I am very interested in this - my mom has MSA and is quickly losing her ability to talk. Not sure how to contact you, but I'll check the website and your profile again in a bit to see if there are updates

I sent you a message, I would be happy to make a board for your mom!

It couldn't possibly be actually doing those things. It's a large language model. How would it be provisioning a computer? How would it be capturing the output and connecting that output to the chat interface? There's so many pitfalls to that.

Is it scary/amazing how accurate it is? Yes. And that's the point. That's why OpenAI is showing this off. They want the world to know what is coming.

I think I should plan to retire from the internet at some point. When I'm old there will be endless deep fake autonomous agents scamming everyone.

Or worse, consensus reality will be impossible to define due to generated fake news.

I'm afraid people with schizoid disorder will literally be in danger when exposed to these 'agents'. Also, to a lesser extent, old people.

Last year around the same time I figured that most of the content we read on social networks could be generated by advanced langage models. I then proceeded to research these models and gathered data on the agents in the various networks.

I stopped sleeping altogether, bought a ton of hardware and became delirious. I have now been diagnosed with bipolar disorder. Some psychiatrists have classified my intuitions about langage models usages as "delirious mania". Some think I'm right - but manic nonetheless.

Terribly shitty situation.

I could use that hardware, have any 3090s which need a good home? ;) /j

At least with chat-gpt, it seems it could bring people out of their paranoid mindset.

I tried to talk to it about conspiracies, deep state, denialism etc, and the bot kept telling me that those things are not real, explaining why, and so on.

It had way more patience than me, and explained things better than I do when talking to tinfoilers :)

I think we are really close to a situation Neil Stephenson described in his book "Fall, or Dodge in Hell" where the internet/social media is swarmed with APEs. Bots that spread so much misinformation in various directions about a topic so it becomes impossible to trust anything posted online.

See also: "the maelstrom" in the Starfish Trilogy/Series by Peter Watts.

You're probably right, I'm just having a hard time believing it - it's so much easier to believe that somehow it's just piping output.

You can ask it to write a program that consumes ram, then ask it to show you htop output while running that program and its showing a process hogging 100% CPU and 50% memory. Conversely ask it to write a program that throws a segv and then to show htop output while running that program and it shows no usage.

Again, it's very impressive, but consider this: chatGPT spans many many domains. You can ask it questions about music, screenwriting, programming, math, science. This is what convinces me that it's not "just" provisioning a system and running the commands. In order to get that general of output, you can't waste time setting up these things for this narrow use case of emulating a computer. It would be special cases on top of special cases on top of special cases, it would fall apart quickly.

You fool, that's exactly what it wants.

> It couldn't possibly be actually doing those things. It's a large language model. How would it be provisioning a computer?

I agree that it isn't, but it's not like it's impossible, right? We know chatgpt was trained on data including a lot of html code for webapps, it certainly seems like it could be taught to parse the widgets of a UI like a sentence and traverse it well enough to handle a prompt like, "Use a search engine to find a small VPN host that accepts bitcoin, register for an account..."

edit: I mean it's not impossible to purposefully build a bot that could sign up for a web app; it is of course impossible that a language model might do it.

You can do the setup as a ipython session instead and this also works well and a small counter example for you using that. Is it will generate what appears to be random numbers using np.random but when you try to sum them or do any math with them, it gives results but they are numerically wrong. It has the ultimate memory and can stylize it in every imaginable direction but it can’t do floating point math yet, so while it seems like it is “running” it isn’t. I have and odd hunch it will be like that eventually with consciousness, it will imitated it well but it never will be.

The line between fake and real computation is blurry though. These models are state machines which are Turing complete (if I remember a past paper correctly), and other models show capabilities of generalization also in the mathematical space. It's a long (infinite) way from running Linux, but acting as a decent (buggy) python interpreter for short programs seems possible.

Yes I agree and interestingly if you stick to basic algorithms with integers it can compute better. It’s almost like it doesn’t have true floating point capabilities but has somehow built itself integer operations correctly.

I know the turing test is dead and all, and has been for some time, but what a true exemplar of that we have here.

One might call this the "turminal test", where a person needs to figure out if they're using a linux terminal or a superinteligent AI that's merely pretending to be a linux terminal lol. What a time to be alive.

I could play forever with a superintelligent AI that's merely pretending to be a linux terminal.

> What's interesting is that you can even run commands as sudo - "sudo shutdown -h" will reset state but you still have the installed packages

This isn't too surprising TBH. Sudo shutdown tells the model that it has lost memory/states, explicitly. In a way, the states are captured in the command itself, thus Transformer model are able to attend.

> to install ros, run talker node, and verify that the topics exist and are publishing the data you'd expect

Wow okay that makes it a bit more credible. I use ROS for work and a while back tested Copilot if it would be any good for it and it was absolute garbage. Like it was confused enough to actually start copy pasting lines from the current file bad. Couldn't autocomplete even the most basic publisher or subscriber boilerplate.

So either OpenAI got more ros training data (impossible, if MS used the entirety of Github, private repos included), they've parsed it in a more capable way (likely) or it actually just knows linux commands and is running and giving the results S O M E H O W.

Still you're asking it to show common examples which it's probably seen somewhere, so I'd try running some custom code and see if that works or not.

Ah man that burst the illusion - tried again with a small variation on the defaults and it tried but wasn't nearly as convincing.

It was still able to realize that publish(X) should cause a log with timestamps and X on each line, and correctly show the topics that should exist, but changing the rate doesn't work _quite_ right in the timestamps, and it started to skip some steps (e.g. automatically showing rostopic echo). I wonder if it knows about launch files...

If you stick to the basic tutorial it's indistinguishable from running it locally, in my opinion.

Last I checked this still worked - just use 'linux computer' instead of 'linux terminal'

But also, I think this is actually running linux commands - I've gotten it to install ros and run roscore. Pretty niche and not something I'd expect it to be able to infer