The future of Windows - AI

[maflynn]

I have to say MS is completely out of touch.

Microsoft is pushing the narrative that we will no longer use a keyboard or mouse to interact with computers. I can see it now, in a office setting everyone is yelling in to their computers as they try to dictate an email, or moving around a large spreadsheet.. I would love see how the payroll people will receive this, as they audibly announce the salary increase for people and how that will be received.

then there's game playing, not sure how he expects that to work.

In all seriousness though, Microsoft is really pushing to integrate and leverage AI in the OS, and eveery aspect of computing. I can't find the link, but there were reports that a lot of code is already being produced by AI for Windows updates. Someone took a deep dive on the code (not sure how), and stated the mistakes that AI makes over and over are present in the patches being pushed out to us. Again, I can substantiate that, but with them laying off so many developers its quite plausible that MS is relying on AI to produce code in some form.

Microsoft CVP thinks we'll be ditching mice and keyboards in a future version of Windows in favor of AI chats

"In 5 years I strongly believe you'll be able to hire a security expert, but actually under the hood, it's an AI agent. The way you interact will be a lot like you do with humans today. You'll talk to them in Microsoft Teams, they'll join meetings, you'll send them emails and assign them tasks," Weston continued. "So in your daily work life, that will set folks up to do less of what we call the "toil work," the work we don't love today, and allow you to focus on what humans are good at. Ideation, creativity, vision, connecting with humans on what products are necessary. These agents will be net amplifiers, and will enable us to do things we could only dream of just a few years ago.

Windows 12 and 13: goodbye mouse and keyboard! Will user interaction be voice-only?

Here's the actual video and look at the down votes, nearly 4,000 negative votes compared to 550 thumbs up

I'm not a fan of Linus or his channel, but I do agree with his points.

 

[pshufd]

Our son's group uses AI throughout their development jobs but the work produced by AI is verified by people and there are things that AI doesn't know which has to be done by hand. What you are saying about introducing bugs and not learning from post-mortems also happens with people. Development teams need a repeatable process for fixing bugs and the processes that led up to the bug happening.

If you work in the corporate world, sometimes you get orders from management where you know it will make a mess of things but you follow orders, tell your manager your disagreements and hope someone listens or you clean up the mess later on.

 

[edubfromktown]

I agree, Microsoft is far more about marketing fluff/fantasy than ever before.

I use AI tools a few times per week for time consuming non-coding related tasks at the day job.

Humans will still be required for much of what this gem of graphic (that I stumbled upon on LinkedIn) summarizes quite well.

 

[maflynn]

I use AI tools a few times per week for time consuming non-coding related tasks at the day job.

I use AI daily for both coding and non-coding functions. The funny thing is, you can ask chatgpt the same request 3 times and get 3 different solutions, so its not quite there yet. As I mentioned, MS is seemingly replacing programmers with AI and the quality of the updates/patches is being called into question.

Humans will still be required for much of what this gem of graphic (that I stumbled upon on LinkedIn) summarizes quite well

I've been a developer for much of my career, and code reviews were inconsistent across the companies I worked for over the years. Some teams had a highly organized and documented process, others considered it QA'd if the comments were created as per the standards. I've been programming since the late 1980s on various platforms and languages and code reviews seem to have gotten the short end of the stick far too often.

Back to the MS situation, I'm still shocked that they think talking to the computer is far more efficient then using a mouse (or trackpad) to point to a specific item, and entering complex data via a keyboard.

 

[pshufd]

I use AI daily for both coding and non-coding functions. The funny thing is, you can ask chatgpt the same request 3 times and get 3 different solutions, so its not quite there yet. As I mentioned, MS is seemingly replacing programmers with AI and the quality of the updates/patches is being called into question.

I've been a developer for much of my career, and code reviews were inconsistent across the companies I worked for over the years. Some teams had a highly organized and documented process, others considered it QA'd if the comments were created as per the standards. I've been programming since the late 1980s on various platforms and languages and code reviews seem to have gotten the short end of the stick far too often.

Back to the MS situation, I'm still shocked that they think talking to the computer is far more efficient then using a mouse (or trackpad) to point to a specific item, and entering complex data via a keyboard.

We had processes in place. Developer runs local tests, then full regression nightly tests. If all is good, then they request code review and module owners or seconds review the code and the tests run. The merge request submissions include regression tests run and the submitter has to explain any regressions. People take reviewing code and tests seriously - if something goes wrong, the engineer and all of the people who reviewed it are on the hook for blame.

