was playing it.
Thanks for this piece. Fabricating multiple source documents to verify a person's background is what intelligence agencies do to give deep cover to spies. The newfound ease of generating many mutually supporting, prima facie plausible frauds is a very interesting—and disturbing—possibility that will only accelerate loss of trust. The fact that Sam Altman invested his fortune into a startup to create reliable a global digital ID for humans to prove they are humans, a step toward ending internet anonymity, is very telling.
Toiling in Oceania's Ministry of Truth, 1984's protagonist Winston Smith devoted himself to re-writing history. Anyone who had been physically erased at the Ministry of Love also had to be erased from memory, so Winston revised books and newspapers and class lists and employment records and whatever else it took to disappear their existence. Sometimes he had to do the opposite: invent somebody to play a part in an imaginary story, such as the tale of a heroic sailor who died fighting East Asia, and whose story is necessary to inspire the populace. Smith could create a well-documented life story for such people.
Our future dystopia will have no need of Winston Smith; ChatGPT will sit in his cubicle, well supplied with reliable power and good internet connections. And of course, ChatGPT won't have to head to the employee auditorium after lunch for the Two-Minute Hate; ol' Chat will just ZOOM in.
I am so glad I read this. I used a chat AI recently and was provided with prevalence data that was incredibly valuable. I asked for the source and was given one that seemed legit. So far, I have been unable to find it! When I quizzed the Chatbot, it insisted it was there. A PMID was provided that PubMed cannot find. Very strange.