> Myself and the OP are more worried about the other path: a dystopia in which the majority of people are forced into something much worse than wage slavery by those in control of the thinking machines.
My fear is that nobody will remain in control of the thinking machines. Imagine an AI agent for hire which maintains its own cryptocurrency accounts and pays its own cloud hosting bills. That's the future I'm worried about.
My cynical prediction: Google search (as we now know it) will be retired. Google will provide AI-generated text which attempts to answer the user's prompt, with links to results in a handful of curated sources such as wikipedia.
I already use GPT for this purpose, and it works terrifyingly well. You can even ask it for citations, and it provides you with a link.
Beware--it's prone to bullshitting. This one time I asked it for the symptoms for pulmonary hypertension: shortness of breath, chest pain, lightheadedness, and fainting. Check. Then I asked it for a citation...
"This study showed that patients with hemorrhoids displayed a 27% increased risk of CHD after adjustment for the confounding factors. Risk factors for the development of CHD are more common among subjects with hemorrhoids than those without hemorrhoids except for diabetes mellitus."
I’ve started doing this already. Just last week I had a question about how to interpret labeling on construction wood. Google came up with nothing for SEO spam. GPT3 explained it completely.
Their fields of research, such as large language models and image generation and the newly demoed “multi search” feature all strongly point towards this being the case.
>Google search (as we now know it) will be retired.
I think so too, but I think it's still pretty far out.
In order to do this, Google's AI-generated text needs to be better than the web results they return today, 100% of the time. It's the same reason why Google has never made 100% of the first 10 results ads. If users make a search and advertisers aren't bidding on the keyword, Google will return 0 results, and the user is less likely to come back and make a new search.
Google's AI needs to consistently be better than the content created by webmasters. Across billions (trillions?) of different search phrases.
Maybe there is an intermediary step where Google uses AI-generated text in a small subset of searches, and then expands out gradually. But there is so many unique searches it seems like a tall task.
Also, if Google displaces webmasters with AI-generated content, what incentives do websites have to create the new content Google will need to train their AI on to answer whatever humans want to know over the next 50 years?
How would this work with new information, e.g. breaking news? Would Google be manually putting in anything new that comes up or would they have the legacy Google crawler running in the background to feed to the AI system?
- To center a div horizontally on a page, simply set the width of the element and the margin property to auto. Or, you can use WebSiteBuilder® which simplifies it for you. Use 'this' affiliate link for 10% off in the first month.
second year:
there are 25 different ways of centering a div with potential fallbacks for older browsers and user technologies, please consider WebSiteBuilder® for all your div centering needs. Otherwise click here to read about the 25 different ways to use, text guaranteed 86% accurate.
If you had to fight for your right to equality, would you not find it unacceptable that someone in such a prominent position profited from the work of others and invested it into a hate group? Why not simply let people be? Gay marriage hurts nobody.
My argument is that opposition to gay marriage isn't intrinsically hateful. The ancient Greeks had no notion of gay marriage, for example, and they weren't exactly averse to gay sex or to men who enjoyed it.
Your system of values, beliefs, and definitions may create a tunnel vision which makes it look like opposing views can only be motivated by hatred or bigotry. But the reality is, other people have different views because they hold different values, beliefs, and definitions. If you cared more about the preservation of traditional culture, and not so much about the peculiar way that homosexuality is expressed in modern western societies, then your views would presumably be in line with those of Eich.
I've never understood what exactly the "open" in OpenAI is supposed to imply. They produce proprietary, gated models - not open in any meaningful sense.
Didn't work on the desktop? I've been running Linux exclusively on my desktops for 15 years. Desktop Linux is better today than it's ever been.
Why does every OS need to be suitable for average consumers? Librem and Pine64 are doing great work in the mobile Linux space on the hardware side, and projects like PostmarketOS are doing great work on the software side. These are niche products for motivated enthusiasts, as they should be. They'll never grow to billions of users, nor should they. The tech industry's "grow massive or the product is worthless" mindset is pathological, in my opinion.
Not only I've been successfully running it on my desktops for about that long, but I also have been running it on my phones for not much shorter. It works.
> Desktop Linux is better today than it's ever been.
I honestly have to strongly disagree with that. Desktop Linux is more complex, more fragile, and more choice-less than 10-15 years ago. It might have a bit better hardware compatibility... but even back then it would run on just about anything.
> Librem and Pine64 are doing great work in the mobile Linux space on the hardware side
But can you actually use them as a phone? Can you use all (or even any) of the apps that people want to use?
> They'll never grow to billions of users, nor should they. The tech industry's "grow massive or the product is worthless" mindset is pathological, in my opinion.
Without that, they'll also never get the things that people want their phones to do.
> Desktop Linux is more complex, more fragile, and more choice-less than 10-15 years ago.
More complex, sure. Bare ALSA was simpler than Pulse/PipeWire. But the benefit of the additional complexity is that nowadays, sound just works. PipeWire gives us flexible audio routing pipelines like Jack, but for all audio applications rather than just pro audio! And as a counter-example, the X11 to Wayland transition considerably reduces the complexity of the display stack.
I'm not sure what you're getting at with "more fragile", but I definitely disagree with more choice-less. Sure, I wouldn't want to run modern Gnome on a full-size desktop/laptop (we have MATE, Xfce, and countless minimal WMs/compositors for that). But modern Gnome is great on tablets. Non-systemd and non-glibc distros are out there if you're into that sort of thing. Flatpak, etc are available if you're into that - if not, nobody's forcing you to use them.
> But can you actually use them as a phone?
Yes, with effort and some sacrifices. The velocity is in the right direction - the situation is much better than it was a year or two ago.
> Without that, they'll also never get the things that people want their phones to do.
I agree that there needs to be a sufficiently large community that things get developed. But this doesn't mean we need billions (or even millions) of users.
They aren't all perfect, but TBH, I would get by. I only miss Threema to switch fully at this point (which I've migrated many of my contacts too, so that's on me), but as work on a Matrix-bridge is ongoing there's a lot of hope. (I could also just use Waydroid + the Android app to fill this gap, but… that kills the battery (at least for the PinePhone).
FWIW battery hungriness of Waydroid should be perfectly fixable. I had a quick look and it doesn't really seem to care about not doing useless work when there's nothing to do and happily wakes up multiple times per second all the time the container is running. A lot of low hanging fruits there.
Counter-argument: DALL-E is smart enough to understand that an astronaut riding a horse makes more sense than a horse riding an astronaut, and therefore assumes that you meant "a horse-riding astronaut" unless you go out of your way to specify that you definitely do, in fact, want to see a horse riding an astronaut.
Because intelligence is more than frequentism. Being able to predict that a dice lands on a given side with probability 1/6 is not a demonstration of intelligence. It feels a bit naive to even suggest this and I suspect that you aren't a ML researcher (and if I'm right, maybe don't have as much arrogance because you don't have as much domain knowledge).
My fear is that nobody will remain in control of the thinking machines. Imagine an AI agent for hire which maintains its own cryptocurrency accounts and pays its own cloud hosting bills. That's the future I'm worried about.