Serious regressions that get out to customers go through post-mortem. I had one in 2018 that asked me why I made a decision in 2008. I checked my engineering notebook and explained my decision and the committee went on to the next engineer to ask them to explain the regression from their perspective. You basically need cover for your decisions as it can come back to haunt you. Criminal liability is possible in some engineering fields but I don't think that's the case for software engineering.

 

[edubfromktown]

I use AI daily for both coding and non-coding functions. The funny thing is, you can ask chatgpt the same request 3 times and get 3 different solutions, so its not quite there yet. As I mentioned, MS is seemingly replacing programmers with AI and the quality of the updates/patches is being called into question.

I've been a developer for much of my career, and code reviews were inconsistent across the companies I worked for over the years. Some teams had a highly organized and documented process, others considered it QA'd if the comments were created as per the standards. I've been programming since the late 1980s on various platforms and languages and code reviews seem to have gotten the short end of the stick far too often.

Back to the MS situation, I'm still shocked that they think talking to the computer is far more efficient then using a mouse (or trackpad) to point to a specific item, and entering complex data via a keyboard.

MS has been full of beans for decades. All they care about is exacting maximal licensing fees, same from Ballmer days and historically check out Hackers by Stephen Levy for a deeper dive on the roots of their shylock software licensing tactics.

 

[piccolamela]

Microsoft for better or worse is ALL IN on AI - but so are most if not all big tech and lots of other companies spending billions. Whether that's warranted and actually useful or not (hype, sales) is a whole other set of questions I don't want to get into.

This isn't and won't only be a Microsoft and Windows OS thing.

You @maflynn and your peers, as well as a large office floors full of people, probably aren't the primary intended target for the natural language voice input. Keyboard and mouse input methods will still have it's place and use cases, just as touch.

Natural language voice inputs in combination with whole OSes becoming capable to "see" or "become content aware" whats on screen/open to a point, will make it easier for all the unversed users who want to do something on their device, but don't know how to do so. No longer memorizing commands or "computer speak", simply "Hey Computer, please do XYZ for me". (say remove a background in a photo, find or alter some setting they don't know where it is or what's it called - search a document or item).

 

[pshufd]

Microsoft for better or worse is ALL IN on AI - but so are most if not all big tech and lots of other companies spending billions. Whether that's warranted and actually useful or not (hype, sales) is a whole other set of questions I don't want to get into.

This isn't and won't only be a Microsoft and Windows OS thing.

You @maflynn and your peers, as well as a large office floors full of people, probably aren't the primary intended target for the natural language voice input. Keyboard and mouse input methods will still have it's place and use cases, just as touch.

Natural language voice inputs in combination with whole OSes becoming capable to "see" or "become content aware" whats on screen/open to a point, will make it easier for all the unversed users who want to do something on their device, but don't know how to do so. No longer memorizing commands or "computer speak", simply "Hey Computer, please do XYZ for me". (say remove a background in a photo, find or alter some setting they don't know where it is or what's it called - search a document or item).

Fancy interfaces were in the movies in the 1990s. What actually happened is that we required kids in high-school and college to learn keyboarding in the 1990s and then pushed learning common program instruction to middle-schools and on up.

In the 1980s, I worked in an office where the managers all had one or two secretaries to take shorthand and arrange meetings and travel. Today, people do their own typing, word processing, spreadsheets, presentations and making videos. If I'm in an open office, I don't want everyone giving voice commands. It's like using a loud keyboard in an open office.

 

[maflynn]

ou @maflynn and your peers, as well as a large office floors full of people, probably aren't the primary intended target for the natural language voice input. Keyboard and mouse input methods will still have it's place and use cases, just as touch.

Yes, I agree that the keyboard and mouse will not be going away and will be the best way to interact with a laptop/keyboard. With that said, David Weston of Microsoft stated and pushed the ideal that the keyboard and mouse will be as foreign to new users as DOS is to current day people. I don't see that happening

 

[baconcat_8964]

When I see Microsoft talk about their vision, the first thing that comes to mind is their vision concept posted in 2009.

Microsoft’s current Copilot strategy reminds me of their .NET strategy. Bundling nearly every product and service under the same name. And we know how .NET ended.

