Long essay that doesn't really say much? What if ChatGPT could be like LARCS? But isn't building something like LARCS the goal of ChatGPT? I don't follow the argument that billions of dollars have been wasted building this, but we could have built that, which is actually the same thing?
Not long, but yes. He starts out as if he has something to say, but doesn't, except to plug something at "I recommend ..." So it's clickbait.
His argument seems headed in the direction of declaring that advertising is bad, but he quits before he gets there. One can make a strong argument that advertising and finance, once rather small industries in terms of employment, have become much too large. They're supposed to be service functions for manufacturing and distribution. For many products, advertising is a significant portion of product cost. When promotional cost becomes a significant fraction of retail price, a company is in the self-promotion industry rather than offering a product. Perhaps tax policy should make this unprofitable. Arguably, it's a driver of unnecessary inflation.
"Advertising", viewed broadly, includes Google and Facebook. We should be thinking of the advertising industry rather than the "tech" industry.
Hmm very interested to hear how you think it relates to advertising? It seems to be about productivity tools.
Also I'd say the first part of the article is a general "the state of affairs sucks" kvetching that is repeated frequently these days in response to the AI surge, as evidenced by the two links. I find the use of the term "plug" in the context of linking those two resources a little needlessly pejorative lol - just illustrating his point quickly by citing existing literature.
The premise, as far as I understand it, is that Star Trek offers a view of the future of technology that is preferrable to the one in Snow Crash. Preferrable because it's about the humans, not the machines.
That's how I read the article anyway. It is a bit hastily written, I think.
Commented elsewhere but I think the point of the article is the focus on the creative use of the system. In his vision I think the "ChatGPT version of LARCS" would be something like the computer searching for answers to questions in its own ways, and if it doesn't know an answer, you're out of luck.
I think it's important to distinguish b/w ChatGPT and LLMs in general - I think he's using the former term to describe monolithic corporate systems, not all LLMs.
> The ability to identify text via search, to summarize and read contents (with just enough contextual capability to be useful) and to output relevant results is rather close, conceptually, to the potential of language models. The difference between what we actually have – competing and discrete systems owned by corporations – and LCARS (besides many orders of magnitude of greater sophistication in the fictional system) is that LCARS is presented as an integrated, holistic and scoped system. LCARS’ design is to be a library that enables access to knowledge and retrieves results based on queried criteria.
But hasn't their been decades worth of work on just that sort of generic system that could work on whatever data sources you have without needing to be trained on a ridiculously huge corpus of data? But that nobody ever figured it out, and the "fuck it, train this thing on all human-generated text we can find" won out over trying to build search + summarization + relevant result generation from first principals?
Sure, you can imagine that future. And people have tried to build it. But I haven't seen any reason to believe that it hasn't worked because of some sort of "too many academics captured by Google" issue. Google didn't even manage to be the first company to commercialize this alternate approach, after all!
>the industry may indeed be building the future but contrary to its claims, it is not a future with human needs centered
It is usually profit centered. There's exceptions of course, but the people working would themselves need to make a living. Naturally, if you had replicators and massive energy production it would be a different story because those things destroy material economics.
I think there's a middle ground between "let technology and society proceed in whatever direction the profit motive of the current power-brokers points" and "eliminate all scarcity altogether". A better society is possible IMO, without the need for any replicators :)
I suggest a game: anyone have some personal activities to do on a desktop like managing a bunch of files, scheduling events with a calendar or something similar and so on. Oh nothing special BUT actual widespread systems what offer?
There are plenty of "calendar app(s)", but you can only ad some entries, no file attachments, no links to relevant mail(s), ... There are plenty of file managers but almost NONE offer anything more than files and directories approach, veeeery few allow for basic tagging and search the tags, just plain text search&narrow, no booleans or something alike for most, even less offer full text search inside files content and files metadata, oh and for all, even Google Drive, mails are not files... Long story short: the actual computing model offer NO INTEGRATION and BIG fragmentation. Initially such choice was for commercial purpose: building a fully integrated environment like Xerox did was extremely complex and do not offer much profit opportunities (especially when it's hard to tell what one can do with such system) while on contrary selling simple systems with limited functionalities and selling extra functionalities a bit at a time, allow BIG profit. But as correctly stated by the Conway's law after a certain timeframe all apps try to do anything since integration is demanded and is not possible in the present model.
