& we will see where we get, and whether we need… ROADS! From searchbox to sidekick: is the humble web browser the secret to mastering the potential of MAMLMs as natural-language interfaces and as...
I'm an old guy and can remember many things that err going to change everything, but didn't. Automation was going to end humans building cars. Mortgage backed securits were totally secure. The new freeways in LA were going to end traffic congestion. Allen Kamp
Include any thoughts on how you think enshittification will likely proceed - e.g., subsequently charging for useful features such as cloud storage of documents, or particular AI-assisted features (Grammarly is doing that). This is not just an ongoing issue, but appears to be expanding in use. If the service shuts down, will there be any way to retrieve documents in the cloud? Can the service work offline or must it be used only online?
Amusing, but the cloud service prevents true ownership of a product. This has been demonstrated by both big and small companies. Even Google has provided cloud services, then discontinued them, stranding the end user or business. No doubt there is a list of companies that have ended services through bankruptcy, deciding to end a service completely for products no longer supported, or simply to screw payments from users who thought they had paid for the service.
Traditional products were owned by the buyer, no matter what happened to the vendor after purchase. But if functionality requires the vendor to both keep supporting the product and stick to the original terms of purchase, then the cloud becomes a tether that the vendor can exploit.
Sometimes the product is not something that can be easily replaced after it is bricked or made too expensive to keep purchasing. A bit like Trump's "trade deals".
Here's a poser: I want the best possible estimate of the historical growth rate of SKUs, to serve as an indicator of product proliferation and market saturation over time. ChatGPT provides a very cogent and convincing analysis, saying there's no reliable overall measure but we can make good guesses using things like Amazon data. But I can't really cite ChatGPT output, and it provides no citations for itself!
The author is writing about her vibes, not about whether the technology works beyond the relatively narrow categories that are hyped as AI and around-the-corner AGI. How many AI summers and winters have we had since the 1960s? For AI to be let loose, especially in teh physical world with robots, it will need very different underlying technology. LLMs are most certainly not it, however scaled it is. Prof. DeLong uses a better term, MAMLM to describe the range of ML technologies that can be used as part of AI, wich can be used to enhance the performance of LLM-based AI. IMO, LLMs should be mostly restricted to teh conversational interface, with other, more robust, technologies to do teh heavy lifting. LLMs are like a typical Joe Six-Pack, somewhat intoxicated, with idiot savant characteristics, holding forth on topics and BS'ing aggressively when the facts are unknown or mistaken. You wouldn't want such a person piloting a commercial aircraft or doing surgery, or even as a babysitter when you go out for the evening. All the hype about job replacement occurs because many people, even bosses, are unaware of what tasks need to be done correctly every time. Try to replace people fully, and it will be quite evident how poor AIs are at replacing people. Even Musk discovered that you can over-automate car manufacturing, something that was known over 30 years ago about factory automation.
Over my lifetime, there have been a number of faddish technologies that came and went. AI will eventually work to AGI level, but it won't be in a year or two, as currently claimed by the hype of spokespeople "talking up their own book."
I'm an old guy and can remember many things that err going to change everything, but didn't. Automation was going to end humans building cars. Mortgage backed securits were totally secure. The new freeways in LA were going to end traffic congestion. Allen Kamp
There is an article in the Guardian about hardware "frugal tech" to avoid the problems of predatory big-tech. There is probably something similar in software. Ironically, AI might democratise coding for almost anyone to build simple software tools. https://www.theguardian.com/commentisfree/2025/aug/02/billionaire-big-tech-frugal-elon-musk-innovation
I look forward to your report.
Include any thoughts on how you think enshittification will likely proceed - e.g., subsequently charging for useful features such as cloud storage of documents, or particular AI-assisted features (Grammarly is doing that). This is not just an ongoing issue, but appears to be expanding in use. If the service shuts down, will there be any way to retrieve documents in the cloud? Can the service work offline or must it be used only online?
Ah yes, the cloud. What if the cloud drifts away, evaporates, gets hit by lightning? The metaphors are endless.
Amusing, but the cloud service prevents true ownership of a product. This has been demonstrated by both big and small companies. Even Google has provided cloud services, then discontinued them, stranding the end user or business. No doubt there is a list of companies that have ended services through bankruptcy, deciding to end a service completely for products no longer supported, or simply to screw payments from users who thought they had paid for the service.
Traditional products were owned by the buyer, no matter what happened to the vendor after purchase. But if functionality requires the vendor to both keep supporting the product and stick to the original terms of purchase, then the cloud becomes a tether that the vendor can exploit.
Sometimes the product is not something that can be easily replaced after it is bricked or made too expensive to keep purchasing. A bit like Trump's "trade deals".
Here's a poser: I want the best possible estimate of the historical growth rate of SKUs, to serve as an indicator of product proliferation and market saturation over time. ChatGPT provides a very cogent and convincing analysis, saying there's no reliable overall measure but we can make good guesses using things like Amazon data. But I can't really cite ChatGPT output, and it provides no citations for itself!
Interesting article on AI Denialism: https://yaschamounk.substack.com/p/the-peculiar-persistence-of-the-ai?utm_source=post-email-title&publication_id=2709399&post_id=169786805&utm_campaign=email-post-title&isFreemail=true&r=15flyq&triedRedirect=true&utm_medium=email
thx v. much... - B.
The author is writing about her vibes, not about whether the technology works beyond the relatively narrow categories that are hyped as AI and around-the-corner AGI. How many AI summers and winters have we had since the 1960s? For AI to be let loose, especially in teh physical world with robots, it will need very different underlying technology. LLMs are most certainly not it, however scaled it is. Prof. DeLong uses a better term, MAMLM to describe the range of ML technologies that can be used as part of AI, wich can be used to enhance the performance of LLM-based AI. IMO, LLMs should be mostly restricted to teh conversational interface, with other, more robust, technologies to do teh heavy lifting. LLMs are like a typical Joe Six-Pack, somewhat intoxicated, with idiot savant characteristics, holding forth on topics and BS'ing aggressively when the facts are unknown or mistaken. You wouldn't want such a person piloting a commercial aircraft or doing surgery, or even as a babysitter when you go out for the evening. All the hype about job replacement occurs because many people, even bosses, are unaware of what tasks need to be done correctly every time. Try to replace people fully, and it will be quite evident how poor AIs are at replacing people. Even Musk discovered that you can over-automate car manufacturing, something that was known over 30 years ago about factory automation.
Over my lifetime, there have been a number of faddish technologies that came and went. AI will eventually work to AGI level, but it won't be in a year or two, as currently claimed by the hype of spokespeople "talking up their own book."