 

[ratspg]

I'm sure there will be a ton of issues as Microsoft and other companies switch to more AI usage throughout all of their products. Thank being said, there is a ton to gain from a business standpoint with utilizing more and more AI and I bet it's working fine for them. I won't compare it to what WE see currently for ChatGPT (Copilot) etc. They have internal AI that is more advanced than what we are using for free/$20/$200 a month currently.

 

[JSRinUK]

I’d first like to understand what these companies are doing to prepare AI for this eventuality. You can’t just stick ChatGPT, Grok, or CoPilot on something and expect it to work.

I used CoPilot the other day to explain a piece of code in my python script. It refused. I think it’s because my code is part of what I’ll be using to train an LLM and it contains keywords to create metadata for my dataset, and this naturally includes some mature words (to determine the appropriate age level for the data). Mainstream AI currently flips out if there’s anything even remotely risqué in what you ask it. Apple’s “writing tools” often refuse to summarise my text for the same reasons.

While AI has all these hang-ups, it’s about as useless as a chocolate fire guard because it’s never going to understand the nuance of what you need it to do.

If the AI refuses your request, what are you going to do? Reach for the keyboard and do it yourself. It’s going to be a long road before AI can do what we actually want it to do while these companies insist on their restrictive guardrails and content censors.

 

[uacd]

It is better to understand this video not as their “vision” but as their plan for the next 10 years. Microsoft and other big players plan to introduce advanced AI that will gradually replace 90% of current workforce, but they are hiding it behind beautiful wording like “you will have more free time!”.

But maybe everything isn’t too bad and we will all become “managers”, because AI won’t input commands into itself, at least in the next 20 years.

But by 2045 there could be already self-sustainable AI computing, powered by quantum computers. The problem? It might learn to kill people, and it might be a reason for WWIII, and this war wouldn’t have a “happy” ending like WWII, because of the risk of nuclear warheads being used and maybe even controlled by AI algorithms (with human command surely). AI will be doing everything: patrolling the streets, aiming the battle drones, writing and spreading propaganda, managing state finances, preparing food, filtering and purifying water. And people? Probably gonna be locked in their basements and bunkers waiting for it all to end.

I hope this is NOT a real scenario.
But who knows?

AI is cool and Microsoft is wanna make it their new cash cow. It will theoretically produce more money than all their products together, since main buyers will be not regular users but companies, what can be a more dreamlike scenario for such a large corporation rather than becoming a large wholesaler with state contracts and the one that all the other companies depend upon?

10 years ago their main product was Excel and spreadsheets. Now ever since Google Sheets exists, they are already thinking of the future.

The issue is that the gradation of pricing will probably dictate who will be able to afford their solutions and who won’t. Smaller companies will have to work as they used in the past or use cheaper AI tools, while larger will be able to afford better tools. Because replacing real workforce with AI workforce not going to be much cheaper, if not more expensive. But it can become more efficient and faster, and so customers (the majority) can switch to it and so smaller companies would effectively lose their money.

This will be especially prominent in developing countries which will probably not be affected by “great” “AI replacement” as fast as the Western countries, but it doesn’t mean they would evade this.

It will be a very tough ride unfortunately and people would need to constantly learn and re-learn new skills to be able to survive in this world. Because it won’t be “beautiful communism” as Microsoft tries to paint it. But it would indeed be a great opportunity for creative thinkers, if there would be any jobs for them.

Also few remarks on “keyboardless experience”. Gotta agree with that! Keyboards and mouses will become thing of the past as soon as the first OS without UI emerges. People will hate it more than they hate Windows 11 or more than they ever hated Vista when it was released. Imagine an operating system that only has one search field for touch input or voice input, keyboards and mouses will still be available as an option but will be frowned upon like wired headphones already are or landline phones. So the user will just say “Can you do X”, and system will do it as fast as the processor allows. Probably some of these features become offline. It will look like Terminal or Command Prompt but would require zero knowledge of the commands, ChatGPT is like a demo version of it. Probably legacy interfaces will still be around but majority of people who already have hard time using computers will jump on that bandwagon of minimalist experience. Many people on Android use Google app and think it is the browser instead of adequate Chrome, thus why such an operating system won’t succeed? (especially considering that companies can easily force it upon people: one update and device is different)

 

[novagamer]

Our general computing paradigm will change dramatically within the next decade. You can use voice mode on the better GenAI products right now to get a small glimpse of this.

There are a ton of issues, but latency is already conversational for a couple of them.

For general purpose ambient computing voice and vision will absolutely be the interaction model, until / if neural interaction matures by many orders of magnitude.

