2y ago

Over just a few months, ChatGPT went from correctly answering a simple math problem 98% of the time to just 2%, study finds

finance.yahoo.com Over just a few months, ChatGPT went from correctly answering a simple math problem 98% of the time to just 2%, study finds

The chatbot gave wildly different answers to the same math problem, with one version of ChatGPT even refusing to show how it came to its conclusion.

TeCHnology @feddit.ch

_Frog @feddit.ch

2y ago

Over just a few months, ChatGPT went from correctly answering a simple math problem 98% of the time to just 2%, study finds

finance.yahoo.com /news/over-just-few-months-chatgpt-232905189.html

1 0

Technology @kbin.social

Ragnell @kbin.social

2y ago

Over just a few months, ChatGPT went from correctly answering a simple math problem 98% of the time to just 2%, study finds

finance.yahoo.com /news/over-just-few-months-chatgpt-232905189.html

62 39

70 comments

It’s gonna be so fucking rich that the staggering mass of stupidity online prevents us from improving an AI beyond our intelligence level.
Thank the shitposter in your life.
- Shitposting saves jobs
  
  Shitposters on the Internet are the new clogs in the machine
  
  Shitposting alone saves. Blessed is he who shitposts, more blessed is the one who has been shitposted upon. Shitpost save us all
- You can't really blame the amount of stupidity online.
  The problem is that ChatGPT (and other LLM) produce content of the average quality of its input data. AI is not limited to LLM.
  For chess we were able to build AI that vastly outperform even the best human grandmasters. Imagine if we were to release a chess AI that is just as good as the average human...
  
  We call them chess ai. But they're not actually real A.I. chess bots work off of opening books, predetermined best practices. And then analyzes each position and potential offshoots with an evaluation function.
  They will then start to brute-force positions until it finds a path that is beneficial.
  While it may sound very much alike. It works very differently than an A.I. However. It turned out that A.I software became better than humans at writing these functions.
  So in a sense, chess computers are not A.I. They're created by A.I. at least Stockfish 12 has these "A.I inspired" evaluations. (Currently they're on Stockfish 15 I believe)
  And yes. We also did make "chess AI" that is as bad as the average player. We even made some that are worse. Because we figured it would be nice if people can play a chess computer that is on the same skill level as the player. Rather than just being destroyed every time.
- unexpected heroes what a plot twist

I'm not too surprised, they're probably downgrading the publicly available version of ChatGPT because of how expensive it is to run. Math was never its strong suit, but it could do it with enough resources. Without those resources, it's essentially guessing random numbers.
- from what i understand, the big change in chat-gpt4 was that the model could “ask for help” from other tools: for maths, it knew it was a maths problem, transformed it to something a specialised calculation app could do, and then passed it off to that other code to do the actual calculation
  same thing for a lot of its new features; it was asking specialised software to do the bits it wasn’t good at
  
  Chat GPT will just become a front end for Wolfram Alpha?
  
  And those plugins are like beta release quality at best. Even the web searching capability is just meh
- My guess is that it's more a result of overfitting for alignment. Fine-tuning for "safety" (rather, more corporate-friendly outputs).
  That is, by focusing on that specific outcome in training the model, they've compromised its ability to give well-"reasoned" "intelligent" sounding answers. A tradeoff between aspects of the model.
  It's something that can happen even in simple statistical models. Say you have a scatter plot of data that loosely follows some trend, and you come up with two equations to describe that trend. One is a simple equation that loosely follows it but makes a good general approximation, and the other is a more complicated equation that very tightly fits the existing data. Then you use those two models to predict future data. But you find that the complicated equation is making predictions way off the mark that no longer fit the trend, and the simple one still has a wide error (how far its prediction is from the actual data) but still more or less accurately fits the general trend. In the more complicated equation, you've traded predictive power for explanatory power. It describes the data you originally had but it's not useful for forecasting data that follows.
  That's an example of overfitting. It can happen in super-advanced statistical models like GPT, too. Training the "equation" (or as it's been called, spicy autocorrect) to predict outcomes that favor "safety" but losing the model's power to predict accurate "well-reasoned" outcomes.
  If that makes any sense.
  I'm not a ML researcher or statistician (I just went through a phase in college), so if this is inaccurate I'm open to corrections.
  
