I find it really astonishing with what naivitë many people are looking at the current capabilities of AI. Unfortunately, many people just don't know what art is all about. For them, a AI generated replica is the same as an artifact created by a human being who tried to express him/herself in an original and deeply personal way. If you are flooded with mediocrity, how would you now the difference. There is so much to unpack here. Thanks for the post, Edward.
_semantic satiation_ sounds like the Big Lie (große lüge) tactically applied for techDazzle (bamboozle the non-technoPriesthood). Part of the problem is the abuse/misuse of artificial "stupidity" since Hollywood has bad analogies (SkyNet, Matrix) to existing art (GPT are statistical parrots). In effect VCs are grooming white collar workers to be (input) slaves to the (inevitable) machine of AGI (magic bullet for world's ills). Fightback ...
1. counter-spin of "Techno-feudalism" and return to corporate "parishes" ruled by AGI-priesthood
2. "give" away (free as in speech) cheap models-as-a-Service to deflate future price-gorging
3. look for realistic machine learning stories ... (think mentat & not Butlerian Jihad)
LLMs do have a very limited use case: replacing low level white collar workers
And I don't mean call centers, I mean lawyers and accountants and graphic artists and technical writers and reporters and columnists, etc etc etc
Is this a killer app? Not at all clear to me. There simply are not a lot of customers who need to do DOGE like things.
Equally, the senior lawyers who can use LLMs to cut down on paralegals and junior lawyers, for example - I think they will all have to die before LLM users get into the senior ranks because they don't do the grunt level IT work as it is - their secretaries and paralegals do it. Hard to see how LLMs are going to replace the only ones able to operate them.
As for the defense angle: my view on Palantir has always been that this is an Oracle-type company, but for online information. As far as I can tell, they have failed miserably at deploying this type of information for real world private sector commercial purposes - you know, where there is actual accountability - so it should not surprise anyone that Palantir is thriving in government/intel agency contracts.
As for Anduril: it is striking just how different Western drones look vs. Ukrainian and Russian ones. Warfighting equipment is like any other real world work gear, only more so: it should only contain what is needed for the job. A steam shovel is a butt ugly monstrosity but it does its job really well. Sleek shiny Anduril drones certainly cost multiples, if not orders of magnitude, more than the real world ones being proven in Ukraine. They are an improvement only compared to the 50 and 100 million dollar plus drones being shot down by Yemenis.
No, I think Zitron has it right: hyperscaler desperation combined with end stage Silicon Valley Uber "value building" has metastasized into its final form: LLM/AI.
I find it really astonishing with what naivitë many people are looking at the current capabilities of AI. Unfortunately, many people just don't know what art is all about. For them, a AI generated replica is the same as an artifact created by a human being who tried to express him/herself in an original and deeply personal way. If you are flooded with mediocrity, how would you now the difference. There is so much to unpack here. Thanks for the post, Edward.
_semantic satiation_ sounds like the Big Lie (große lüge) tactically applied for techDazzle (bamboozle the non-technoPriesthood). Part of the problem is the abuse/misuse of artificial "stupidity" since Hollywood has bad analogies (SkyNet, Matrix) to existing art (GPT are statistical parrots). In effect VCs are grooming white collar workers to be (input) slaves to the (inevitable) machine of AGI (magic bullet for world's ills). Fightback ...
1. counter-spin of "Techno-feudalism" and return to corporate "parishes" ruled by AGI-priesthood
2. "give" away (free as in speech) cheap models-as-a-Service to deflate future price-gorging
3. look for realistic machine learning stories ... (think mentat & not Butlerian Jihad)
This is a nice rant but off base.
LLMs do have a very limited use case: replacing low level white collar workers
And I don't mean call centers, I mean lawyers and accountants and graphic artists and technical writers and reporters and columnists, etc etc etc
Is this a killer app? Not at all clear to me. There simply are not a lot of customers who need to do DOGE like things.
Equally, the senior lawyers who can use LLMs to cut down on paralegals and junior lawyers, for example - I think they will all have to die before LLM users get into the senior ranks because they don't do the grunt level IT work as it is - their secretaries and paralegals do it. Hard to see how LLMs are going to replace the only ones able to operate them.
As for the defense angle: my view on Palantir has always been that this is an Oracle-type company, but for online information. As far as I can tell, they have failed miserably at deploying this type of information for real world private sector commercial purposes - you know, where there is actual accountability - so it should not surprise anyone that Palantir is thriving in government/intel agency contracts.
As for Anduril: it is striking just how different Western drones look vs. Ukrainian and Russian ones. Warfighting equipment is like any other real world work gear, only more so: it should only contain what is needed for the job. A steam shovel is a butt ugly monstrosity but it does its job really well. Sleek shiny Anduril drones certainly cost multiples, if not orders of magnitude, more than the real world ones being proven in Ukraine. They are an improvement only compared to the 50 and 100 million dollar plus drones being shot down by Yemenis.
No, I think Zitron has it right: hyperscaler desperation combined with end stage Silicon Valley Uber "value building" has metastasized into its final form: LLM/AI.
Then I’m sure you’ll love our episodes together next week!