Google employees slam CEO for “dumpster fire” response on ChatGPT

Enlarge / Google’s headquarters.

When Google’s ChatGPT competitor event was announced for last week, we wrote that it seemed like a rush job to reassure investors, and since then this event has happened and gone worse than anyone could have imagined. Google’s event did the opposite of what it intended, with the stock down nearly 12 percent from its recent high just before the event. Even Google employees are beginning to take notice, as CNBC’s Jennifer Elias writes that employees are internally criticizing CEO Sundar Pichai for what they call a “rushed, botched” announcement of Google’s new chatbot.

CNBC says it was able to view several messages from Google’s internal “Memegen” employee forum, and while these are typically light-hearted, the report states, “Posts following the Bard announcement took a more serious tone, even going straight to Pichai. “

“Dear Sundar,” one post began, “The launch of Bard and the layoffs were rushed, botched and short-sighted. Please return to a long-term perspective.” Another highly upvoted post read, “Sundar and leadership deserve a Perf NI. They’re oddly short-sighted and un-Googleish in their quest for ‘grain focus.’” A “Perf NI” refers to the lowest employee performance rating of “Needs Improvement,” as if someone were evaluating Pichai’s performance as CEO.

Google announced the unveiling of a ChatGPT competitor last Wednesday, but then made the bizarre decision to spoil that announcement with a blog post two days earlier. Apparently the reason was that Microsoft had already sent out invites for this Tuesday’s ChatGPT-powered “New Bing” launch, ahead of Google’s announcement by a day, and Google wanted to get ahead of the frontrunners.

The two events were intertwined, but could not have been more different. Microsoft announced the “New Bing” on Tuesday and launched a product. It has a lot of rate caps at the moment, but Bing + ChatGPT is out in the wild where it will try to compress search results into a readable paragraph and answer questions. It’s also built into the Microsoft Edge browser, where it can answer questions, help with content authoring, or summarize a page. It’s a real product for people to try.

Google’s Monday announcement of its “Bard” chatbot only gave an overview of Google’s planned features, with details so vague that anyone could have written them (it’s a chatbot within Google Chat that answers questions!). Google’s post included a sample answer, and that answer turned out to be incorrect, earning it an embarrassing rectification article in Reuters. The blog post has since been updated with a new example, but it clarified the AI’s tendency to generate plausible-sounding but incorrect answers.

Google’s event did not include a public launch of its chatbot technology, instead opting for a private “Trusted Testers” program. It seemed like the company was struggling to throw this entire event together, so of course it wasn’t ready to launch anything just yet, but the back-to-back events only showed how far behind Google was. Unlike Microsoft, Google’s event was mostly filler material, with just five minutes of the 40-minute event devoted to its ChatGPT competitor, and almost all of that information was already covered in Monday’s spoiler blog post. The remainder of the event covered minor updates to Google Translate, Maps and Lens, including a section that had to be skipped after a presenter told the crowd that Google had “lost the phone” it needed for the presentation. The live stream ended with an error message and was marked “private” on YouTube, making it unplayable for several hours.

According to the CNBC report, one contributor wrote that “Bard’s rushing into the market in a panic confirmed the market’s anxiety around us.” Another posted an image of a dumpster fire with a Google logo on it, saying it was representative of “how everything has felt since last year.” Another slammed how the company seems to be stock-focused lately, saying, “When 12,000 employees are laid off, the stock goes up 3%, a rushed AI presentation takes it down 8%.”

The really embarrassing thing for Google is that it invented the key technology that underpins ChatGPT. The “GPT” in ChatGPT stands for “Generative Pre-trained Transformer,” where a “transformer” is a neural network architecture invented and open-sourced by the Google Brain team in 2017. Back then, Google described Transformer as “particularly good for language understanding,” but never created a product using the technology. ChatGPT’s owner, OpenAI, is turning AI research into products like ChatGPT and DALL-E that people can actually use while Google locks them in a lab The two events, one with and one without a product launch, only reinforced that narrative last week.

Leave a Reply

Your email address will not be published. Required fields are marked *