I smell a “welcome to the NFL” moment coming for this piece of OpenAI’s ambitions. But I also know that it is much too early and information is still much too thin for speculation to be worthwhile...
I don’t understand the problem Ive & Altman are trying to solve, much less whether they can actually solve it. It seems managing the device would require more time & effort than going directly to the task.
Not long ago I saw a review of Apple’s new realtime translation software. Why should I rely on it? At present I cannot rely on Apple’s software to render accurately an English voicemail as English text.
I hereby nominate enshi**ification as word of the year.
As you say, the latency problems mean that it is not really always on.
Even if they resolve the latency / battery problem, the security issue looks menacing and will invite all sorts of things … the lapel pin trading crypto or gambling on the NFL without the user exactly knowing because the default setting requires opting out of many obscure bits ?
Do we want to hear machine voices? Another solution in search of a problem.
In a world without want (for people with money), the most profitable technology is one that creates demand. All roads return to social media and video.
The original, 1965-vintage Moore's Law was "every two years." But he revised that in 1975 to be "every 18 months." So Apple's Newton to iPhone was more than 9 (not 7) Moore's Law generations.
As we are now closer to 4 than 5 years out from Keynes's "Project 2030", the question of friend-vs-butler is apt. My qualm with Keynes is that at the end of his paper he blames the British upper class for doing a poor job of teaching the lower ranks how to make use of copious spare time. Obviously, he was not keeping up with the adventures of Bertie Wooster. I propose changing butler to valet, and adding the Jeeves stories to the language model. Keynes championed the enrichment of the average person's life to include that which was formerly the privilege of the few. The model for the man/machine relationship is the feudal spirit, and the upgrade to Siri is Jeeves.
Yes: a 'bot that treated you like Jeeves treated Bertie Wooster would produce a truly amusing version of Brave New World. I think that is, more or less, a lot of the worldbuilding and much of the underplots of Iain M. Banks's "Culture" novels. Characters like Unaha-Closp and Skaffen-Amtiskaw steal the show:
> Albert Short: As we are now closer to 4 than 5 years out from Keynes's "Project 2030", the question of friend-vs-butler is apt. My qualm with Keynes is that at the end of his paper he blames the British upper class for doing a poor job of teaching the lower ranks how to make use of copious spare time. Obviously, he was not keeping up with the adventures of Bertie Wooster. I propose changing butler to valet, and adding the Jeeves stories to the language model. Keynes championed the enrichment of the average person's life to include that which was formerly the privilege of the few. The model for the man/machine relationship is the feudal spirit, and the upgrade to Siri is Jeeves...
Also cf.: John Maynard Keynes: Economic Possibilities for Our Grandchildren.
Making AI do more than gather information and put it in words and perhaps pictures is the problem. I agree with you that I don't see a solution, because AI, for all its gathering and translation abilities, cannot think, and it certainly cannot create. See Putnam.
I’m still skeptical that on-device models can be built along the lines of LLMs but with a smaller local dataset. SLMs still require a lot of energy and heat intensive compute. Maybe they don’t need an array of thousands of Bxxx chips drawing gigawatts of power but even needing a single chip drawing 1000w on the local device is going to make this thing the size of a laptop. Perhaps their investment in AMD is a sign that the instinct line’s competitiveness on price/performance is the way to go. Even so, that’s still a 750w part. Apple’s a19 chipset in the new iPhone is ~10w at peak load. That’s what it’ll take to fit Ive and Altman’s device in the palm of your hand. If the compute is somewhere else, this is not a problem but you run into your latency/reliability problem.
It’s a huge challenge and not one I think anyone is solving quickly.
I don’t understand the problem Ive & Altman are trying to solve, much less whether they can actually solve it. It seems managing the device would require more time & effort than going directly to the task.
Not long ago I saw a review of Apple’s new realtime translation software. Why should I rely on it? At present I cannot rely on Apple’s software to render accurately an English voicemail as English text.
I hereby nominate enshi**ification as word of the year.
I do not like this “thing”.
Why do I want “always on ?”
As you say, the latency problems mean that it is not really always on.
Even if they resolve the latency / battery problem, the security issue looks menacing and will invite all sorts of things … the lapel pin trading crypto or gambling on the NFL without the user exactly knowing because the default setting requires opting out of many obscure bits ?
Do we want to hear machine voices? Another solution in search of a problem.
In a world without want (for people with money), the most profitable technology is one that creates demand. All roads return to social media and video.
One tiny nitpick:
The original, 1965-vintage Moore's Law was "every two years." But he revised that in 1975 to be "every 18 months." So Apple's Newton to iPhone was more than 9 (not 7) Moore's Law generations.
As we are now closer to 4 than 5 years out from Keynes's "Project 2030", the question of friend-vs-butler is apt. My qualm with Keynes is that at the end of his paper he blames the British upper class for doing a poor job of teaching the lower ranks how to make use of copious spare time. Obviously, he was not keeping up with the adventures of Bertie Wooster. I propose changing butler to valet, and adding the Jeeves stories to the language model. Keynes championed the enrichment of the average person's life to include that which was formerly the privilege of the few. The model for the man/machine relationship is the feudal spirit, and the upgrade to Siri is Jeeves.
Yes: a 'bot that treated you like Jeeves treated Bertie Wooster would produce a truly amusing version of Brave New World. I think that is, more or less, a lot of the worldbuilding and much of the underplots of Iain M. Banks's "Culture" novels. Characters like Unaha-Closp and Skaffen-Amtiskaw steal the show:
> Albert Short: As we are now closer to 4 than 5 years out from Keynes's "Project 2030", the question of friend-vs-butler is apt. My qualm with Keynes is that at the end of his paper he blames the British upper class for doing a poor job of teaching the lower ranks how to make use of copious spare time. Obviously, he was not keeping up with the adventures of Bertie Wooster. I propose changing butler to valet, and adding the Jeeves stories to the language model. Keynes championed the enrichment of the average person's life to include that which was formerly the privilege of the few. The model for the man/machine relationship is the feudal spirit, and the upgrade to Siri is Jeeves...
Also cf.: John Maynard Keynes: Economic Possibilities for Our Grandchildren.
Making AI do more than gather information and put it in words and perhaps pictures is the problem. I agree with you that I don't see a solution, because AI, for all its gathering and translation abilities, cannot think, and it certainly cannot create. See Putnam.
Way above VAR for me today. This gives me a lot to think about.
Buzzing my buddy at Apple to see if I can poke that bear and get a response to learn more.
Thanks much... -B.
I’m still skeptical that on-device models can be built along the lines of LLMs but with a smaller local dataset. SLMs still require a lot of energy and heat intensive compute. Maybe they don’t need an array of thousands of Bxxx chips drawing gigawatts of power but even needing a single chip drawing 1000w on the local device is going to make this thing the size of a laptop. Perhaps their investment in AMD is a sign that the instinct line’s competitiveness on price/performance is the way to go. Even so, that’s still a 750w part. Apple’s a19 chipset in the new iPhone is ~10w at peak load. That’s what it’ll take to fit Ive and Altman’s device in the palm of your hand. If the compute is somewhere else, this is not a problem but you run into your latency/reliability problem.
It’s a huge challenge and not one I think anyone is solving quickly.