Really interesting. I'm more on the side of you get some really interesting outlines, a seriously good bit of help in research, but I wouldn't rely on ChatGPT for much more. I grew up in journalism with the understanding that the most important thing about information was the source, and I've not had much luck getting sources out of ChatGPT or Bard. I know they are large models with a limited date range but, well, as Chat GPT pointed out, there's a lot of room for bias. It's also interesting to run the same question in CGPT and Bard and see the differences. I do see great potential with these tools and eventually will learn how to do more than ask questions. Thanks for the articles on your experiences.
It took me a while to get the hang of how to prompt it, but I have been having pretty good luck with it. I plan to use it on my next book. I have not had much luck with Bard or Bing but I didn't work that hard with them.
You must not have read Mr. Hinton very closely. He is one of the most well informed experts and he is decidedly pessimistic. His interview May 4 with Computerworld rattled me, especially an anecdote about his recent interaction with ChatGTP, in which he found reasoning at an IQ level of 80-90 and potential for rapid development. He said that ML cannot pull levers but it will be able to persuade humans to pull levers for better or worse. He cited the example that social media has already enabled sketchy politicians to persuade large groups to do harmful things (like break into the US Capitol). Said politicians were able to do this from afar and probably avoid accountability. Social media is quite crude compared to the prospects of ML, and our societal/political institutions are not equipped to control ML.
We might not want to bomb the data centers yet, but should consider installing self-destruct or disarming systems systems like the large rockets and military hardware. The big tech companies will not even attempt to control or "pause" ML. Governments (looking at you, China and Russia) should start NOW on interventions in the real world to control the cyber world.
Really interesting. I'm more on the side of you get some really interesting outlines, a seriously good bit of help in research, but I wouldn't rely on ChatGPT for much more. I grew up in journalism with the understanding that the most important thing about information was the source, and I've not had much luck getting sources out of ChatGPT or Bard. I know they are large models with a limited date range but, well, as Chat GPT pointed out, there's a lot of room for bias. It's also interesting to run the same question in CGPT and Bard and see the differences. I do see great potential with these tools and eventually will learn how to do more than ask questions. Thanks for the articles on your experiences.
It took me a while to get the hang of how to prompt it, but I have been having pretty good luck with it. I plan to use it on my next book. I have not had much luck with Bard or Bing but I didn't work that hard with them.
You must not have read Mr. Hinton very closely. He is one of the most well informed experts and he is decidedly pessimistic. His interview May 4 with Computerworld rattled me, especially an anecdote about his recent interaction with ChatGTP, in which he found reasoning at an IQ level of 80-90 and potential for rapid development. He said that ML cannot pull levers but it will be able to persuade humans to pull levers for better or worse. He cited the example that social media has already enabled sketchy politicians to persuade large groups to do harmful things (like break into the US Capitol). Said politicians were able to do this from afar and probably avoid accountability. Social media is quite crude compared to the prospects of ML, and our societal/political institutions are not equipped to control ML.
We might not want to bomb the data centers yet, but should consider installing self-destruct or disarming systems systems like the large rockets and military hardware. The big tech companies will not even attempt to control or "pause" ML. Governments (looking at you, China and Russia) should start NOW on interventions in the real world to control the cyber world.