Imagine that your company has hired a new talent, a rising star in the big leagues so much so that a rival company has hired the same person. The buzz around them is intoxicating. Everyone seems to agree, from the CEO to the shareholders, that this person is the future of the entire business.
Then you learn that the adult has what is called an “intellectual problem(Opens in a new tab)” Every time they open their mouths, they have 15 to 20 chances to create something.(Opens in a new tab). A professor at Princeton calls the man a cow generator(Opens in a new tab). They cannot tell truth from fiction(Opens in a new tab). They are going on stage to unveil a new drug in five minutes. Do you still push them into the spotlight?
For Microsoft and Google this week, the answer was yes. Hot on the heels of the success of OpenAI’s ChatGPT, an Artificial Intelligence chatbot with 100 million monthly users two months after its launch, Microsoft made a surprise last-minute event to announce that OpenAI will bring ChatGPT-style search to the Bing and Edge search engines. browser. Google announced its AI search tool, Bard, yesterday, and unveiled it at an event in Paris the next day – but it ran into a problem of its own.
“A new race begins today,” Microsoft CEO Satya Nadella told reporters in Redmond, Wash. Yes, it’s not nice to think like that(Opens in a new tab)? Microsoft, always the unpopular child of technology, would like you to think that Bing – sorry, “The New Bing” – is in a search competition with Google for everything.
Google’s response to Bard’s announcement(Opens in a new tab) he came down with humility: “We took control of the company six years ago,” wrote Google CEO Sundar Photosi.
Google is a ‘vision problem’
Which is a known fact. Google, the world leader in search, has had many years to integrate AI, and its ChatGPT competitor, Bard, is not at the beta stage with a small group of testers. For all of Fotosi’s hipster preferences, Bard’s reveal had an unexpected twist. Google seems to have been caught off guard by all the ChatGPT buzz as well.
How to explain Bard’s embarrassing mistake in the whole show during the installation – not in the event itself, where some shows are expected, but in a pre-made GIF? A user is shown asking Bard what facts to tell his 9-year-old son about the James Webb Space Telescope.
One of the “facts”, that JWST took the first image of an exoplanet, is not true. Bard was reasonable(Opens in a new tab). (UPDATE: As a Financial Times reporter says The Bard’s words were technically correct(Opens in a new tab)(which requires reading a language that no human can use – which is another problem in AI research.)
Not surprisingly, parent company Alphabet lost nearly 8 percent of its stock on the day Bard launched. Google has put the big problem in AI research front and center, adding that the company can’t use its vast database to find out the facts.
Google should know better, because it already had an “intellectual problem” with its words(Opens in a new tab) at the top of search results back in 2017. The snippets algorithm seemed to particularly enjoy making false statements about US presidents. Again, what would happen?
Never trust a single source: The new rules for learning anything online
In other words, launch your AI search engine too early and you risk playing yourself out. Microsoft was lucky, because nothing wrong was shown at its launch event. But if ChatGPT-based search is not riddled with errors, why is it at such a low level? Side note: If you want to do a free AI QA for Bing, there is a signup sheet.(Opens in a new tab).
“There’s a lot of work to be done there,” Sarah Bird, Microsoft’s Director of AI (a familiar title!) said in response to a question from Microsoft. Wired about the ChatGPT hallucination problem(Opens in a new tab). Yes, no kid: a 15 percent preview came from a company that is on their brand to create a ChatGPT monitor.(Opens in a new tab). (FACTS: a New York Times A breathless report by a columnist on New Bing(Opens in a new tab) (she revealed that it can’t even find good math, or even a child-friendly list.)
Bird added that previous apps could help users plan school shootings, but that the service has been discontinued. Good to know! What could go wrong next? Of course there aren’t any unexpected results lurking in the beta search that could embarrass the giant and not officially recognized.
Clippy. Zune. The new Bing.
Microsoft knows from shame, yes: It is the company that gave us the biggest problem in the history of software, Clippy. The paperclip assistant was famous for giving unnecessary advice. ChatGPT is not Clippy, in the sense that we come with questions.
But the fact that it often checks for answers — or, more often than you might think, gives users a variation of “I can’t answer” — can make ChatGPT-connected Bing a version of Clippy on LSD. If regular users of the “New Bing” get bad results, then that’s what will be recalled.
It doesn’t matter if the thing will be fine later; The first known answer is the one that can turn into a punchline. Microsoft should know that, too; it gave us the Zune. Releasing a ChatGPT object before it is actually ready is no different.
“The New Bing” is already asking to be a punchline, honestly. Or are you ready to give up Google search and your Chrome browser for Bing and Edge, the next one should win the AI search competition, what does “win” mean here? I didn’t think so. Tech inertia is greatly reduced as a power.
ChatGPT is awesome at times – real estate agents especially love it(Opens in a new tab) on lists – writing – and calling out fear in others. But any talk of his interference seems trivial, once you dig beneath the surface. It will lead to a wave of student plagiarism! Except it will also tell you when the paper is written by ChatGPT(Opens in a new tab), disrupting its vulnerability. He passed the law school exam! Except it was just cut by C-plus(Opens in a new tab).
Here’s the thing: creating a digital equivalent of the human brain, known in AI circles as “General AI”, is actually go hard. We have not yet reached the insect intelligence stage(Opens in a new tab), another long-standing goal of AI. Can you really trust ChatGPT to give you results, instead of just clicking links yourself?
The answer may depend on how much you, dear reader, have hallucinations.