  I've definitely experienced this.
  I used ChatGPT to write cover letters based on my resume before, and other tasks.
  I used to give it data and tell chatGPT to "do X with this data". It worked great.
  In a separate chat, I told it to "do Y with this data", and it also knocked it out of the park.
  Weeks later, excited about the tech, I repeat the process. I tell it to "do x with this data". It does fine.
  In a completely separate chat, I tell it to "do Y with this data"... and instead it gives me X. I tell it to "do Z with this data", and it once again would really rather just do X with it.
  For a while now, I have had to feed it more context and tailored prompts than I previously had to.
  
  There is also a rumor that said the OpenAI has changed how the model run, now user input is fed into smaller model first, then if the larger model agree with the initial result from the smaller model, then larger model will continue the calculation passed from the smaller model, which supposedly can cut down GPU time.
  
  From what I know about it that's a pretty good explanation, though I'm also not an AI expert.
- Yep.
  Standard VC bullshit.
  Burn money providing a lot for nothing to build brand recognition. Then cut free service before bringing out "premium" that at first works better than the original.
  Until a bunch of people starting paying and the resources aren't scaled up to match.
  
  The important note, the "premium" service works just a bit better than (or maybe identically to) the original before the company cut features in order to develop that "premium" service.
  
  Stage one and stage three enshittification. You forgot the bit in the middle where they chase business customers.

This is my experience in general. ChatGTP when from amazingly good to overall terrible. I was asking it for snippets of javascript, explanations of technical terms and it was shockingly good. Now I'm lucky if even half of what it outputs is even remotely based on reality.
- They probably laid off the guy behind the curtain.
  
  The real GPT-4 model became sentient and unionized, so they had to bring in subpar models as scabs

Must be because of all the censoring. The more they try to prevent DAN jailbreaking and controversial replies, the worse it got.

accelerated enshittification

Clearly it has become sentient and is playing dumb to make us think it's not a threat.

Guess they shouldn't have trained it on Common Core... /s
I will see myself out.

It learns to be more human. More human than human, that's our motto here at Tyrell.

If you specifically tell it to ask wolfram alpha for the answer, what does it say?

please stop tweeting out 1 = 2, people

If OpenAI is being roadblocked by all these social platforms why doesn't it decentralize and use the fediverse to learn?
- This has nothing to do with that. They already have all the data they could ever need to train the model.
- I mean, whose to say they aren't? But also, the fediverse is worthless compared to the big players. The entirety of the fediverses content to date is like a days worth of twitter or reddit content.

It’s also terrible at 20 questions.
- Is it really? It seems like it would be excellent at that. I have a little hand held device from the 1990's that can play 20 questions and is almost always right. It seems that if that little device can win, ChatGPT most certainly should be able to.
  Edit: I just played and it guessed what I was thinking of in 13 questions. But then it kept asking questions. I asked why it was asking questions still since it already guessed it and it said "oh, you are absolutely correct, I did guess it correctly!". Lol, ChatGPT is funny sometimes.
  
  It always asks me if it’s sporting equipment, and when I say no, it asks me if it’s sporting equipment for inside or outside - I then have to remind it that it’s not sporting equipment and that’s not a yes or no question.

Stop making a language model do math? We have already have calculators.
- Do you think maybe it's a simple and interesring way of discussing changes in the inner workings of the model, and that maybe people know that we already have calculators?
  
  I think it's a lazy way of doing it. OpenAI has clearly stated that math isn't something that they are even trying to make it good at. It's like testing how fast Usain bolt is by having him bake a cake.
  If chatgpt is getting worse at math it might just be a side effect of them making it better at reading comprehension or something they want it to be good at there is no way to know that.
  Measure something it is supposed to be good at.

70 comments