capableweb

You don't think already starting with world-class talent (Sutskever, Karpathy, Zaremba and more being part of the founding team) lead to OpenAI being able to get more world-class talent, rather than world-class talent joining because of Altman?

Is OpenAI's current success attributed more to its excellent business and startup management, or does it stem from its superior technology and research that surpasses what others have developed?

Everything points towards this being last minute both for people outside and people inside. Microsoft caught with their pants down, announcement before markets closed rather than just waiting a bit, and so on.

Seemingly, OpenAIs priorities shifted after the public ChatGPT release, and seems to be more and more geared towards selling to consumers, rather than a research lab that it seemed they initially were aiming for.

I'm sure this was part of the disagreement as Sam is "capitalism incarnated" while Ilya gives of much different feelings.

Indeed, but probably kind of relevant as Ilya is one of the few left at the reign of OpenAI, so useful to know what their thoughts are as the board didn't get rid of them.

Unless Google (or someone else) outright acquires GPT4 and future GPT research, it's unlikely that someone will suddenly overtake OpenAI. From what we've seen, no one is close to GPT4, sometimes not even GPT3.

> re the website name (aipanic) and the attitude that the real problem is the people reaction to ai and not ai itself.

The website seems to be about fighting "ai panic" though, not producing/sharing the panic. At least that's how I understand the last sentence:

> It was after this meeting that I decided I wanted to fight the publicly-facing AI hype and doom - even more.

> but IMO it's cringe to use it for consequential official statements like this

This is funny to me, as Twitter is the platform for "deliberately low-effort" posts, but you see it as a platform for official statements. How times change...

So you know that they are rule and norm breakers, no rule too tiny to be broken and skipped in the name of productivity/health/insert-hot-thing-of-the-day

To be fair, we don't really know if OpenAI is successful because of Altman or despite Altman (or anything in-between).

There is a mod for that, unsure if it works with this new release though, but try it out and report back :)

https://github.com/sultim-t/xash-rt

Just like a lot of early 3D PC games weren't possible to play without a graphics accelerator card (GPU for you younglings) which not a lot of people had, these new VR games are not possible to play without VR hardware.

Doesn't mean early 3D PC games weren't PC games, and it doesn't mean that contemporary VR games aren't also PC games.

No, Microsoft is keeping all of that stuff under the wraps. They have a "secret scanning partner program" where they allow companies to have a endpoint GitHub can use for figuring out if something is a secret or not, so it's not just a library with a bunch of regex, seems like a service in itself and Microsoft doesn't really open source stuff like that.

Thank you for attempting to explain it.

So what I don't understand is how "drunk driving ... helps people to get to work on time" is close to true and what I said that made it seem like "impossible to say if it's good or bad" about ChatGPT?

> It doesn't mean that these LLMs are about linguistics and grammar specifically.

Correct. But it does mean they're about "language" rather than knowledge. It's meant to read, parse and generate language, no matter if it includes prose, facts, knowledge or whatever.

> Clearly, they aren't as likely to generate "Colorless green ideas sleep furiously" as "It's important to feed your dog".

Clearly, as the training set probably contains much more instances of the latter, than the first example. And if it's more probable to generate, that's the text it'll generate. It doesn't understand why or how, just that it's more likely, so that's the way it's going.

If you're gonna be pissy about not being able to explain something so others can understand it, why even spend the time writing the comment?

> And if it's "easy" to verify what AIs say, then it should be "easy" to teach AI to verify before it speaks

Sure, if that's the only thing you want the LLM model to be able to do, you could do that. But there are other use cases beyond getting facts, that these models usually also want to support.

Huh? You mind explaining how that applies to my comment, in terms of the things we're actually talking about instead of analogies?

But what if you're unable to even get any sort of answer? It's impossible to verify it, as there is nothing. ChatGPT can help you at least arrive at that first step, in those situations.

Not knowing what you don't know, is really hard to verify. But if you get any sort of answer, it's usually pretty trivial to verify.

Where I found ChatGPT to be really helpful is when I don't know anything regarding something, and ChatGPT can give me something I could at least verify as true or false.

Compared to before, where I just got stuck on some things, as I couldn't find out what I was actually looking for at all.

It seems like the author is missing any sort of nuance. Yes, sometimes it produces bullshit, but often, it doesn't, and can save you a huge amount of time. But yeah, sometimes it leads you down the wrong path or give outright incorrect answers.

How you deal with that is up to you. You can either try to work with what it gives you, setup tooling to easily verify things, and increase your productivity, or you can dig your head down in the sand and say it's all incorrect and you want nothing with it.

Both are valid approaches :)

> I'm not sure what that means, the loan guarantee already seems like quite a risk. Are they going to go for nationalization of the power sector or something like that?

Seems large parts already is nationalized. Vattenfall is a huge power company and is owned by the Swedish state.

The outages you're talking about were at OpenAI, and doesn't seem to have impacted Microsoft's service at all.

Does it matter if the code-execution happens at `pip install` or `python myapp.py`? Using 3rd party libraries inevitably means you're allowing code-execution to 3rd parties, that's the point after all.

Depends widely on scope, complexity, client and consultancy. Example from Trail of Bits regarding blockchain audits:

    +---------+---------------------------+-------------------------+----------------------+
    | Size    | Small                     | Medium                  | Large                |
    +---------+---------------------------+-------------------------+----------------------+
    | Project | ERCs (20, 71, 4626, ...)  | Standalone arithmetic   | AMM or lending       |
    |         |                           | lib                     | protocol             |
    +---------+---------------------------+-------------------------+----------------------+
    | Pricing | $25k                      | $25-50k                 | $50-100k             |
    +---------+---------------------------+-------------------------+----------------------+
    | Timeline| 1 week                    | 1-2 weeks               | 2-4 weeks            |
    +---------+---------------------------+-------------------------+----------------------+

https://www.trailofbits.com/services/software-assurance/

Maybe it depends on the platform? For me, on Windows with Chrome, it only opens the context menu once you've stopped holding down the button. In Blender (on Windows), the menu opens as soon as you hold the button.

I think it's UX that commonly works differently in consumer software vs pro software. I think parent is talking about professional software.

Pro software made for speed and efficiency, you usually want to be able to do things quickly, even if sometimes people not used to the software might screw up. Holding right-click, selecting menu item and releasing I think is one of those things.

In consumer software, users would be confused because maybe they long-press the right-click, drag the mouse a little while holding down then releasing, and the menu would just appear and disappear. Confusing UX for most users, I bet.

> Brooooo how do you even. I wish I had that time it takes to get this good at Blender. I might actually end up releasing some solid indie games.

Keep in mind that the Blender Open Movies are made by professionals who've been doing what they been doing for a long time, and there is a whole team making those, with roles specific to the area.

I don't think you could single-handedly create something like that at the same timescale as they created it. They basically have made a proper studio at this point and they're fine-tuning the workflows and processes of Blender by doing these movies.

So don't feel bad if you never would be able to create this alone, it is a team effort after all.

I do realize this, I didn't mean literally yesterday as in "the day before today", but more "earlier in time".

> Presumably now Netflix will have to pay Broadcom some amount to stream over HEVC.

Presumably, Netflix will in the short-term pay X amount for this but work probably started yesterday or earlier to replace it with something that doesn't have fees to use.