Since years few timid ways of integration have seen the light, MOSTLY only when someone have found a way to avoid giving much power to the end user. For instance we have seen the web 2.0 as a successor (aspiring, at least) of classic widget-based UIs and desktop environments, those have evolved to be a very complex WebVM [1] bootloaders and the web 2.0 is a kind of modern original integrated environment "document based" where anything can be linked, most parts can be integrated in a "newly composed page" and so on. Of course, in the original desktop the users do write the "active documents" or "the UI", extending the ones offered by the environment, in the web 2.0 model the user is practically powerless since writing personal JS/CSS etc to bend someone else UIs is a painful and pointless exercise, but that's is. End users programming is came back a little bit, again in limited and limiting ways with notebook UIs, "low code" environments and so on, since well... No point-and-click UIs can cover all functionalities while simple textual strings known as "code" can.
Long story short again: no, "AI" is not an alternated vision, a potential totally different future. The alternate vision is the original desktop model, of people who have a fully integrated, flexible, user-programmable, environment and can communicate with other peers between them. The future is no need of a crappy website per bank to operate but simple APIs, functions the user can call/integrate in hes/shes own documents "notes" to operate. This alternative vision of computing was from the past, still exists a little bit in the present (see https://youtu.be/B6jfrrwR10k as a simple Emacs usage demo) and is the future, the potential future, if enough people realize it.
This model do not happen mostly because those who can spread it do not want at all giving such power to anyone who want to learn. Giving it means the end of the platform goose that lays golden eggs, giving up any position/size advantage and so on, while most ignore it's existence so do not push back against alternate evolution.
The really future model, beside the above would be the integration of human brain and an artificial exobrain, so we do not need specific UIs limited by their physical IO interfaces and concept model. But as per the "AI", the much hyped "neuralink" and alike are just very expensive toys so far, we are far behind the dawn of such era.
This is basically HN-heresy, but I was interested in this comment but was struggling to read it b/c of the style, so summarized it via ChatGPT. Very interesting thoughts, I agree with your end conclusion as I understand it! Especially this bit: "AI" is not an alternated vision, a potential totally different future. The alternate vision is the original desktop model
_The summary:_
The commenter criticizes the current computing model, stating that it is overly fragmented and lacks integration. They point out the limitations of current systems like calendar apps and file managers, noting their inability to link related files and emails or effectively search file content and metadata. They argue that this model was developed for commercial purposes, prioritizing profit over functionality, and that true integration is impossible in this framework.
The commenter identifies a few modest efforts towards integration but believes these fall short as they limit the end user's power. They highlight the shift to web 2.0 and modern integrated environments where everything can be linked, but express disappointment at the lack of end user programming capabilities.
They suggest an alternative vision, rooted in the original desktop model, where users operate in a fully integrated, flexible, and programmable environment. This approach allows users to communicate more effectively and eliminates the need for individual websites for various operations, such as banking.
Finally, they speculate about a future model that integrates the human brain with an artificial exobrain, but dismiss current attempts like "neuralink" as expensive toys. They see potential in this idea, but believe we are far from reaching this level of technology.
> There is a potential, latent within language models and hybrid systems – indeed, within almost the entire menagerie of machine learning methods – to create a unified computational model for a universally useful platform. This potential is being wasted, indeed, suppressed as oceans of capital, talent and hardware is poured into privately owned things such as ChatGPT.
If that is the core point of the article, then it is incredibly vague. All large projects are built by organized groups of humans, and profit-seeking companies are just one form. The article feels like gesticulating at the problem without outlining an actual solution. Building a product requires people (and the capital to support them), coordinated towards a common goal and it can’t be designed by everyone in the world.
I don't think they're objecting to the fact that these products are produced privately, they're concerned that the focus is on large companies building huge purpose-defined systems that humans can just access/ask to do things. As opposed to building a set of linguistic tools that individuals can mix, match, and apply according to their preferences and personal viewpoint on the work.
When I say it like that, it kinda sounds like a restatement of the Unix philosophy - Do One Thing, and Do It Well. Will have to chew on that one...
I think the author is saying that dedicated apps built from raw, open models (like LLaMA) are the path to building LCARS-like systems, not cloud megaservices like ChatGPT.
And there is something to this... Certain LLaMA finetunes can be way more specialized and focused than what OpenAI offers at this time.
I believe what they mean is. It is not the technology that is bad but the way large corporations are using it and how investors are promoting it that is bad.
meta comment: the article have zero substance besides hinting at the frivolity of tech today. and got instant 11 comments. people here are anxious to hear a why (article offers none)
I would say that's a subjective assessment of its substance :). I'd agree it's not very explicit, but I see substance.
