'Why Use ChatGPT:' Author Of 'Black Swan' Says OpenAI's Chatbot Requires Deep Expertise, Makes Mistakes Only A 'Connoisseur Can Detect'

Nassim Nicholas Taleb, renowned author of “Black Swan,” has once again voiced his skepticism about OpenAI’s ChatGPT, stating that the functionality of this AI-powered chatbot comes with a condition. 

What Happened: Over the weekend, Taleb took to X, formerly Twitter, and posted his "verdict" on ChatGPT, stating that the chatbot is only usable if one has in-depth knowledge of the subject. 

He went on to point out that ChatGPT often makes errors that can only be detected by a "connoisseur," citing an example of an incorrect linguistic interpretation.

"So if you must know the subject, why use ChatGPT," he asked.

See Also: How To Use ChatGPT On Mobile Like A Pro

VERDICT ON ChatGPT
It is ONLY useable if you know the subject very, very well. It makes embarrassing mistakes that only a connoisseur can detect. For instance, "bar" is Aramaic for "son", not a Yiddish "pronounciation shift".

So if you must know the subject, why use ChatGPT? pic.twitter.com/U36VCH9DVT

— Nassim Nicholas Taleb (@nntaleb) February 18, 2024

He went on to say that he uses the chatbot for writing "condolences letters" and it fabricates "quotations and sayings."

3) I just got a call from this cyclist friend of mine (who usually asks me to put questions to my X* friends) wants to know if ChatGPT can do powerpoint slides.

* Formerly Twitter.

— Nassim Nicholas Taleb (@nntaleb) February 18, 2024

In the comment section, people suggested Taleb consider ChatGPT as a sophisticated typewriter instead of a definitive source of truth. One person said that OpenAI's AI-powered chatbot is not the "smartest assistant on the planet but you can correct and direct work to move faster."

It is like being given a personal assistant. Not the smartest assistant on the planet but you can correct and direct work to move faster. For tasks like programming, the iterative process works very well. Better than just working by yourself and it will continue to improve.

— Kevin Rossi (@KevinRossi) February 18, 2024

You should not think GPT as a source of truth but as a smart typewriter.

Give it what (context, facts, …) you want to write and ask for edits you would do.

It's a freaking good typewriter that can write, summarize, reorganize, you name it

But it needs *your* input. Always.

— Pierre Alex (@pa_schembri) February 18, 2024

However, some people also agreed with him saying that ChatGPT is "too risky" for certain work assignments. 

Exactly. My work requires extreme accuracy — with stiff legal penalties for getting things wrong.

Far too risky to let LLMs near it

— Superbubble (@TheSuperbubble) February 18, 2024

Why It's Important: This isn’t the first time that Taleb has critiqued ChatGPT’s limitations. 

Last year, he highlighted the chatbot's inability to grasp the ironies and nuances of history. Taleb has also expressed his frustration with the ChatGPT's lack of wit in conversations. 

The same year, it was reported that a lawyer's use of ChatGPT for legal assistance backfired when the chatbot fabricated nonexistent cases. 

However, during the same time, it was highlighted by several reports that not only ChatGPT but other generative AI models like Microsoft Bing AI and Google Bard, now called Gemini, tend to hallucinate and provide even made-up facts with utmost conviction.

In fact, in April 2023, Google CEO Sundar Pichai admitted that AI's "hallucination problems" saying, "No one in the field has yet solved the hallucination problems. All models do have this as an issue." 

Check out more of Benzinga’s Consumer Tech coverage by following this link. 

Read Next: ‘Look What’s Happened’: Elon Musk Questions OpenAI’s Path After Nvidia’s AI Supercomputer Donation To ChatGPT-Maker In 2016

Image via Shutterstock