It doesn't seem like you're addressing the HGP example, which at the time, would have been an "individual procedure" as well.
> If an API call can cause a disaster then fix the API By "API" I'm not referring just to publicly facing REST endpoints. I mean things like shell access for system maintenance, that normally only human professionals…
Not sure if you meant this comment rhetorically or not, but try asking ChatGPT to do something useful like convert some text into JSON, or what conjugation of a verb you should use, or for writing feedback on an email.…
I mean that the AI is what's being attacked. It's likely that backend LLM agents will have access to sensitive non-public APIs.
What is the difference? The original prompt isn't what gets used anyways.
> Is prompt injection even a problem worth worrying about? It depends what API access the AI has. If it's just a chat bot, prompt injection can only reveal facts about its language model. But if the AI has POST access…
> But for most usecases, online models trained by megacorps still win. Whisper and whisper.cpp have gotten us close to the tipping point.
You can ask it why yourself. That might have something to do with it.
You might like some of the work being done under the label "Factored Cognition". It's an approach that treats LLMs as building blocks instead of being complete AIs. Instead of asking the LM to solve a problem directly…
> A few errors show quickly there is no such concept being weilded I would have given similar examples to show that ChatGPT makes the same kinds of mistakes that humans do. The first one is good, because ChatGPT can…
Looks like IVF costs around $12,000 today. Considering that this is for your own child, I bet it would be one of the least expensive parts of raising them. I also expect it would become very cheap after it went…
"statistical re-mixer" doesn't describe these systems very well. I see this complaint a lot, that supposedly DL models can only manipulate existing content without creating anything of their own. That's just false,…
What would be evidence that a prediction machine had developed a theory of mind?
I'm always confused. The NWO conspiracy people say that gene editing will be prohibitively expensive (like owning a yacht is), but the real life examples I see don't seem that way. What are some good examples of…
> and it's not pretty. It feels somewhat ironic that when I looked up "Fisherian Runaway" the first thing I saw was an image of a peacock. Evolution by "persistent, directional female choice" seems like exactly what our…
> Reality, nor even driving, is not anywhere close to that. This reminds me of the kinds of intuitions people had about Chess and Go. In retrospect they seem silly, but it made plenty of sense to them at the time. The…
> It'll just go "lol I encountered 3 486 864 games with this particular board pattern and 49 652 lead to a victory lemme pick one for the lulz". That isn't really an accurate description of how the modern DL bots work.…
I think that's already been done: https://en.wikipedia.org/wiki/MuZero
> And then the slippery slope, “we can also improve the odds of higher intellect, being taller , thinner, etc”. I would love to have been born in a world that had slid to the bottom of that particular slope.
> It requires a theory of mind, which means understanding human goals. Like chess?
> endless progress or a good future The optimist believes progress will halt at a convenient moment. The pessimist fears otherwise.
> Artists will need to get real jobs again like the rest of us and fund their art as a side project. You said you were surprised that the creative jobs are being automated first. Are you expecting the "real jobs" to…
"It's a cook book!" hasn't yet entered the Overton window.
It doesn't seem like you're addressing the HGP example, which at the time, would have been an "individual procedure" as well.
> If an API call can cause a disaster then fix the API By "API" I'm not referring just to publicly facing REST endpoints. I mean things like shell access for system maintenance, that normally only human professionals…
Not sure if you meant this comment rhetorically or not, but try asking ChatGPT to do something useful like convert some text into JSON, or what conjugation of a verb you should use, or for writing feedback on an email.…
I mean that the AI is what's being attacked. It's likely that backend LLM agents will have access to sensitive non-public APIs.
What is the difference? The original prompt isn't what gets used anyways.
> Is prompt injection even a problem worth worrying about? It depends what API access the AI has. If it's just a chat bot, prompt injection can only reveal facts about its language model. But if the AI has POST access…
> But for most usecases, online models trained by megacorps still win. Whisper and whisper.cpp have gotten us close to the tipping point.
You can ask it why yourself. That might have something to do with it.
You might like some of the work being done under the label "Factored Cognition". It's an approach that treats LLMs as building blocks instead of being complete AIs. Instead of asking the LM to solve a problem directly…
> A few errors show quickly there is no such concept being weilded I would have given similar examples to show that ChatGPT makes the same kinds of mistakes that humans do. The first one is good, because ChatGPT can…
Looks like IVF costs around $12,000 today. Considering that this is for your own child, I bet it would be one of the least expensive parts of raising them. I also expect it would become very cheap after it went…
"statistical re-mixer" doesn't describe these systems very well. I see this complaint a lot, that supposedly DL models can only manipulate existing content without creating anything of their own. That's just false,…
What would be evidence that a prediction machine had developed a theory of mind?
I'm always confused. The NWO conspiracy people say that gene editing will be prohibitively expensive (like owning a yacht is), but the real life examples I see don't seem that way. What are some good examples of…
> and it's not pretty. It feels somewhat ironic that when I looked up "Fisherian Runaway" the first thing I saw was an image of a peacock. Evolution by "persistent, directional female choice" seems like exactly what our…
> Reality, nor even driving, is not anywhere close to that. This reminds me of the kinds of intuitions people had about Chess and Go. In retrospect they seem silly, but it made plenty of sense to them at the time. The…
> It'll just go "lol I encountered 3 486 864 games with this particular board pattern and 49 652 lead to a victory lemme pick one for the lulz". That isn't really an accurate description of how the modern DL bots work.…
I think that's already been done: https://en.wikipedia.org/wiki/MuZero
> And then the slippery slope, “we can also improve the odds of higher intellect, being taller , thinner, etc”. I would love to have been born in a world that had slid to the bottom of that particular slope.
> It requires a theory of mind, which means understanding human goals. Like chess?
> endless progress or a good future The optimist believes progress will halt at a convenient moment. The pessimist fears otherwise.
> Artists will need to get real jobs again like the rest of us and fund their art as a side project. You said you were surprised that the creative jobs are being automated first. Are you expecting the "real jobs" to…
"It's a cook book!" hasn't yet entered the Overton window.