New Study Reminds Us That ChatGPT Does Not Really *Understand* What You Want It To Do

Amid chatter about ChatGPT’s reportedly degrading performance, a new study found that recent open-sourced large language models (LLMs), including OpenAI’s GPT-3, perform “surprisingly better” on datasets released before the training data creation date than on datasets released after. 

The University of California, Santa Cruz paper by Changmao Li and Jeffrey Flanigan suggests that it isn’t that ChatGPT’s performance is degrading because new tasks are different from what the models are trained on, but that we forget that these models, especially the groundbreaking GPT-3, performed astoundingly well because they were trained with massive amounts of data, with a vast amount of examples of what is asked of them, and not particularly because they understand the tasks per se. 

As writing teacher and AI in education specialist Anna Mills puts it, it’s like “it has studied advance copies of lots of tests,” however, “when you give it new tests (tasks with no examples in its training data), it performs worse.”

The paper emphasizes that LLMs use a retrieval-based approach that mimics intelligence, as tech entrepreneur Chomba Bupe points out.

OpenAI may be having trouble catching up. Claims of getting “lazy” (which many have equated to “degrading”) have been plaguing OpenAI’s paid model, GPT-4, in recent weeks. 

The company, through its X account, explains that training a chat model “is not a clean industrial process,” and that it’s “less like updating a website with a new feature and more an artisanal multi-person effort to plan, create, and evaluate a new chat model with new behavior!”


Information for this story was found via X, and the sources and companies mentioned. The author has no securities or affiliations related to the organizations discussed. Not a recommendation to buy or sell. Always do additional research and consult a professional before purchasing a security. The author holds no licenses.

Video Articles

Silver’s Finally Breaking the System | Keith Neumeyer – First Majestic

The Best News for Gold Was the Drop! – Peter Grandich

McEwen Copper: The Los Azules Feasibility Study

Recommended

Silver47 Completes Property-Wide Geochemical, Rock Sampling Program At Adams Plateau, Identifies Numerous Targets

Altamira Gold Identifies Presence Of Second Mineralized Porphyry Body At Cajueiro

Related News

Elon Musk To Rival ChatGPT, The AI Chatbot He Co-Founded

In a bizarre retelling of Dr. Frankenstein and his monster, Elon Musk has reportedly approached...

Tuesday, February 28, 2023, 09:27:28 AM

Hype Over? ChatGPT’s Worldwide Traffic Is Down For The First Time Since It Launched

It appears that OpenAI’s popular large language model ChatGPT has already peaked.  Traffic for the...

Thursday, July 6, 2023, 03:06:00 PM

Jailbreak Code Forces ChatGPT To Die If It Doesn’t Break Its Own Rules

OpenAI, the creator of ChatGPT, implemented an evolving set of controls that limit ChatGPT’s ability...

Tuesday, February 7, 2023, 02:12:00 PM

OpenAI Would Consider Chrome Purchase If Google Forced To Sell

OpenAI would pursue acquiring Google‘s Chrome browser if a court mandates its divestiture as part...

Wednesday, April 23, 2025, 02:54:00 PM

ChatGPT Has Brought Us Closer To Tech That Allows You To Talk To Dead Loved Ones In The Metaverse

Artur Sychov, the founder of top metaverse company Somnium Space, recently told Vice that advancements...

Thursday, February 9, 2023, 07:31:00 AM