Same. I still don’t know the social script in these scenarios. Do I individually thank everyone for their birthday wishes? Do I thank everyone with one message after it seems like everyone’s done? How long do I wait? What if someone jumps in after I do that with a belated happy birthday?

I went the friend-to-romantic-partner route, so it was pretty much just like hanging out with any other friends.

I’ve had the opposite experience. The older the pasta, the longer it takes to cook. If it cooks faster, that suggests to me that it has absorbed water during storage, which allows harmful bacteria to grow and the pasta might actually be expired.

I’ve never found Bing chat to match up with the free ChatGPT. It often just refuses to answer my question while ChatGPT will at least take a guess and give me something to work with.

Quickly filtering out a subset of them to prioritize so that we get the most value possible out of the time that humans spend on it.

LLMs cannot:

• Tell fact from fiction
• Accurately recall data from its training set
• Count

LLMs can

• Translate
• Get the general vibe of a text (sentiment analysis)
• Generate plausible text

Semantics aside, they’re very different skills that require different setups to accomplish. Just because counting is an easier task than analysing text for humans, doesn’t mean it’s the same it’s the same for a LLM. You can’t use that as evidence for its inability to do the “harder” tasks.

Sounds to me like a 50% improvement over zero human eyes.

It certainly would be. Thankfully, there’s many more than zero human eyes involved in this.

Considering that it’s a language task, LLMs exist, and the cost, it’s a reasonable assumption. It’d be pretty silly to analyse a bag of words when you have tools you can use with minimal work with much better results. Even sillier to spend over $200 for something that can be run on a decade old machine in a few hours.

Perfectly fine tool, but they should not be used when you’re being evaluated on your ability to do arithmetics.

They do not have sales data, so they use two different proxies: number of reviews, and number of active players.

I don’t see mention of how they get the number of active players. I’m assuming it’s through stats in Steam or something similar. If that’s the case, then their assumption of this number being biased towards being larger than the true number would be wrong. If you choose to both pirate and buy the game, chances are good that you’ll be playing the pirated version, and therefore would not get counted towards active players.

Same here. That’s why I try to stock up on more than I think I’ll need when things go on sale.

Why not both?

I can’t tell if you’re trying to say Alpine skiing is scary or that you’re into all the stuff people consider to be extreme sports.

My online persona is definitely different from IRL, and it differs IRL depending on who I’m interacting with. But these are all the real me. My ability to communicate via text is generally better than spoken, so that is reflected in how I write, what I write about, as well as how little I speak in person.

Secondly, in person communication has clearer continuity. If I have multiple conversations with a given person, I learn a bit about them and their communication style, allowing me to adjust how I speak to be better understood by that person. Online, I rarely remember who I’m talking to, so I just write in whatever way feels most natural to me.

The real time nature of in person communication also limits what you can bring up and when. Anything you say requires the other party to respond immediately, and if you recognize that they’re not in the mood to think particularly hard, then you don’t bring up difficult topics. Online conversations don’t come with this kind of information, but it does give you the flexibility to answer whenever you want, or not at all, so many things that I would not deem acceptable in an IRL setting can be acceptable online.

So in summary, different situations do call for different behaviours. But that’s not problematic any more than behaving differently at a party and at a funeral is problematic.

If you’re blending it up into a powder anyway, wouldn’t it make more sense to add the paprika at the end? Does adding it before baking actually make a difference?

You mean at the tail end of a thread that opened with me pointing at the environmental costs?

Exactly! Hence my confusion. If you care about energy costs, then shouldn’t saving energy be a good thing? Why would the benefit be 0?

Weren’t you just telling me that the environmental cost has no impact on your stance?

It sounds like you don’t like how LLMs are currently used, not their power consumption.

I agree that they’re a dead end. But I also don’t think they need much improvement over what we currently have. We just need to stop jamming them where they don’t belong and leave them be where they shine.

Yeah, they operate very opaquely, so we can’t know the true cost, but based on what I can know with certainty given models I can run on my own machines, the numbers seem reasonable. In any case, that’s not really relevant to this discussion. Treat it as a hypothetical, then work out the math later to figure out where we want to be and what threshold we should be setting.

Indeed. Though what we should be thinking about is not just the cost in absolute terms, but in relation to the benefit. GPT-4 is one of the more expensive models to run right now, and you can accomplish very good results with their smaller GPT-4o mini at 0.5% of the energy cost^[1]. That’s the cost of running 0.07 LED bulbs over an hour, or running 1 LED bulb over 0.07 hours (i.e. 5min). If that saves you 5min of time writing an email while the room is lit with a single LED bulb and your computer is drawing energy, that might just be worth it, right?

[1] Estimated by using https://huggingface.co/spaces/genai-impact/ecologits-calculator and the pricing difference between GPT-4o, 4o mini, and 3.5 (https://openai.com/api/pricing/). The assumption I’m making is that the total hardware and energy cost scales linearly with the API pricing.