Þe "Spartan Mirage" of Plutarch, &, perhaps, the German army 1866-1944 & the U.S. Marine Corps after Guadalcanal; American music industry revenues by product type; Ezra Klein & Robinson Meyer on...
Let me be the first to say that Deneen's proposed society is dishonestly portrayed and will never be realized -- right wing "populism" is always and everywhere a lie. He does not merit "debate".
"And at present it seems that the direction AI applications are going in is the opposite one – the use of specialised training data to make models which are better at giving correct answers in specific problem domains. It seems to me that these models, when they fail, will fail hard and will tend to double down on mistakes like a politician. "
Back in the days of GOFAI - hard-coded, rules-based, expert systems, it was found that these systems proved very brittle as they reached the limits of their rules as "edge cases". They would tend to fail... badly.
It seems to me that the author's argument is suggesting that these new LLM AIs trained on more selective data will increasingly seem to operate like this expert systems, except that rather than just failing, insist on proclaiming the wrong answer as correct. I think we have all experienced Chat-GPT insist that it was right about a fact, until confronted about its mistake.
How long before such AIs will respond like politicians: answer a different question to the one posed, deflect the question by changing the subject, either refuse to answer, or keep blathering to prevent the questioner from interceding (this last was the Boris Johnson MO).
Let me be the first to say that Deneen's proposed society is dishonestly portrayed and will never be realized -- right wing "populism" is always and everywhere a lie. He does not merit "debate".
GPT-LLM-ML
"And at present it seems that the direction AI applications are going in is the opposite one – the use of specialised training data to make models which are better at giving correct answers in specific problem domains. It seems to me that these models, when they fail, will fail hard and will tend to double down on mistakes like a politician. "
Back in the days of GOFAI - hard-coded, rules-based, expert systems, it was found that these systems proved very brittle as they reached the limits of their rules as "edge cases". They would tend to fail... badly.
It seems to me that the author's argument is suggesting that these new LLM AIs trained on more selective data will increasingly seem to operate like this expert systems, except that rather than just failing, insist on proclaiming the wrong answer as correct. I think we have all experienced Chat-GPT insist that it was right about a fact, until confronted about its mistake.
How long before such AIs will respond like politicians: answer a different question to the one posed, deflect the question by changing the subject, either refuse to answer, or keep blathering to prevent the questioner from interceding (this last was the Boris Johnson MO).