Microsoft Launched Bing Chatbot Despite OpenAI Warning It Wasn't Ready 23
According to a report from the Wall Street Journal, the partnership between Microsoft and OpenAI has become "awkward" due to tension and confusion. Ars Technica reports: Not only has this tension and confusion extended to Microsoft's internal AI team -- which apparently is dealing with budget cuts and limited access to OpenAI technology -- but sources said it also clouded Microsoft's controversial rollout of AI-powered Bing search last February. At that time, Bing was found to be vulnerable to prompt injection attacks revealing company secrets and providing sometimes inaccurate and truly unhinged responses to user prompts. According to WSJ, OpenAI warned Microsoft "about the perils of rushing to integrate OpenAI's technology without training it more" and "suggested Microsoft move slower on integrating its AI technology with Bing." A top concern for OpenAI was that Bing's chatbot, Sydney, might give inaccurate or unhinged responses, but this early warning seemingly was easily ignored by Microsoft. In a Wired interview published today, Microsoft CEO Satya Nadella suggested that any hiccups with Sydney at first were just part of Microsoft's plan for training the chatbot to respond to real-world prompts that couldn't be tested in a lab. "We did not launch Sydney with GPT-4 the first day I saw it, because we had to do a lot of work to build a safety harness," Nadella told Wired. "But we also knew we couldn't do all the alignment in the lab. To align an AI model with the world, you have to align it in the world and not in some simulation."
So that's partly why Microsoft rushed ahead anyway, but sources told WSJ that the rush was also partly due to Microsoft executives who had "misgivings about the timing of ChatGPT's launch last fall." Because OpenAI started ChatGPT's public testing while Microsoft was still working on integrating OpenAI tech into Bing, tension seemingly spiked between the partners, who also stood as rivals in an AI race to capture the world's attention. As ChatGPT's success grew, some Microsoft employees raised concerns that ChatGPT was stealing Bing's "thunder," WSJ reported. Others sensibly posited that Microsoft could learn valuable lessons ahead of Bing's rollout from ChatGPT's early public testing. [...] Of course, ChatGPT ultimately won the AI race, instantly attracting the fastest-growing user base in history. Meanwhile, "the new Bing," released a month later, "has yet to come close to the breakout success of ChatGPT," WSJ reported. Citing data from analytics firm YipitData, WSJ reported that ChatGPT "has nearly double the average number of daily search sessions as Bing search does."
Further tension and confusion has brewed within Microsoft's in-house AI team, which has "complained about diminished spending." Most employees are set back by a lack of "access to the inner workings" of OpenAI's technology, which is particularly painful for employees attempting to integrate that tech into various Microsoft products. There's also the awkward reality that OpenAI's and Microsoft's sales teams "sometimes pitch the same customers." Much of this "drama" amounts to typical infighting that happens any time two companies pair up, WSJ reported, but there's no ignoring the conflict inherent to both sides attempting to maintain independence while reaping maximum profits by selling access to the same technology. Despite these tensions, Nadella told Wired that OpenAI "bet on" Microsoft, and Microsoft "bet on" OpenAI. He still envisions "a good commercial partnership" between the independent companies and considered Microsoft's investment in OpenAI as "a long-term stable deal." Increasingly, it looks like one way to assuage tension is to bring the companies even closer together in partnership. WSJ noted that Nadella announced last month that the Bing search engine would soon be integrated into ChatGPT, which he said was "just the start of what we plan to do with our partners in OpenAI to bring the best of Bing to the ChatGPT experience."
So that's partly why Microsoft rushed ahead anyway, but sources told WSJ that the rush was also partly due to Microsoft executives who had "misgivings about the timing of ChatGPT's launch last fall." Because OpenAI started ChatGPT's public testing while Microsoft was still working on integrating OpenAI tech into Bing, tension seemingly spiked between the partners, who also stood as rivals in an AI race to capture the world's attention. As ChatGPT's success grew, some Microsoft employees raised concerns that ChatGPT was stealing Bing's "thunder," WSJ reported. Others sensibly posited that Microsoft could learn valuable lessons ahead of Bing's rollout from ChatGPT's early public testing. [...] Of course, ChatGPT ultimately won the AI race, instantly attracting the fastest-growing user base in history. Meanwhile, "the new Bing," released a month later, "has yet to come close to the breakout success of ChatGPT," WSJ reported. Citing data from analytics firm YipitData, WSJ reported that ChatGPT "has nearly double the average number of daily search sessions as Bing search does."
