bauldursdev

If you don't verify identity it's easy to fake.

Sounds really cool, were you able to verify the correctness of the results?

Why wouldn't inference get less expensive? The frontier will probably always be expensive but imagine if in 5-10 years Fable 5 is considered a low tier antiquated model, it is still somewhat capable.

They do, but unethical people don't announce themselves. In fact an ethical person is probably more likely to admit their faults, which usually doesn't play well in elections.

From a software POV, I feel like it makes it easier to implement stuff. Whether that stuff is good or bad. If you know what you're doing, vet the output, and use it properly, you can get a nice productivity boost while still producing good code. If you don't know what you're doing, you are prone to go down rabbit hole after rabbit hole of unwise decisions.

Is there anything that would point to this possibly being the case?

For it to fix the bug it has to identify the bug. If the bug is a security vulnerability then it will have to identify the security vulnerability to fix it. What's the alternative, have it ignore vulnerabilities/bugs? It wouldn't be a very good coding companion in that case.

I'd pay less attention to the prompt and more attention to the output when interpreting this story. (I'm not saying I agree with the decision, but this is how they are looking at it.)

Unless by next door you mean a different continent I think they'd still be in Europe. Although there are edge cases, such as if the fab is turkey but the office managing it is in a neighboring central Asian country.

I think they said increasing false positives because it would make it easier to generate at a mass scale. IDK the merits of the argument or what exactly they're saying would be done, but imagine pre-AI it might take someone quite a bit of manual effort to manufacture a plausible document regarding nuclear developments, but with AI it doesn't require so much work and is easier.

Can't solve a problem that doesn't exist yet

To me it seems like it's more likely to refuse the harder the problem is. I wonder if it's cover for a model that's not as good as advertised. Even when I ask questions in biology it is switching me.

No I'm paid to write code.

There's such a spectrum between "give it everything" and "give it nothing". Imagine you just want to use it to code and want to make sure any commands it runs doesn't mess up your actual machine.

Ya if you just run VMs but now a lot of cloud vendors have you build stuff using their proprietary features so decoupling becomes harder and can be risky.

No, it's like, if you could buy 100 things before, but you can only buy 96 things now, then you have accumulated less value :D

A human just predicts the next best action.

I don't know how it is in Europe but when I worked at restaurants, it was just part of the managers job to bring the cash to the bank. It took 20-30 minutes, we'll say 1 hour if it's busy. The bank didn't charge a fee to deposit the cash. Pay was like $20/hr, let's say 50% overhead so $30 to deposit the cash. The frequency was just about weekly, and would be into the thousands. Let's say it's only $1,000. That's 3%. And they generally did it when there was down time anyways, so I think it's wrong to attribute 100% of that to depositing cash, since it would be an expense regardless.

But this is just my anecdotal experience, do you have actual numbers/statistics on this?

Because information asymmetry benefits those with the information. If the devil understands your argument, and you don't understand the devil's argument, the devil will have information advantage.

It's not just mind either. I know at least one person whose around 60, still very smart actually, but has diminished vision and hearing. So less able to pick up the obvious signs.

I think they mean limited like it might be limited in performance or some other characteristics. :D

I haven't read the paper but it seems like it's saying rude prompts are better, so isn't it reasonable to assume that's what they meant? If we want to talk about directness, that's kind of a tangent right? I see directness as an entirely different dimension, you can be very direct and polite, you can be very rude and indirect (e.g. passive aggressive). Maybe they should do a follow-up study on how well AI responds based on level of directness.

Ya my experience is that many people honestly don't produce output as good as AI. An educated (formally or informally), experienced person who is putting forward good effort is better than AI, but I do know people who honestly just produce results having AI do it for them.

Yeah it's the same thing. No shortage of uninformed opinions on each side (not like I know WTF is going to happen, but at least I don't go around presenting my uninformed beliefs as fact).

No point? There certainly is a point. The point is that I don't sound like AI now. Sure, one day, some AI lab might build models so that everyone has a model that's indistinguishable from them, and at that point, this method won't work.

But saying it's not worth it is like going back in time and telling a peasant to stop harvesting their crop manually because in 10 years someone will invent the tractor. I get trust of others now, which is something that is useful now. Any benefits I accumulate don't get undone.

Unless you mean they'd target me specifically... which I find very unlikely.

Maybe worth looking into other providers. I've never had issues disputing charges. Granted I dispute maybe 1% of the transactions I make, because I only do it if I legit got scammed. I think paying for a month and not getting a month is a scam.

I know, I could've worded my comment clearer. I meant that I specifically introduce them in a way to make it obvious my text was written by me.

I despise when someone just passes my prompt to an AI, but I do honestly think that there are a minority of people who do better work with it. Not that the work is good, but they don't care to try and at least the AI is eager.

Agree wholeheartedly. I have actually started introducing small idiosyncrasies into my text to make it clear that my words come from me and not a bot.

This is much much more complex than a traditional program, which can be followed line by line. Trying to understand every bit of the literal logic is like trying to understand a person by thinking about the neurons fired in their brain to make them say or do something.

Unfortunately you have to learn to let go, and say, "I'll never be able to keep this all in my head", and learn to think about it in terms of of the outputs/inputs and how you can create a model capable of efficiently modeling your problem and how parameters can be nudged to get an output which is kinda shaped how you want.

Maybe some really genius savant could keep it all in their head but I doubt it, like I said it'd be like trying to understand a person by reasoning about their neural pathways.

Yea, isn't a lot of software already developed like this? I doubt the lead developers of Chrome know the implementation details of every component of chrome. They know the high level architecture and leave a lot of the specific implementation to various teams.