In 1980, as an undergraduate with ADHD and a typewriter, I was not capable of writing something as long as a dissertation. In 1987, using the lab’s shared Apple computer and word processor, I wrote a dissertation. Cut, Copy, and Paste makes it possible for me to write thing. Spellcheck too.
It is worth adding that - as Cory Doctorow pointed out in a recent essay for Locus Magazine - and I paraphrase: "this sh!t is expensive!" The massive resources required to run the liquid-cooled server farms is enormous. This is not financially viable as a profit model if all we have to show for it in the long run are "low-stakes, high cost" gains (e.g upgrading from ps 3 to ps 5). Copyright regulation is a barrier (the ghost of Napster-past) and so are hype-driven investment subsidies that makes the service seem more profitable than it actually is (the ghost of Uber-present)
I just recently had to talk down my supervisor about so called AI after the CIO laid down an edict that our organization is going to 'leverage AI to save money'. An insane prospect when ChatGPT is almost literally burning money. I had a conversation with the man, and it's obvious he's fallen for the hype. Like he thinks he can replace programmers with decades of institutional knowledge who work on obscure heavily customized software with a chatbot.
Insane.
My advice to my manager (and everyone else) is as much as possible we (people who do the actual work) punt the can down the road until the hype collapses.
I especially like your point about the already existing and deeply problematic uses of machine learning. I immediately thought of the use of predictive algorithms like the ones they talk about over at AI Snake Oil and The Markup. Using upgraded machine learning to guess who will do well in school or stay out of jail will suck for the humans being judged and carry significant social costs. Worth putting some focus there rather than on the Shoggoths.
ChatGPT looks impressive, in part, because it is a radically improved version of Google, which has got steadily worse for 25 years. If ChatGPT had been the result of gradual progress over that quarter-century, no one would think twice about it.
ChatGPT is a great little tool for revealing data that has already been stolen. When working with ChatGPT, one can't help realizing the vast quantity of data about us and our world that the tool has incorporated. ChatGPT and other machine learning tools need more data each year in order to stay relevant. Hopefully, this realization will cause people to finally take a serious look at how their data is being collected and used.
Very helpful. Techno-optimism’s inextricable link to investment-optimism takes rationality and sober predictions out of the equation. Like so many things - hype first, reality later.
Agree with (1) as not agreeing flies in the face of obvious economic necessities. (2) probably not because the 'big data analytics' AI and 'Generative AI' are very different beasts with generally disjunct, and sometimes even opposing use cases.
In 1980, as an undergraduate with ADHD and a typewriter, I was not capable of writing something as long as a dissertation. In 1987, using the lab’s shared Apple computer and word processor, I wrote a dissertation. Cut, Copy, and Paste makes it possible for me to write thing. Spellcheck too.
It is worth adding that - as Cory Doctorow pointed out in a recent essay for Locus Magazine - and I paraphrase: "this sh!t is expensive!" The massive resources required to run the liquid-cooled server farms is enormous. This is not financially viable as a profit model if all we have to show for it in the long run are "low-stakes, high cost" gains (e.g upgrading from ps 3 to ps 5). Copyright regulation is a barrier (the ghost of Napster-past) and so are hype-driven investment subsidies that makes the service seem more profitable than it actually is (the ghost of Uber-present)
I just recently had to talk down my supervisor about so called AI after the CIO laid down an edict that our organization is going to 'leverage AI to save money'. An insane prospect when ChatGPT is almost literally burning money. I had a conversation with the man, and it's obvious he's fallen for the hype. Like he thinks he can replace programmers with decades of institutional knowledge who work on obscure heavily customized software with a chatbot.
Insane.
My advice to my manager (and everyone else) is as much as possible we (people who do the actual work) punt the can down the road until the hype collapses.
I especially like your point about the already existing and deeply problematic uses of machine learning. I immediately thought of the use of predictive algorithms like the ones they talk about over at AI Snake Oil and The Markup. Using upgraded machine learning to guess who will do well in school or stay out of jail will suck for the humans being judged and carry significant social costs. Worth putting some focus there rather than on the Shoggoths.
ChatGPT looks impressive, in part, because it is a radically improved version of Google, which has got steadily worse for 25 years. If ChatGPT had been the result of gradual progress over that quarter-century, no one would think twice about it.
ChatGPT is a great little tool for revealing data that has already been stolen. When working with ChatGPT, one can't help realizing the vast quantity of data about us and our world that the tool has incorporated. ChatGPT and other machine learning tools need more data each year in order to stay relevant. Hopefully, this realization will cause people to finally take a serious look at how their data is being collected and used.
Very helpful. Techno-optimism’s inextricable link to investment-optimism takes rationality and sober predictions out of the equation. Like so many things - hype first, reality later.
Agree with (1) as not agreeing flies in the face of obvious economic necessities. (2) probably not because the 'big data analytics' AI and 'Generative AI' are very different beasts with generally disjunct, and sometimes even opposing use cases.