OpenAI CEO admits ChatGPT risks. What now? | The AI Beat

Over the weekend, OpenAI CEO Sam Altman suddenly jumped into the Twitter fray around ChatGPT, the company’s recently-released conversational text-generation model, with a surprisingly firm note of caution: 

“ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness,” he tweeted on Saturday. “It’s a mistake to be relying on it for anything important right now. It’s a preview of progress; we have lots of work to do on robustness and truthfulness.” 

The thread concluded with a note that for ChatGPT, “fun creative inspiration; great! reliance for factual queries; not such a good idea. We will work hard to improve!” 

OpenAI grapples with ChatGPT hype

What prompted Altman’s comments? After all, when OpenAI released ChatGPT on November 30, while he did caution that it was an early demo and research release with “a lot of limitations,” he also hyped up future applications: “Soon you will be able to have helpful assistants that talk to you, answer questions, and give advice,” he tweeted. “Later you can have something that goes off and does tasks for you. [E]ventually you can have something that goes off and discovers new knowledge for you.” 

Altman’s latest cautionary notes probably aren’t responding to last week’s AI Beat column. More likely, they emerge as a result of the past week-and-a-half of massive hype — and drumbeat of criticism. So far, OpenAI’s ChatGPT has been described as everything from a “sensation” and “the most disruptive technology since [fill in the blank]” to a “world-class bull**** artist” and “kind of like that drunk guy or gal you meet at the bar who never stops talking, blathers on and on with an engaging combination of facts and random bullshit, but that you’d certainly never want to take home to your parents.”

Perhaps, most notably, Altman is seeing the effects of OpenAI’s research demo hitting the mainstream with a wallop. The New York Times, the Wall Street Journal, the Washington Post, the Atlantic and even Fox Weather have covered it in just the past few days.  

What happens now with ChatGPT? 

For now, it seems the ChatGPT horse is fully out of the barn and running down Main Street. There are zero signs of its popularity slowing down: In fact, OpenAI seems to be having trouble keeping up with capacity. Some have reported receiving notes saying “Whoa there! You might have to wait a bit. Currently we are receiving more requests than we are comfortable with.” 

But while the dizzying pace of discourse around ChatGPT continues — from those pointing fingers at Google for supposedly lagging behind in LLMs, to concerns about the future of college essays — it seems those developing these models are keeping their heads down, aware of the fierce competition ahead. 

On the OpenAI side, it appears clear that the company is using this period of widespread community experimentation with ChatGPT to get RLHF — reinforcement learning from human feedback — for a highly-anticipated future release of GPT-4.

Of course, while Stability AI CEO Emad Mostaque says exactly that, he also represents the other side of the coin: those rapidly working around the clock to produce an open-source variant of ChatGPT. 

LAION, one of the creators of Stable Diffusion, says it is already actively working on that:

So, we wait. In the meantime, I promise that none of the above text was written by ChatGPT. I’m holding my own — for now.