travis-weber.bsky.social
Software Developer from Arizona
www.travis.gg
35 posts
8 followers
15 following
Getting Started
Active Commenter
comment in response to
post
Fair enough, I concede GPT misled me here. Perhaps it is worse with history than I thought.
Another problem might be my leading question, "Was paros abandoned due to piracy in the 1600s?"
Either way, I will trust it much less in the future.
comment in response to
post
Experts will give better answers than LLMs, but professors only reply to cold-emailed questions <50% in my experience.
LLMs give instant answers and are always accessible via internet. Not an end-all-be-all, but another great tool for learning!
And not all papers can be found in the local library!
comment in response to
post
Feeble analogy. Some things are better than others.
iPhone has more capabilities than a flip phone. GPT is more capable than Gemini. Grok3 might be better than both.
comment in response to
post
Source?
comment in response to
post
Sure, if you are a full-time academic then read all day. Perhaps that is why we have such different perspectives.
I am a full-time software dev. The LLMs give me working code, speeding up my workflow. In the spare time I have to explore curiosities, they speed up my learning.
comment in response to
post
Yeah, Gemini sucks. You gotta judge the models individually though.
They are improving quickly. There is room for improvement across the board.
comment in response to
post
But I did not block you sarah, nor anyone else. Not sure if I am the "he" you are referring to.
comment in response to
post
I was blocked by the original poster under which we were talking earlier, and a couple of the people I was talking to, too.
Pretty strange! Not used to getting blocked
comment in response to
post
It's similar to a librarian or a search engine. Ask a question and then for sources, and you get a pretty good starting place for research.
Example: chatgpt.com/share/67b4fc...
It would be hard to get a person to give this much of an answer, and they would not be as fast.
comment in response to
post
Yes, it's best to ask for links to sources so that you can go read the original. In this way, GPT is best used in the same way as Wikipedia.
comment in response to
post
Montessori is the way
comment in response to
post
It's okay to not be convinced! No need to mute, sorry if I'm annoying.
Wikipedia is great too! And it is risky business to trust AI blindly, as your example shows.
Tools can be used well or poorly, and for good or bad.
AI-literacy means knowing how to use AI well. It will rise over time!
comment in response to
post
There is no human that I could have asked that would have given such a thorough set of links and responded as quickly to these questions as GPT:
chatgpt.com/share/67b4b2...
Not enough information to write a research paper, but the best way to scratch a curiosity and be a bit more knowledgeable.
comment in response to
post
Librarians are not always knowledgeable on a topic, and professors only respond to cold email questions at about a 40% rate.
ChatGPT is probably the greatest advancement in mass access to information since Google.
Here is another conversation where GPT helped me out:
chatgpt.com/share/67b4b2...
comment in response to
post
Check out this conversation I had with GPT last week: chatgpt.com/share/679ee4...
Using GPT was the fastest path to this set of information. A lot of the papers mentioned are not stored in a library, and hard to find via only a keyword search on Google Scholar.
It's not complete, but a start.
comment in response to
post
Not being prepared to fact-check an AI is a problem.
Everyone will get better at using them with experience.
AI literacy is set to create a great advancement in both learning and productivity!
comment in response to
post
We aren't disagreeing much. Everything you say is true.
My degree is in math so I am over-prepared to care about the semantics.
LLMs are still very useful for learning even though they give the occasional false answers, it usually leads to more discussion. Asking GPT for sources works well
comment in response to
post
This article elaborates a bit.
travis.gg/posts/scienc...
comment in response to
post
It's not that hard, but it is harder than using LLMs. LLMs are another tool in learning, and should be celebrated as such!
If you don't believe what it is saying, then you can ask for sources. GPT pulling sources is often better than google.
comment in response to
post
When the brakes on your car don't work, serious injury is risked. Do not use AI if serious injury is at risk.
However, in low-consequence situations such as drafting code, the speed that LLMs introduce is valuable.
Also, which person does not sometimes answer questions wrong?
comment in response to
post
Yes. Wait actually...no.
Which of these two was not an answer? One is incorrect, but they are both answers.
comment in response to
post
Also, when you say "steal", what is it you don't have access to anymore?
Art and literature are often meant to inspire other minds, the only new thing here is that some are now artificial.
comment in response to
post
LLMs are a much faster way to learn than libraries. Responses happen in fractions of a second rather than hours. Both are useful, but LLMs are very helpful and hopefully improve in the future.
comment in response to
post
Although this is a hard claim to defend these last few weeks....
comment in response to
post
This post showed up in a search for "replicators", meaning Bluesky runs some photo-to-text model.
comment in response to
post
One of the most worthwhile podcast episodes I've heard in a bit:
open.spotify.com/episode/6beC...