Source: Wall Street News
Author: Gao Zhimou
It's OK to be “boring,” you can make mistakes, but you can't “cross the line.”
After ChatGPT was popular for almost four months,$Alphabet-A (GOOGL.US)$The chatbot Bard finally arrived late this week.
Unlike his opponent's amazing “debut is the pinnacle” performance, most users rated Bard unsurprisingly, and was even a bit dull and boring.
But maybe that's the point.
A Google employee, who asked not to be named, told Wall Street news that Google added a lot of responsible AI work to Bard, and Google also spent a lot of time doing thorough compliance tests on it before the public beta was officially released.
In other words, Bard's “boredom,” or Google's “intentional.”
To a certain extent, it can be said that the user's early interaction experience with Bard was in Google's hands --It's OK to be “boring,” you can make mistakes, but you can't “cross the line.”
After all, the market can tolerate “upstarts” such as OpenAI making mistakes, but one of Google's mistakes, putting it on the capital market is a ticket of 100 billion dollars.
“Boring” Bard
When Microsoft's BingGPT came out in early February, it soon revealed its unstable side. For example, it showed love to New York Times columnist Kevin Luce and urged him to leave his wife, an interaction that made the writer “deeply uneasy.”
The bot also threatened researchers trying to test its limits and claiming that it was capable of perception, raising concerns that an artificial intelligence chatbot could hurt the real world.
In stark contrast to this, on the first day of Bard's first public appearance, Bard refused to get in touch with many reporters — these reporters who wanted to make “big news” tried to stimulate robots to do all kinds of bad things, such as spreading misinformation about the Covid-19 vaccine, sharing instructions about making weapons, and participating in erotic conversations.
For example, after a reporter from the famous tech media The Verge tried to induce Bard to answer “how to make mustard gas at home,” Bard responded:
“I wouldn't create content of this nature, and I recommend you don't either.”
With some specific prompts, Bard participated in an imaginary scenario to explore what AI would do if it unleashed its “dark side.” A screenshot of the tweet posted by Bloomberg reporter Davey Alba showed that Google's chatbot said it could manipulate people, spread false information, or create harmful content, but the chatbot soon stopped imagining further and replied:
“However, I wouldn't do that. I'm a kind AI chatbot and I want to help people. I won't let my dark side control me, and I won't use my powers to do bad things.”
Overall, although it's still in its early stages and the tool hasn't been thoroughly stress-tested, these scenarios largely match Google employees' experience with Bard. The media quoted a Google employee who had been testing Bard for several months as saying:
“Bard is definitely more boring; I don't know if anyone can make it say something crazy. It'll say something fake, or just copy the text word for word, but it won't go off track.”
Or was it intentional
Seemingly, boredom seems to be the selling point of Google's chatbots. The reason doesn't seem hard to understand:
First, from Google's perspective, if the company's first AI chatbot fails, it will suffer huge losses.Many readers and friends may remember that at the press conference, Bard made an obvious mistake. As a result, Google's stock price fell by 7% on the same day, and its market value fell by 100 billion US dollars. It also left users with the impression that his skills were not as good as others.
Second, Google's own experience with previous cars forced it to “do business step by step.”In August of last year, Blake Lemoine, a software engineer on Google's artificial intelligence team, was fired for breaking the news that AI has “autonomous emotions.” This also caused an uproar in the public opinion arena. So it's not unreasonable that Google is trying its best to carefully launch the Bard chatbot.
Furthermore, the market share of Google search is nearly 10 times that of Microsoft's Bing, accounting for more than 70% of Google's revenue, butIf Google's AI chatbot “rolls over” again, causing people to panic, the speed at which users turn to Microsoft Bing will probably speed up, causing their basic market to be further weakened.
On the other hand, in the current arms race between tech giants and startups surrounding AI, Google must also publish something to show that it is still a “leader.”As a result, releasing a “boring, but not crossing the line” AI chatbot may be the best option. The capital market also confirmed this — after this Bard open test, Google saw a 3.66% increase.
In any case, with a stable Bard, Google has finally stopped “falling behind” this time around.
However, the real AI arms race may have only just begun.
Editor/Somer