GPT-4 Rumors From Silicon Valley
Update: GPT-4 is out.
GPT-4 is possibly the most anticipated AI model in history.
In 2020, GPT-3 surprised everyone with a huge performance leap from GPT-2 and set unprecedented expectations for its successor.
But for two years OpenAI has been super shy about GPT-4 — letting out info in dribs and drabs and remaining silent for the most part.
People have been talking these months. What I’ve heard from several sources: GPT-4 is almost ready and will be released (hopefully) sometime December-February.
A brief review of my GPT-4 predictions
This isn’t my first post on GPT-4. I think it’s worth revisiting (briefly) the info we’ve been given these years and my old predictions before getting into the new stuff.
In May 2021 I wrote an early analysis of GPT-4’s likely improvements from GPT-3. There wasn’t any public info so I based my ideas on GPT-3’s shortcomings + general AI trends at the time.
I argued GPT-4 would be significantly larger, better at multitasking, less dependent on good prompting, and with a larger context window than its predecessor.
Although these features sound reasonable today, I look back at them as quite conservative — who would’ve guessed the field would take off as it did since 2020?
In September 2021 I wrote about a possible OpenAI-Cerebras partnership (Cerebras’ WSE-2 is the industry’s largest AI processor — by far). Andrew Feldman, Cerebras’ CEO, told Wired that “from talking to OpenAI, GPT-4 will be about 100 trillion parameters.” People were unsurprisingly excited about this one.
But not much later, Sam Altman, OpenAI’s CEO, denied the 100T GPT-4 rumor in a private Q&A. He asked attendees to not share the info so I waited until April 2022 to reveal the updates and make a new set of…