Also accepting that assessment as a premise, perhaps Cunningham's Law applies: _the best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer._
Nothing against Star Trek metaphors but its a sign how awkward our internalization of the role of computation in society when the only wholesome vision goes back seventy years, to Douglas Engelbart, augmenting human intellect and collective intelligence.
The greed and fear unleashed by the LLM algorithms is indicative of how dis-empowered and uninformed are vast numbers of people about what digital information technology can and cannot do.
The wasted potential is indeed gargantuan. But power structures never cared about wasted potential, their only concern is perpetuation.
The question is whether there is a threshold of tech availability and power where collective intelligence will emerge spontaneously and despite its oppression.
In other words, is Wikipedia an aberration or the shape of the things to come?
Sci-fi has been providing alternative visions of computation pretty much as long as it's existed. Star-trek has more mainstream awareness, but Bank's Culture series is barely a decade old (1987-2012). If you want more recent works, The Ancillary Justice and Murderbot Diary series' both explore the relationship between AI and society, as well as scores of others.
The important thing here is the mainstream bit. The fact that these visions exist doesn't really move the needle because the popular narrative of rogue AI is such a dominant and constantly reinforced cultural narrative that it drowns out anything else.
By the way, does anyone know why it was named LCARS (by Okuda, presumably)? Neither the abbreviation nor the expansion are particularly catchy or straightforward.
Perhaps the author can also address how our transportation needs can be better solved with teleportation, rather than airplanes, railways, busses, ships, cars, and bikes we have now.
I have a feeling that people who look too much to sci-fiction in describing the future are not that much different than people who read apocalyptic religious writings and try to make real world immediate predictions from them.
And this is why Peter Thiel says Star Trek is communist but Star Wars is capitalist. If I interpret that correctly public LLMs might not lead us to lightsabres but ghosts of dead Jedi masters might /s
29 comments
[ 4.6 ms ] story [ 68.7 ms ] threadNot long, but yes. He starts out as if he has something to say, but doesn't, except to plug something at "I recommend ..." So it's clickbait.
His argument seems headed in the direction of declaring that advertising is bad, but he quits before he gets there. One can make a strong argument that advertising and finance, once rather small industries in terms of employment, have become much too large. They're supposed to be service functions for manufacturing and distribution. For many products, advertising is a significant portion of product cost. When promotional cost becomes a significant fraction of retail price, a company is in the self-promotion industry rather than offering a product. Perhaps tax policy should make this unprofitable. Arguably, it's a driver of unnecessary inflation.
"Advertising", viewed broadly, includes Google and Facebook. We should be thinking of the advertising industry rather than the "tech" industry.
Also I'd say the first part of the article is a general "the state of affairs sucks" kvetching that is repeated frequently these days in response to the AI surge, as evidenced by the two links. I find the use of the term "plug" in the context of linking those two resources a little needlessly pejorative lol - just illustrating his point quickly by citing existing literature.
That's how I read the article anyway. It is a bit hastily written, I think.
I think it's important to distinguish b/w ChatGPT and LLMs in general - I think he's using the former term to describe monolithic corporate systems, not all LLMs.
But hasn't their been decades worth of work on just that sort of generic system that could work on whatever data sources you have without needing to be trained on a ridiculously huge corpus of data? But that nobody ever figured it out, and the "fuck it, train this thing on all human-generated text we can find" won out over trying to build search + summarization + relevant result generation from first principals?
Sure, you can imagine that future. And people have tried to build it. But I haven't seen any reason to believe that it hasn't worked because of some sort of "too many academics captured by Google" issue. Google didn't even manage to be the first company to commercialize this alternate approach, after all!
It is usually profit centered. There's exceptions of course, but the people working would themselves need to make a living. Naturally, if you had replicators and massive energy production it would be a different story because those things destroy material economics.
It'll happen to other forms of supply decoupling like replicators and infinite energy.
I have no proof, yet I'm absolutely certain.
There are plenty of "calendar app(s)", but you can only ad some entries, no file attachments, no links to relevant mail(s), ... There are plenty of file managers but almost NONE offer anything more than files and directories approach, veeeery few allow for basic tagging and search the tags, just plain text search&narrow, no booleans or something alike for most, even less offer full text search inside files content and files metadata, oh and for all, even Google Drive, mails are not files... Long story short: the actual computing model offer NO INTEGRATION and BIG fragmentation. Initially such choice was for commercial purpose: building a fully integrated environment like Xerox did was extremely complex and do not offer much profit opportunities (especially when it's hard to tell what one can do with such system) while on contrary selling simple systems with limited functionalities and selling extra functionalities a bit at a time, allow BIG profit. But as correctly stated by the Conway's law after a certain timeframe all apps try to do anything since integration is demanded and is not possible in the present model.
