5 ms·
As others suggested, this is likely a summer vacation dip, but speaking for myself as a non-student, my GPT (4 + 3.5) usage has plummeted. If you use it in a "
by shrimpx 3y ago
As others suggested, this is likely a summer vacation dip, but speaking for myself as a non-student, my GPT (4 + 3.5) usage has plummeted.
If you use it in a "what should be done" or "what is the right thing" sense, you get useless generalities. "It is important to remember that ... <there are many sides to the situation>". Suppose for example that you talk to it about combating judicial bias. It will largely insist that judges are unbiased because they're supposed to be unbiased according to their job description... You won't walk away with any insights from engaging it in ethics/morals topics.
If you dive into depth with some subject, it will invariably make errors and contradict itself in the details. I dove into how electricity works and it was saying and enthusiastically confirming electrical energy is encoded in electrons' kinetic energy. Then it started claiming that electrical energy is definitely not in the kinetic energy of electrons, that electrons don't lose speed when they bump into atoms, and instead the energy is "in the field". Ok... YouTube it is.
It's at least less convenient than a search engine at routine fact-finding due to its knowledge cutoff, inability to display images and links, etc. Ditto with translation. It's good but so is Google Translate and Google Translate has in-line suggestions for misspellings, a menu of languages, etc.
Ok, it's pretty good at writing hilarious poems, imitate people ("in the style of", etc.). Friends and I sometimes text each other hilarious things GPT said. But even there the machine is sort of laid bare, and you can see its creative limits. Its poetry is ultimately pretty lame, basically, and gets predictable. It's like an insistent and repetitive 8 year old with a lot of knowledge to pull from. It's not something I do often.
It's good at writing boilerplate code but honestly I just don't reach for that use case often. I like to stay in my own flow rather than constantly delegate to a robot and check its work, which I find flow-breaking, especially if it doesn't understand my APIs and makes mistakes often.
I do, sometimes, engage in architectural discussion about how I should design something at a high level. It's full of mistakes but bouncing ideas around can be useful. But I only feel the need to do that once in a great while.
These issues are exacerbated by being forced to downgrade to 3.5 after using 4 for a bit. You might feel like you're getting somewhere with 4, and then 3.5 starts laying doozies on you. This is another UI/product issue, but affects usability.
- IanCal 3y ago> Suppose for example that you talk to it about combating judicial bias. Ok, here you go. https://chat.openai.com/share/b083b98d-6904-4c85-94fc-abdc76a22038 https://chat.openai.com/share/b083b98d-6904-4c85-94fc-abdc76... A quick discussion of where things are "now", some proposals, a selection of three and then ethical concerns about implementing the proposals. I propose a method of increasing diversity and have it critique it.
- ChatGTP 3y agoI did a Google search for the same topic and personally found the top link full of much more useful and practical information. YMMV but a high level summary like that isn’t necessarily what people want. You can probe the bot harder but why bother when I could just Google and read a little? It’s actually less effort for better information in some sense.
- IanCal 3y agoI'm not arguing here what I did was better than a search, but that it's not some forbidden information. I often feel like complaints about what you can do with these things come from not knowing how to use them. It did also take virtually no effort. A benefit is being able to ask questions though, you can see I proposed a specific idea and had it reply - you can delve more or less into that. Don't want a high level summary? Ask for more details on the part you're interested in. Want a higher level one or simpler one? Ask. As things get less direct this gets better. I talked to it about projects for my son, narrowed down through different interests and got some great recommendations. One involved building a radio, which it then could explain at several levels of detail for children. That gets into "hope someone made a list and then go through loads for relevant things, lookup more detail from them then find several different resources for kid friendly explanations".
- shrimpx 3y agoBtw I didn't mean to imply it's forbidden information. It's just not that good/impressive/sharp. It often just doesn't inspire going back for more. Also like I said, I find it useful for software design. It ultimately makes a bunch of mistakes but I find the conversation useful and I walk away with usable ideas. I don't often start new software, so I don't go to it often for this. I think it's useful at the high level stuff; giving you ideas and laying out options, like the chat log you posted. But as you dive into detail I find two things happen: (a) It starts to make egregious errors. (b) You bias it with your more detailed questions. When I talked to it about electricity, it started contradicting itself about how energy is carried, whether by electrons' kinetic energy or by "the field". I asked it "so is electrical energy encoded in electrons' kinetic energy or not?" and then it started to say that oh yes, that's how it's encoded. It seems it used my question to reinforce that knowledge which previously it was confused about. Since I personally don't know how electrical energy is carried, I lost confidence that I'm learning correctly. This happens often. "I'm sorry about the confusion. You are right. ..."