Further tension and confusion has brewed within Microsoft's in-house AI team, which has "complained about diminished spending." Most employees are set back by a lack of "access to the inner workings" of OpenAI's technology, which is particularly painful for employees attempting to integrate that tech into various Microsoft products. There's also the awkward reality that OpenAI's and Microsoft's sales teams "sometimes pitch the same customers." Much of this "drama" amounts to typical infighting that happens any time two companies pair up, WSJ reported, but there's no ignoring the conflict inherent to both sides attempting to maintain independence while reaping maximum profits by selling access to the same technology. Despite these tensions, Nadella told Wired that OpenAI "bet on" Microsoft, and Microsoft "bet on" OpenAI. He still envisions "a good commercial partnership" between the independent companies and considered Microsoft's investment in OpenAI as "a long-term stable deal." Increasingly, it looks like one way to assuage tension is to bring the companies even closer together in partnership. WSJ noted that Nadella announced last month that the Bing search engine would soon be integrated into ChatGPT, which he said was "just the start of what we plan to do with our partners in OpenAI to bring the best of Bing to the ChatGPT experience."
All board the hype train! First adopter wins all! (Score:4)
Because ChatGPT will turn anything it touches into pure unicorn magic!
Big whoop (Score:1)
Title says it all.
Ok but (Score:5, Informative)
Maybe this article isn't the place for this discussion but
Why do people presume that there will ever be a time when LLMs are not subject to prompt injection?
The belief that these systems will eventually be what people imagine them to be in narrative fiction runs throughout writing on the technology. It suggests the majority of people writing about it have absolutely no idea how it functions.
Yes by the way, I know what I'm saying is obvious to people who do understand how it functions.
Re: (Score:2)
Why do people presume that there will ever be a time when LLMs are not subject to prompt injection?
I’m not sure, but it might be related to all the prompting in the information they consume.
Re: Ok but (Score:3)
Man, turns out deploying commercial tech based on algorithms that are not fully understood to the market was a really bad idea huh.
We basically deserve the upcoming AI winter.
Re: (Score:2)
current generation of LLMs can't really avoid prompt injections. The models themselves have to have their own moral compass of sorts, to avoid being tricked into saying stuff that "they" wouldn't normally say...
e.g. you don't wanna train a LLM on Mein Kampf and have it come out a Nazi, you want it to be able to consume Mein Kampf and still maintain its views... once we have that capability (model's own personality or moral compass), then prompt injecting will be less of an issue...
...that said, once we give
In other words (Score:2)
Microsoft experiments on its users.
Beats google (Score:5, Insightful)
Sure, it's not perfect. No webpage is, and no search engine or chatbot is. Getting bored with the hysteria, such as this story. There is no clear-cut metric by which it would be "ready."
Re: (Score:2)
I'm failing to see where the hysteria is in this article. I read the whole thing and some terms I'd use to describe it are: level-headed, reasoned, apparently factual, and even dry.
Do you know what hysteria means, and can you point to any part of the article that fits? You seem to be under the impression it means "something I'm not interested in".
Re:Beats google (Score:5, Insightful)
One major factor is Google has regressed... a lot.
For "long tail" queries (in other words, non common ones), it tries to help and fails making my job much more difficult.
Quotation marks, AND/OR operators, plus/minus signs don't work most of the time. When they do, they usually return no usable results. However it would gladly transform my query into a common one, thinking I made a mistake:
Compared to that Bing / ChatGPT combo usually becomes more helpful. Not always, but most of the time, when I query niche stuff, GPT does better.
Re: (Score:2)
R.E. "Google has regresses". Yeah, remember the good old days when (from the "Why use Google?" page twenty-four years ago.)
Google only returns pages that include the terms you type.
Unlike many other search engines, Google only produces results that match all of your search terms, either in the text of the page or in the link anchors pointing to the page. No more frustration with results that have nothing to do with your search terms.
Those were the days. FWIW, I personally find "verbatim" mode does make it d
Re: (Score:1)
Google basically ruined their own value by putting in "too much" advertising and selling page position "too much".
I can't define "too much" but I know I haven't trusted google results for a couple years now. And even to the degree I do trust them, they are much less useful (more noise less signal) than they used to be. Pure greed is killing their golden goose.
Re: (Score:1)
So it's not like they didn't know!
Oh dear (Score:4)
Nice partner (Score:2)
Not only will OpenAI compete with their customers, they will throw them under the train for media points. Only Apple should consider a major contract with OpenAI, because in a media slapfight they'd win.
Microsoft should cut their losses.
Released Early? (Score:1)
Oh really? (Score:2)
Nadella made the right call (Score:1)
ChatGPT technology is the first thing to come along that can disrupt the search market and Bing needs to wield it against Google.
There's always people saying "it's not ready." Often it's more important to be first (learn the lesson of the DEC Rainbow).
Re: (Score:1)
NIH (Score:2)
MS has an internal NIH problem.
Microsoft would do *that*?! (Score:3)
PC Magazine, 1995, reported that M$ was about to introduce its new o/s, Windows95, with over 64,000 known bugs or issues....
Re: (Score:3)
In Microsoft's defense, the x86 architecture at the time limited them to 65,535 bugs. Later, IBM increased this limit to 262,144 bugs when they released OS/2.