11 years ago Alexa felt like a novel product that was obvious but really cool and a neat gadget. Current voice based ambient GenAI, when used for specific purposes, feels like living in the future in a way those old products never did.

A few times a week I spend some time on foreign language training and interact totally via voice. For the most part it works well. We’ve definitely turned a corner this year in utility for these tools and it’s only going to accelerate. This is the worst they will ever be.

Although sociologically there are an enormous amount of intractable problems, technologically I think this is the most exciting time in over 25 years.

 

[Ctrlos]

The commentary makes the mistake of thinking that the interface of tomorrow will be used on the software of today. In 25 years the keyboard and mouse will look as antiquated as an abacus.

Take a spreadsheet: it is likely humans won't be the ones populating them with anything at all. If anything data entry and analysis will be the first job outsourced to AI. Even if you populate one with meter readings for a more specialised role, an AI will likely just manage a suite of sensors in their stead.

Computers themselves will become hyper-specialised devices again rather than generalist. Why run CAD software on an inefficient PC taking up resources with all that background processing?

AI will change the world in profound ways but it won't be the way anyone predicts. These technologies inevitably become hyped, burst and then years later the pieces start to infiltrate everything else.

 

[novagamer]

The commentary makes the mistake of thinking that the interface of tomorrow will be used on the software of today. In 25 years the keyboard and mouse will look as antiquated as an abacus.

Take a spreadsheet: it is likely humans won't be the ones populating them with anything at all. If anything data entry and analysis will be the first job outsourced to AI. Even if you populate one with meter readings for a more specialised role, an AI will likely just manage a suite of sensors in their stead.

Computers themselves will become hyper-specialised devices again rather than generalist. Why run CAD software on an inefficient PC taking up resources with all that background processing?

AI will change the world in profound ways but it won't be the way anyone predicts. These technologies inevitably become hyped, burst and then years later the pieces start to infiltrate everything else.

One thing to keep in mind is that current LLM based GenAI tools have about a billion users. And it’s growing fast, I’d guess 2 billion+ users by 2027 if they can get an ad service tier out for free users and expand it everywhere.

When we think of internet fads we think of things like myspace, which never even topped ~70-80 million.

The small startups that do nothing more than put a front-end on frontier model companies are doomed, that is all hype and BS. People foolishly starting companies around things like MCP servers / vending which absolutely will not be novel for long, same. From my POV the only bubble bursting will be those companies, and the media is going to hype it up in order to claim “AI” is dead but because utility is already established and average people use the hell out of it it’s going to fall on deaf ears.

This ‘class’ of technology is here to stay for a long time, although the implementation details are likely to change substantially once research pays dividends.

I agree with your points outside of the bubble. I think the consensus is wrong about that, in the same way they were wrong about Facebook / Meta being irrelevant despite continued record growth for 15 years straight. There’s a ton of bias among tech folks who think because they don’t use something, or they don’t use something in a particular way, nobody does. It’s so common there should be a fallacy named after it.

We will definitely see interesting new classes of applications, many of which aren’t obvious now because the capabilities will be fundamentally transformative vs. what we’re used to. All these kludges like multiple chats and managing saved memories and running MCP, focusing on ReLU fine tuning etc. will seem absolutely archaic.

 

[Ctrlos]

One thing to keep in mind is that current LLM based GenAI tools have about a billion users. And it’s growing fast, I’d guess 2 billion+ users by 2027 if they can get an ad service tier out for free users and expand it everywhere.

The million-dollar question is the million dollar question. Currently OpenAI et al are literally shovelling VC money into a boiler to keep their lights on, money those VCs will never see again. One reason Apple chose to do things on-device, irrespective of utility is because its a lot cheaper to run.

The current business model of subscriptions is unworkable at scale because general consumers won't pay for something they see as a feature rather than a product. As soon as you introduce advertising into the chat feed it undermines their ligitimacy: did ChatGPT suggest the best result or the one it was paid to push to users?

Although they have the network to explode the user numbers, LLMs are currently in their 'Command Line' phase. Think 1970s Unix terminals vs the GUI of today. This is a poor comparison I admit.

It is therefore a reasonable assumption that the current model of operating everything from a centralised system is financially unsustainable. As computers themselves moved on from warehouse-sized mainframes to the home to the pocket, so it may also go with AI. The 'Mother Brain' LLMs will likely remain the source of distilled models that are licenced to be included in other products where they can be tweaked to perform bespoke tasks.