Since years few timid ways of integration have seen the light, MOSTLY only when someone have found a way to avoid giving much power to the end user. For instance we have seen the web 2.0 as a successor (aspiring, at least) of classic widget-based UIs and desktop environments, those have evolved to be a very complex WebVM [1] bootloaders and the web 2.0 is a kind of modern original integrated environment "document based" where anything can be linked, most parts can be integrated in a "newly composed page" and so on. Of course, in the original desktop the users do write the "active documents" or "the UI", extending the ones offered by the environment, in the web 2.0 model the user is practically powerless since writing personal JS/CSS etc to bend someone else UIs is a painful and pointless exercise, but that's is. End users programming is came back a little bit, again in limited and limiting ways with notebook UIs, "low code" environments and so on, since well... No point-and-click UIs can cover all functionalities while simple textual strings known as "code" can.
Long story short again: no, "AI" is not an alternated vision, a potential totally different future. The alternate vision is the original desktop model, of people who have a fully integrated, flexible, user-programmable, environment and can communicate with other peers between them. The future is no need of a crappy website per bank to operate but simple APIs, functions the user can call/integrate in hes/shes own documents "notes" to operate. This alternative vision of computing was from the past, still exists a little bit in the present (see https://youtu.be/B6jfrrwR10k as a simple Emacs usage demo) and is the future, the potential future, if enough people realize it.
This model do not happen mostly because those who can spread it do not want at all giving such power to anyone who want to learn. Giving it means the end of the platform goose that lays golden eggs, giving up any position/size advantage and so on, while most ignore it's existence so do not push back against alternate evolution.
The really future model, beside the above would be the integration of human brain and an artificial exobrain, so we do not need specific UIs limited by their physical IO interfaces and concept model. But as per the "AI", the much hyped "neuralink" and alike are just very expensive toys so far, we are far behind the dawn of such era.
[1] the monsters mostly known as browsers...
_The summary:_
The commenter criticizes the current computing model, stating that it is overly fragmented and lacks integration. They point out the limitations of current systems like calendar apps and file managers, noting their inability to link related files and emails or effectively search file content and metadata. They argue that this model was developed for commercial purposes, prioritizing profit over functionality, and that true integration is impossible in this framework.
The commenter identifies a few modest efforts towards integration but believes these fall short as they limit the end user's power. They highlight the shift to web 2.0 and modern integrated environments where everything can be linked, but express disappointment at the lack of end user programming capabilities.
They suggest an alternative vision, rooted in the original desktop model, where users operate in a fully integrated, flexible, and programmable environment. This approach allows users to communicate more effectively and eliminates the need for individual websites for various operations, such as banking.
Finally, they speculate about a future model that integrates the human brain with an artificial exobrain, but dismiss current attempts like "neuralink" as expensive toys. They see potential in this idea, but believe we are far from reaching this level of technology.
If that is the core point of the article, then it is incredibly vague. All large projects are built by organized groups of humans, and profit-seeking companies are just one form. The article feels like gesticulating at the problem without outlining an actual solution. Building a product requires people (and the capital to support them), coordinated towards a common goal and it can’t be designed by everyone in the world.
When I say it like that, it kinda sounds like a restatement of the Unix philosophy - Do One Thing, and Do It Well. Will have to chew on that one...
And there is something to this... Certain LLaMA finetunes can be way more specialized and focused than what OpenAI offers at this time.
Also accepting that assessment as a premise, perhaps Cunningham's Law applies: _the best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer._
The greed and fear unleashed by the LLM algorithms is indicative of how dis-empowered and uninformed are vast numbers of people about what digital information technology can and cannot do.
The wasted potential is indeed gargantuan. But power structures never cared about wasted potential, their only concern is perpetuation.
The question is whether there is a threshold of tech availability and power where collective intelligence will emerge spontaneously and despite its oppression.
In other words, is Wikipedia an aberration or the shape of the things to come?
The important thing here is the mainstream bit. The fact that these visions exist doesn't really move the needle because the popular narrative of rogue AI is such a dominant and constantly reinforced cultural narrative that it drowns out anything else.
I have a feeling that people who look too much to sci-fiction in describing the future are not that much different than people who read apocalyptic religious writings and try to make real world immediate predictions from them.