Nothing is below the line for me to start my inquiry with LLMs. The bigger question is where is the line for whether I follow up with my own verification. There's two reasons for me not to verify: 1) I don't care enough, or 2) I'm confident enough in the answer.

The latest thing that I asked ChatGPT about in which I didn't care to verify because I don't care, not because I was confident in the answer, is explaining to me why Europeans don't do college sports like Americans do, and why their sports youth pipelines come in the form of professional academies instead of college athletics.

Any time I need to figure out the subtle ways that libre office differs from the Microsoft suite, i'm happy to make chatgpt my little bitch.

I use AI to research things I care about, because I care about too many things to spend huge amounts of time on it. If you're paying for bleeding edge models they are more efficient at researching basically anything, given that you are steering them.

We are now habitually using it in cooking, gardening, personal finance, developing policy proposals for local political groups, many things.

It's kind of like online dating back in the days before tinder, when you had to pretend you didn't have a profile because it was considered something for desperate losers. Ironically online dating created way more desperate losers than were on it back then when it was uncool. Probably something similar will happen when the bubble pops and the economics make the current delivery models unsustainable, when OpenAI embraces the ad model and Anthropic gets acquired by Google and subsumed by it's ad services.

Like the internet itself this productivity window might be brief, supported by venture capital in order to overcome behavioural barriers only to sucker punch the consumer like social media.

Interesting research problem haha. I personally never ask an LLM to solve a problem end-to-end because it never satisfies me, so I ask much smaller things. Is it possible that the problem statement is too broad and that's why it didn't come up with anything?

Anyway, I mostly use it to search stuff for me, like I need to get background on something I know but don't remember the source for, so I don't really do things like these. The last one I did was to have it figure out for me why there were differing stories about Starlink capabilities for monitoring usage in Africa.

Sounds like this is also the reason why we can't have home Bitcoin miners that cool the house.

I haven't done much of this kind of research. Looks like it could be fun to mess around with.

Honestly, so far with my limited experience, it's mostly helped me out kick my coverage. I like to learn, and it's a pretty powerful tool in my opinion.

Una de las cosas que me pongo a preguntar es como hacer cerveza o vino, siempre quise aprender a hacerlo, pues por no saber las medidas termino convirtiendolo en vinagre, y lo más loco veo bricolajes en Youtube para luego hacer mis intentos, me supero supongo 🤣💪

Yeah, same here AI is great for probing the edges of curiosity without the time sink it’s less about answers and more about cheaply learning why something doesn’t exist.

I once asked GPT chat who would win in a fight between Kirby and sun wukong, Kirby is a monster