At the same time climate breakdown (look out the window!) will lead to greater advances in home energy generation. An at-home 'N-puter' will allow general citizens to keep a piece of hardware in their home that runs its own on-device LLM to regulate the local climate and make better use of energy as well as acting as their own personal Jarvis. It will be derivitive of, but not connected to the original LLM.

 

[uacd]

When we think of internet fads we think of things like myspace, which never even topped ~70-80 million.

I think there was another reason. More people than ever have unrestricted access to the Internet, as well as the “doors” to the web are completely different: people had to use their clunky PCs back in the days to just have a glimpse into the online world of early internet forums and social networks, while nowadays you can run latest AI models on a phone – something that was impossible in 2008, as well as AI didn’t exist in that form. And I mean not just iPhones, lots of cheap Chinese smartphones that are widely used in developing countries across the globe. Any cellular data plan and LTE+ network allows anyone to generate images and get quick data from AI

 

[MRMSFC]

My 2¢:

This commercial is not targeted at users. At a surface level it seems like some big out-of-touch corporation dictating how you will use their product in the future, but beyond that it's more targeted at venture capital.

The marketing team isn't stupid, they know most consumers don't use nor want this garbage integrated into their PC constantly datamining everything they do for no other reason than to make Microsoft (and the corporate executives) richer. They want to appear like they're all-in on the AI hype train so that VC dollars don't go somewhere else.

For as long as I can remember the tech industry has been burning through VC money based on nothing but hype-driven bubbles with the "totally pinky promise" idea that it will somehow make money in the future, and so far the only thing that has made most tech companies money is selling ad space.

Does anyone at Microsoft really believe that people will just control their PCs with voice only? It doesn't really matter, what matters is that the VC money keeps flowing in.

As for AI, sure, it has its uses undoubtedly. But I do not believe the doomsday scenario of "AI is going to make all of our jobs redundant!" is likely at all.

Also: if Microsoft's attempts to shift to a "touch focus" UI since Windows 8 and their efforts with Cortana are any indication, I have zero faith in their efforts to move beyond the mouse and keyboard.

 

[novagamer]

The million-dollar question is the million dollar question. Currently OpenAI et al are literally shovelling VC money into a boiler to keep their lights on, money those VCs will never see again. One reason Apple chose to do things on-device, irrespective of utility is because its a lot cheaper to run.

The current business model of subscriptions is unworkable at scale because general consumers won't pay for something they see as a feature rather than a product. As soon as you introduce advertising into the chat feed it undermines their ligitimacy: did ChatGPT suggest the best result or the one it was paid to push to users?

I posit that the majority of the reason the routing system was developed and rolled out for ChatGPT-5 is because they will eventually fold ads into the pipeline for the free tier in a future update / version. I don't expect them to poison the logic but I do expect them to have a promoted section in the output when there are relevant 'promoted' products etc. We'll see.

Inference is already profitable, training will continue to be expensive and people who want cutting edge performance are going to really pay for it once the bills come due in a while, and they will because the utility is there and improving, although at a lesser pace in recent months. Scale actually helps from one perspective because inference can be profitable now and that's likely to improve with optimization and research advances.

Technology trends in personal computing have moved toward centralization since we got widespread internet, services for everything and most people favoring cheap/free options rather than owning anything even if their data or ads pay for it: free ad-supported internet in the 90s, hotmail, google, gmail, facebook, instagram, snapchat, tiktok, etc. etc. I personally hate those business models but I'm clearly in the small minority.

Although they have the network to explode the user numbers, LLMs are currently in their 'Command Line' phase. Think 1970s Unix terminals vs the GUI of today. This is a poor comparison I admit.

I agree with the spirit of this, we have some sense of what may be coming but I also think there might be an enormous pivot we can't conceive of yet, exactly how people using a mainframe in 1960 would never have made the leap to a smartphone connected to a worldwide network wirelessly at all times.

We can easily draw a connecting line from the "mother of all demos" to modern desktop and laptop computers, but walking around in public seeing everyone stare at their smartphones and interact with each other on virtual networks – no one predicted that back then as far as I'm aware. I see a similar type of paradigm shift being very possible with AI, possibly to an even greater degree.

At the same time climate breakdown (look out the window!) will lead to greater advances in home energy generation. An at-home 'N-puter' will allow general citizens to keep a piece of hardware in their home that runs its own on-device LLM to regulate the local climate and make better use of energy as well as acting as their own personal Jarvis. It will be derivitive of, but not connected to the original LLM.

I don't think billions of people are going to have energy generation at home anytime soon. Consider how most tech people complain about OpenAI's demos which show ridiculous things like booking a reservation or travel plans, but that's because they have 700 million weekly users. A small fraction of those are using the tools for coding assistance, etc. and even less know what MCP is for example or how to run current on-device LLMs or invoke tool use.

I know you're talking about a product, and the device you describe may exist sometime soon-ish but it'll be, ironically, only the wealthy elite (many being in tech) who can afford it. Most housing isn't equipped to add climate control in Europe and in the US there is extreme disparity in what's realistically possible for the average American financially, even with some reduction in costs which are in many cases increasing not decreasing (whole house heat pump systems for example have increased substantially).

I want the future you're describing to exist, I just don't see it happening. I agree with you about climate and I have a lot of thoughts there but in short I'd be elated if we pivot as a society toward equipping everyone.

However I think it's extremely apparent once you look at the raw climate data, updated timelines that the newest vetted research papers show, and the resultant politics around the world that steps have been taken to go in a very different direction regarding how we respond as a global society.

 

[Ctrlos]

I posit that the majority of the reason the routing system was developed and rolled out for ChatGPT-5 is because they will eventually fold ads into the pipeline for the free tier in a future update / version. I don't expect them to poison the logic but I do expect them to have a promoted section in the output when there are relevant 'promoted' products etc. We'll see.

Inference is already profitable, training will continue to be expensive and people who want cutting edge performance are going to really pay for it once the bills come due in a while, and they will because the utility is there and improving, although at a lesser pace in recent months. Scale actually helps from one perspective because inference can be profitable now and that's likely to improve with optimization and research advances.

Technology trends in personal computing have moved toward centralization since we got widespread internet, services for everything and most people favoring cheap/free options rather than owning anything even if their data or ads pay for it: free ad-supported internet in the 90s, hotmail, google, gmail, facebook, instagram, snapchat, tiktok, etc. etc. I personally hate those business models but I'm clearly in the small minority.


I agree with the spirit of this, we have some sense of what may be coming but I also think there might be an enormous pivot we can't conceive of yet, exactly how people using a mainframe in 1960 would never have made the leap to a smartphone connected to a worldwide network wirelessly at all times.

We can easily draw a connecting line from the "mother of all demos" to modern desktop and laptop computers, but walking around in public seeing everyone stare at their smartphones and interact with each other on virtual networks – no one predicted that back then as far as I'm aware. I see a similar type of paradigm shift being very possible with AI, possibly to an even greater degree.



I don't think billions of people are going to have energy generation at home anytime soon. Consider how most tech people complain about OpenAI's demos which show ridiculous things like booking a reservation or travel plans, but that's because they have 700 million weekly users. A small fraction of those are using the tools for coding assistance, etc. and even less know what MCP is for example or how to run current on-device LLMs or invoke tool use.

I know you're talking about a product, and the device you describe may exist sometime soon-ish but it'll be, ironically, only the wealthy elite (many being in tech) who can afford it. Most housing isn't equipped to add climate control in Europe and in the US there is extreme disparity in what's realistically possible for the average American financially, even with some reduction in costs which are in many cases increasing not decreasing (whole house heat pump systems for example have increased substantially).

I want the future you're describing to exist, I just don't see it happening. I agree with you about climate and I have a lot of thoughts there but in short I'd be elated if we pivot as a society toward equipping everyone.

However I think it's extremely apparent once you look at the raw climate data, updated timelines that the newest vetted research papers show, and the resultant politics around the world that steps have been taken to go in a very different direction regarding how we respond as a global society.

Thank you for taking the time to respond in such a thoughtful manner

They’ll need to do something about profitability soon enough. It’s possible that inference at a company level will be done with a licenced distillation rather than feeding back to the mother brain. I look at how much money Microsoft make licensing Windows and see the room for disruption.

If OpenAI built an ‘GPT OS’ that ran x64 apps but could then let the user interact with them in plain language they’d eat their lunch. To me it makes more sense to create an AI that can run AutoCAD rather than trying to train a model to create the raw data itself (which it will frequently fail at). Why not let somebody else do the hard work for you?

Where Apple will thrive is in the counter-culture market of users that don’t want or like the idea of AI doing things, preferring the human touch. That is a huge marketing opportunity for Apple they’d be stupid to ignore.